Next Article in Journal
A Comprehensive Methodology for the Statistical Characterization of Solar Irradiation: Application to the Case of Morocco
Previous Article in Journal
Voice Interaction Recognition Design in Real-Life Scenario Mobile Robot Applications
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Applied Sciences
Volume 13
Issue 5
10.3390/app13053363
applsci-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Neural Image Analysis for the Determination of Total and Volatile Solids in a Composted Sewage Sludge and Maize Straw Mixture

by
Sebastian Kujawa
1,*,
Gniewko Niedbała
1,
Wojciech Czekała
1 and
Katarzyna Pentoś
2,*
1
Department of Biosystems Engineering, Poznań University of Life Sciences, Wojska Polskiego 50, 60-627 Poznań, Poland
2
Institute of Agricultural Engineering, Wrocław University of Environmental and Life Sciences, 37b Chełmońskiego Street, 51-630 Wrocław, Poland
*
Authors to whom correspondence should be addressed.
Appl. Sci. 2023, 13(5), 3363; https://doi.org/10.3390/app13053363
Submission received: 4 February 2023 / Revised: 28 February 2023 / Accepted: 3 March 2023 / Published: 6 March 2023
(This article belongs to the Section Agricultural Science and Technology)
Browse Figures
Versions Notes

Abstract

:
Waste management is one of most important challenges in environmental protection. Much effort is put into the development of waste treatment methods for further use. A serious problem is the treatment of municipal sewage sludge. One method that is useful for this substrate is composting. However, it is reasonable to compost a sewage sludge mixed with other substrates, such as maize straw. To carry out the composting process properly, it is necessary to control some parameters, including the total solids and volatile solids content in the composted mixture. In this paper, a method for the determination of the total solids and volatile solids content based on image analysis and neural networks was proposed. Image analysis was used for the determination of the colour and texture parameters. The three additional features describing the composted material were percentage of sewage sludge, type of maize straw, and stage of compost maturity. The neural models were developed based on various combinations of the input parameters. For both the total solids and volatile solids content, the most accurate models were obtained using all input parameters, including 30 parameters for image colour and texture and three features describing the composted material. The uncertainties of the developed models, expressed by the MAPE error, were 2.88% and 0.59%, respectively, for the prediction of the total solids and volatile solids content.

1. Introduction

The volume and variety of the waste generated are increasing every year [1]. Waste associated with the municipal sector plays a crucial role. It is produced every day by each inhabitant of the planet. One such type of waste is municipal sewage sludge, defined according to Polish law as “sludge from digestion chambers and other installations for the treatment of municipal sewage and other sewage with a composition similar to that of municipal sewage” [2]. Sewage sludge management is challenging. Any action taken should consider the legal, environmental, social, and economic aspects. Sludge is a problematic material that must be managed in a legal manner [3]. In many countries around the world, sewage sludge is landfilled, although this is increasingly limited by regulations. It is therefore necessary to look for other options for their management. One solution may be to use sewage sludge in the reclamation of degraded and devastated areas. This issue has been addressed in reports by Bęś et al. [4] and Halecki et al. [5]. Another solution for managing sewage sludge is to use it as a fertiliser in agriculture. Research on this topic has been carried out by Salinitro et al. [6], Ragonezi et al. [7], and Wydro et al. [8].
One method of municipal sewage sludge management is composting [9,10]. It is one of the simplest and most popular methods [11]. However, due to the fact of its structure and chemical composition, it is unreasonable to use sewage sludge as a monosubstrate. Therefore, it is mixed with a suitable structuring material that, on the one hand, increases the porosity of the composted mixture and, on the other hand, improves the ratio of the C:N [12,13]. Such a material could be maize straw, which is the residue left over from the cultivation of maize for grain, as well as other types of straw, e.g., wheat, paddy, rapeseed, or sugarcane. However, under Polish conditions, maize straw has far fewer potential customers than straw obtained from the cultivation of other cereals or oilseed rape. Therefore, it is particularly interesting for use as a structuring material.
When sewage sludge is composted properly, there is significant heating of the composted biomass. This promotes pasteurisation and the destruction of pathogens [14,15]. The material is assumed to be properly hygienised when its temperature is maintained at a level of at least 55 °C for at least 1 day. Alternatively, it should maintain a level of at least 70 °C for at least 1 h [16,17]. When considering the agricultural use of compost from sewage sludge, it is crucial to ensure that the heavy metals content is sufficiently low. Suitable compost for this purpose can usually be produced on the basis of sewage sludge from municipal sewage treatment plants located in areas without excessive industry. In addition, it is important to check that the compost produced does not contain pathogens so that it is safe to use for fertiliser purposes [18].
In recent years, a considerable amount of attention has been paid to the study of biomass composting processes involving sewage sludge. Studies have been aimed at gaining a better understanding of these processes and optimising them regarding the minimisation of the time required to obtain a final product of suitable quality [19,20,21,22,23]. This research requires the implementation of experiments, which are often carried out at the laboratory scale using specialised bioreactors [15,24]. During such tests, several physicochemical parameters of the composted material are monitored, as well as the concentrations of oxygen and carbon dioxide in the air exiting the bioreactor and the emissions of ammonia, hydrogen sulphide, and methane [25,26]. Based on these parameters, decisions regarding the termination of a composting process may be taken at the right time, i.e., when the aerobic decomposition process has slowed down considerably.
Modern IT methods are increasingly being used to study the composting process [17,22]. Methods based on computer image processing and analysis in combination with neural modelling [27,28,29] are widely used in solving classification and prediction problems in many different fields, including life sciences [30,31,32,33]. Nevertheless, they are rarely used in studies of composting processes. In previous studies, these techniques have been successfully employed to identify the early maturity stage of a composted mixture of sewage sludge and maize straw [16] and sewage sludge and rapeseed straw [34]. A particular type of neural network (convolutional neural networks (CNNs)) was applied to the identification of cosubstrate composted with sewage sludge [35] and to the maturity classification of composted sewage sludge and rapeseed straw mixtures [17].
For reaching the early maturity of the composted material, in addition to the temperature, parameters such as the total solids (TS) and volatile solids (VS) content in the composted mixture are important. The measurement of these parameters is usually carried out during aeration of the composted biomass mixture [36]. There are studies on the possibility of using spectrometric methods to measure the moisture content of compost [37]. However, this type of analysis requires expensive laboratory equipment. In reports by Zaborowicz et al. [38] and Wojcieszak et al. [39], a preliminary analysis of the possibility of using image analysis and neural modelling to determine the content of total solids and volatile solids in compost is presented. Unfortunately, the results obtained should be regarded as preliminary research due to the small number of learning cases in relation to the number of inputs of the neural models. Therefore, it was decided to extend the research on the use of neural image analysis to determine the physicochemical parameters of the composted material. The aim of this study was to develop neural models for determining the total solids and volatile solids content in a composted mixture of sewage sludge and maize straw based on the information extracted from images of the composted material. Image analysis and neural networks are useful tools for research into the composting process. Their use can help in analysing the selected parameters of the composting process and in controlling it, which was conducted as part of this research.

2. Materials and Methods

2.1. Research Material

A composted mixture of sewage sludge and maize straw was used as the test material. The sludge came from the municipal sewage treatment plant located in Szamotuły (Greater Poland Voivodeship, Poland), while the straw was obtained from a branch of the Swadzim Agricultural Experimental Farm located in Złotniki (Greater Poland Voivodeship, Poland). The composting processes were carried out under controlled conditions at the Ecotechnology Laboratory (Poznań University of Life Sciences) using a dedicated bioreactor [15,24,26,40]. Two types of maize straw were used in the study: untreated and ensiled. A total of eight composting experiments lasting between 29 and 39 days were carried out (Table 1). The values of the total solids content were determined using the standard PN-EN 14346:2011 [41], while the values of the volatile solids content were determined using the standard PN-Z-15011-3:2001 [42].

2.2. Image Acquisition

The samples of composted material were subjected to image acquisition in a specialised photographic chamber illuminated with visible light (Figure 1). The construction of this chamber was described in detail by Kujawa et al. [43]. As a light source, fluorescent lamps, Sylvania Luxline Plus F15W/865 (Feilo Sylvania, Budapest, Hungary), were used. The images were acquired using a Nikon D80 digital single-lens reflex camera (DX format sensor), equipped with a fixed lens Nikkor 35 mm f/1.8G AF-S DX (Nikon Corporation, Tokyo, Japan). The exposure parameters were set according to the rules of photography, taking into account the intensity of the light illuminating the photographed sample. The ISO number of the camera’s sensor was set to ISO100, the aperture was set to f/5.6, and the shutter speed was set to 1/25 s. A total of 1536 images of composted material were acquired with a resolution of 968 × 648 pixels covering an area of 98 × 65 mm. The 576 images were of a composted mixture with untreated straw and 960 with ensiled straw. At the same time, 865 images were of immature material, while 672 were of material that had reached the early maturity stage. Example images taken at the beginning and the end of each composting experiment are shown in Figure 2. It is worth noting that the composted material filled the entire frame of the photograph. Therefore, there was no problem of separating the material from the background at a later stage of the research.

2.3. Image Processing and Features Extraction

Each of the acquired images was extensively analysed, resulting in values for 25 colour and 5 texture parameters. Some of these parameters were obtained from the images in their original form (24-bit JPEG format). However, to obtain some of them, the following transformations were performed:
  • Image conversion from a 24-bit RGB colour space model to an 8-bit grey scale (256 grey levels); the brightness of each pixel was determined as the weighted sum of the R, G, and B components according to the following relationship:
GS = 0.2989R + 0.5870G + 0.1140B,
  • Greyscale image binarisation using 4 threshold values: 0.05, 0.1, 0.15, and 0.20;
  • Conversion from the RGB model to the HSV model.
The greyscale images were used in the texture analysis process. For each image, a grey-level co-occurrence matrix (GLCM) was created. Eight pixel brightness classes were included, and the neighbourhood was considered as one pixel, symmetrically along the 4 main directions: 0, 45, 90, and 135°. A full list of the obtained colour and texture parameters is presented in Table 2.
The colour and texture parameters were extracted from the acquired images in an automated manner. For this purpose, the Compost Image Analysis software (version 2.0, Sebastian Kujawa, Poznań, Poland), developed in the MATLAB programming language, was used. The formulas presented in [16] were used to determine the texture parameters.
Based on the extracted image parameters, as well as additional information on the composted material, 10 datasets were developed for the training of neural models to determine the total solids content of the composted material. For training of models to determine the volatile solids content, the separate 10 datasets were prepared. The additional information on the composted material was as follows:
  • Sewage sludge percentage;
  • Maize straw type (0—untreated, 1—ensiled);
  • Compost maturity stage (0—immature compost, 1—compost at early maturity).
An integral part of the datasets was the expected output information of the neural network on the total solids or the volatile solids content determined during the laboratory tests. Each dataset was composed of 1536 independent cases and was divided in a 2:1:1 ratio into training (768 cases), validation (384 cases), and test (384 cases) subsets. The datasets varied in terms of the input parameters (Table 3).

2.4. Neural Models Development

Based on the datasets detailed in Table 3, neural models were developed to determine the content of TS and VS in the analysed material. The models were created in the MATLAB computing environment using the Deep Learning Toolbox. As a neural network, the multilayer perceptron (MLP) with ten neurons in the hidden layer and one neuron in the output layer was employed. MLP is one of the most popular feedforward neural network topologies [44]. It has a high degree of flexibility and is widely used to solve a variety of classification and regression problems. Its advantage over more complex networks, such as CNNs, is its much simpler structure. As a result, it is not necessary to have huge amounts of data to train (determine weights and biases) this network. One hidden layer and the number of neurons in this layer were chosen on the basis of the authors’ experience and previous preliminary analyses. In the authors’ experience, a good starting point for similar regression problems is usually between 5 and 30 neurons in the hidden layer. A sigmoidal function was taken as the activation function in the hidden layer and a linear function in the output layer. The number of inputs to the MLP depended on the number of input parameters in the dataset. The networks calculated a single output in terms of total solids and volatile solids content. The models were trained using a back-propagation algorithm in the form of Bayes regularisation (BR). The mean square error (MSE) was used as the error function. The maximum number of training epochs was set at 1500. The training process was stopped earlier if there was no reduction in the value of the network error function with respect to the validation set within 100 consecutive epochs.

3. Results and Discussion

The information on the neural models developed to determine the total solids content in the composted material is presented in Table 4. The MSE error of the models trained only on the basis of some of the image parameters was in a range from 20.346 to 52.211 for the test dataset. The MLP 25-10-1 TS5 model, considering all colour parameters, was the one with the smallest error. The MSE error of the MLP 5-10-1 TS6 model, which was based on the texture parameters, was equal to 42.737. In the case of the MLP 30-10-1 TS7 model trained with the dataset containing all image parameters, the MSE error was 19.359. Additional information on the composition of the composted material (percentage of sludge and type of straw) and the maturity stage of the compost at the input of the network had a significant effect on improving the prediction of the total solids content. This is probably due to the differences in the appearance of the composted material with different proportions of substrates used, both during intensive composting and after reaching the early maturity stage. The best accuracy was observed for the MLP 33-10-1 TS10 network, developed based on all of the image parameters and three additional features of the material analysed. The MSE error of this network was 2.470.
Detailed information on the models developed for the prediction of volatile solids content in the composted material is presented in Table 5. In the case of the models trained with the use of the selected image parameters, the MSE error was in a range of 8.582–14.418. The smallest MSE error in this group of models was observed for MLP 25-10-1 VS5, trained based on all colour features. The MSE error of the model developed on the basis of texture parameters (MLP 5-10-1 VS6) was equal to 13.308. The accuracy of the MLP 30-10-1 VS7 model trained based on all image parameters was lower than the accuracy of the model developed with the use of only colour features, which was unexpected. The use of parameters describing the composition of the composted material and the maturity stage of the compost as additional inputs to the model significantly improved the prediction of the volatile solids content. A similar phenomenon was observed for the models of the total solids content in the composted material. The lowest error of 0.494 was calculated for the MLP 33-10-1 VS10 model, developed taking into account all image parameters and three additional features describing the material.
MLP 33-10-1 TS10 and MLP 33-10-1 VS10 proved to be the best models for determining the total solids and volatile solids content in the composted mixture of sewage sludge and maize straw, respectively. The high accuracy of both neural models as predictive tools is confirmed by the regression issue statistics and error values (Table 6 and Table 7). The linear regression between the values of the total solids content predicted by the MLP 33-10-1 TS10 model and the experimental values of this parameter for the test dataset is depicted in Figure 3. The information on the error in the determination of the analysed parameter is presented in the form of a histogram. Analogous information on the volatile solids content and the MLP 33-10-1 VS10 model is included in Figure 4. The correlation coefficient for these two models was 0.9865 and 0.9876, respectively. Their uncertainties measured by the MAPE error were 2.88% and 0.59%.
The modelling of the composting of various substrates, including sewage sludge, to better control and optimize the process has been the subject of some scientific reports. Dogan et al. [45] used four artificial intelligence methods to model the process of the cocomposting of sewage sludge (dewatered by a decanter and separator) and biomass fly ash: feedforward neural network (FFNN), feedback neural network (FBNN), cascade forward neural network (CFNN), and deep cascade forward neural network (DCFNN). They found that the DCFNN model was the most accurate for the prediction of the pH, electrical conductivity, and NH4+/NO3− ratio, with MAPE values lower that 1%. The only exception was the composting process of sewage sludge dewatered by a separator, with a MAPE value of 1.99%. The three algorithms, namely, feedforward neural networks, Elman recurrent neural networks, and response surface methodology (RSM), were used by Dümenci et al. [46] to predict the compost maturity efficiency. In their study, olive mill wastes mixed with natural mineral materials (montmorillonite, kaolinite, sepiolite, and expanded vermiculite) were composted. The authors stated that the neural models (MAPE < 2%) were of better accuracy than the RSM model (MAPE > 10%). Higashikawa et al. [47] employed Fourier transform infrared spectroscopy and the partial least squares regression (PLS) method to predict the stability and maturity of compost-based substrates. The accuracy of the prediction depended on the maturity index. The coefficient of determination (R2) between the experimental and predicted values for the test dataset varied from 0.55 for the NH4+/NO3− ratio to 0.92 for the degree of polymerisation. The same techniques were used by Meissl et al. [48] to determine the humic acids content in composts. For all of the PLS models developed in this research, the R2 between the experimental and predicted values exceeded 0.8.
The topic of determining the total solids and volatile solids content in composted material using computer image analysis and neural modelling methods is very rare and, therefore, innovative. It has only been analysed in a few publications. Such research was conducted by Zaborowicz et al. [38] and Wojcieszak et al. [39]. They developed neural networks to predict the values of these parameters for composted mixtures of sewage sludge with maize, rape, and wheat straw. The model for determining the total solids content from compost images acquired under visible light presented in their report had an RMSE error of 0.0922. The model for the prediction of the volatile solids content was less accurate, with an RMSE error of 0.1722. The neural models presented in this work are described by a higher RMSE error. However, they were developed on the basis of a significantly larger dataset (1536 independent learning cases vs. 84 learning cases). Consequently, the predictive models presented in this work are characterised by a much higher generalisability and lower recall. The uncertainties of these models, expressed by the MAPE error, were 2.88% and 0.59%, respectively, for the prediction of the total solids and volatile solids content. These results may be considered very satisfactory.
Changes in the total solids and volatile solids content are crucial for monitoring a composting process [49]. The total solids content is a parameter that is subject to change during the process. This is due, among other things, to the evaporation of significant quantities of water contained in the initial mixture of substrates prepared for the composting. The analysis of changes in the volatile solids content provides information about the correct course of the composting process. This is due to the fact that some of the volatile solids contained in the substrate mixture decompose over time. Classical methods for the determination of total solids and volatile solids content are based on sampling from a reactor or pile and determining these parameters using drying and combustion processes. The disadvantage of these methods is that more than one day is needed to produce the results. The method proposed in this article is less invasive than the classic sampling of the composted mixture from the chambers. The parameters are determined on the basis of the photographs taken, and there is no need to irretrievably remove a portion of the compost from the reactor that will no longer be returned to it.

4. Conclusions

The management of biodegradable waste is one of the most important environmental challenges. Due to the fact of its relatively low investment costs and well-known technology, composting is a popular method for managing waste, including sewage sludge. In this study, twenty prediction models based on MLP topology were developed to determine the total solids and volatile solids content in a composted mixture of sewage sludge and maize straw. In these models, the input information was the image parameters describing the samples of composted material. Some of the models included additional information on the material. The best prediction results, for both the total solids and volatile solids content, were obtained using models with 33 input parameters. These parameters included 30 colour and texture parameters and three parameters for the percentage of sewage sludge, the type of straw, and the maturity stage of the compost, respectively. The uncertainty of the best of the models for determining the total solids content, expressed using the MAPE error, was 2.88%. The MAPE error in the case of the best model for determining the volatile solids content was 0.59%. The MAPE values, as well as the values of the other error metrics obtained for these two best models, demonstrated their high accuracy.
The results of this research indicate that neural networks combined with image analysis are a suitable tool for determining the selected physicochemical parameters of composted material. Neural image analysis has proven to be a fast, noninvasive method for predicting the total solids and volatile solids content of analysed material. The use of this method may provide an alternative to traditional time-consuming and invasive methods for determining these parameters based on drying and combustion. However, further research should be carried out to fully exploit the advantages of the proposed technique. This research should be related to the analysis of the composting processes of other common substrates or even the introduction of other machine learning methods (e.g., convolutional neural networks).

Author Contributions

Conceptualisation, S.K.; methodology, S.K. and W.C.; software, S.K.; validation, S.K., G.N. and K.P.; formal analysis, S.K. and K.P.; investigation, S.K.; resources, S.K. and W.C.; data curation, S.K.; writing—original draft preparation, S.K., G.N., W.C. and K.P.; writing—review and editing, S.K., G.N., W.C. and K.P.; visualisation, S.K. and G.N.; supervision, S.K.; project administration, S.K. and G.N. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Not applicable.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Czekała, W.; Jasiński, T.; Grzelak, M.; Witaszek, K.; Dach, J. Biogas Plant Operation: Digestate as the Valuable Product. Energies 2022, 15, 8275. [Google Scholar] [CrossRef]
  2. Ustawa z dnia 14 Grudnia 2012 r. o Odpadach (Dz.U. 2013 poz. 21). Available online: https://isap.sejm.gov.pl/isap.nsf/download.xsp/WDU20130000021/U/D20130021Lj.pdf (accessed on 10 January 2023).
  3. Cieślik, B.M.; Namieśnik, J.; Konieczka, P. Review of sewage sludge management: Standards, regulations and analytical methods. J. Clean. Prod. 2015, 90, 1–15. [Google Scholar] [CrossRef]
  4. Bęś, A.; Sikorski, Ł.; Szreder, K. The Effect of Mineral-Based Mixtures Containing Coal Fly Ash and Sewage Sludge on Chlorophyll Fluorescence and Selected Morphological Parameters of Deciduous and Coniferous Trees. Minerals 2021, 11, 778. [Google Scholar] [CrossRef]
  5. Halecki, W.; López-Hernández, N.A.; Koźmińska, A.; Ciarkowska, K.; Klatka, S. A Circular Economy Approach to Restoring Soil Substrate Ameliorated by Sewage Sludge with Amendments. Int. J. Environ. Res. Public Health 2022, 19, 5296. [Google Scholar] [CrossRef] [PubMed]
  6. Salinitro, M.; Montanari, S.; Simoni, A.; Ciavatta, C.; Tassoni, A. Trace Metal Accumulation and Phytoremediation Potential of Four Crop Plants Cultivated on Pure Sewage Sludge. Agronomy 2021, 11, 2456. [Google Scholar] [CrossRef]
  7. Ragonezi, C.; Nunes, N.; Oliveira, M.C.O.; de Freitas, J.G.R.; Ganança, J.F.T.; de Carvalho, M.Â.A.P. Sewage Sludge Fertilization—A Case Study of Sweet Potato Yield and Heavy Metal Accumulation. Agronomy 2022, 12, 1902. [Google Scholar] [CrossRef]
  8. Wydro, U.; Jankowska, M.; Wołejko, E.; Kondzior, P.; Łozowicka, B.; Kaczyński, P.; Rodziewicz, J.; Janczukowicz, W.; Pietryczuk, A.; Cudowski, A.; et al. Changes in Soil Biological Properties after Sewage Sludge and Pesticide Application in Wheat Cultivation. Appl. Sci. 2022, 12, 11452. [Google Scholar] [CrossRef]
  9. Siles-Castellano, A.B.; López-González, J.A.; Jurado, M.M.; Estrella-González, M.J.; Suárez-Estrella, F.; López, M.J. Compost Quality and Sanitation on Industrial Scale Composting of Municipal Solid Waste and Sewage Sludge. Appl. Sci. 2021, 11, 7525. [Google Scholar] [CrossRef]
  10. Rusănescu, C.O.; Rusănescu, M.; Voicu, G.; Paraschiv, G.; Biriș, S.; Popescu, I.N. The Recovery of Vermicompost Sewage Sludge in Agriculture. Agronomy 2022, 12, 2653. [Google Scholar] [CrossRef]
  11. Kulikowska, D. Kinetics of organic matter removal and humification progress during sewage sludge composting. Waste Manag. 2016, 49, 196–203. [Google Scholar] [CrossRef]
  12. Dach, J. Wpływ poziomu C:N na wielkość emisji amoniaku z kompostowanych osadów ściekowych. J. Res. Appl. Agric. Eng. 2010, 55, 14–18. [Google Scholar]
  13. Czekała, W.; Dach, J.; Ludwiczak, A.; Przybylak, A.; Boniecki, P.; Koszela, K.; Zaborowicz, M.; Przybył, K.; Wojcieszak, D.; Witaszek, K. The Use of Image Analysis to Investigate C:N Ratio in the Mixture of Chicken Manure and Straw; Falco, C.M., Jiang, X., Eds.; SPIE: Bellingham, WA, USA, 2015; Volume 9631, p. 963117. [Google Scholar]
  14. Haug, R.T. Compost Engineering. In Principles and Practice; Ann Arbor Science Publisher Inc.: Ann Arbor, MI, USA, 1980; p. 655. [Google Scholar]
  15. Wolna-Maruwka, A.; Schroeter-Zakrzewska, A.; Dach, J. Analysis of the growth and metabolic activity of microorganisms in substrates prepared on the base of sewage sludges and their impact on growth and flowering of garden verbena. Fresenius Environ. Bull. 2012, 21, 325–336. [Google Scholar]
  16. Kujawa, S.; Nowakowski, K.; Tomczak, R.J.; Dach, J.; Boniecki, P.; Weres, J.; Mueller, W.; Raba, B.; Piechota, T.; Rodríguez Carmona, P.C. Neural image analysis for maturity classification of sewage sludge composted with maize straw. Comput. Electron. Agric. 2014, 109, 302–310. [Google Scholar] [CrossRef]
  17. Kujawa, S.; Mazurkiewicz, J.; Czekała, W. Using convolutional neural networks to classify the maturity of compost based on sewage sludge and rapeseed straw. J. Clean. Prod. 2020, 258, 120814. [Google Scholar] [CrossRef]
  18. Czekała, W.; Jeżowska, A.; Chełkowski, D. The Use of Biochar for the Production of Organic Fertilizers. J. Ecol. Eng. 2019, 20, 1–8. [Google Scholar] [CrossRef]
  19. Boniecki, P.; Dach, J.; Pilarski, K.; Piekarska-Boniecka, H. Artificial neural networks for modeling ammonia emissions released from sewage sludge composting. Atmos. Environ. 2012, 57, 49–54. [Google Scholar] [CrossRef]
  20. Kujawa, S.; Nowakowski, K.; Tomczak, R.J.; Boniecki, P.; Dach, J. Image Parameters for Maturity Determination of a Composted Material Containing Sewage Sludge; Wang, Y., Yi, X., Eds.; SPIE: Bellingham, WA, USA, 2013; p. 88782K. [Google Scholar]
  21. Malinska, K.; Zabochnicka, M. Selection of bulking agents for composting of sewage sludge. Environ. Prot. Eng. 2013, 39, 91–103. [Google Scholar] [CrossRef]
  22. Białobrzewski, I.; Mikš-Krajnik, M.; Dach, J.; Markowski, M.; Czekała, W.; Głuchowska, K. Model of the sewage sludge-straw composting process integrating different heat generation capacities of mesophilic and thermophilic microorganisms. Waste Manag. 2015, 43, 72–83. [Google Scholar] [CrossRef]
  23. Kulikowska, D.; Gusiatin, Z.M. Sewage sludge composting in a two-stage system: Carbon and nitrogen transformations and potential ecological risk assessment. Waste Manag. 2015, 38, 312–320. [Google Scholar] [CrossRef]
  24. Czekała, W.; Malińska, K.; Cáceres, R.; Janczak, D.; Dach, J.; Lewicki, A. Co-composting of poultry manure mixtures amended with biochar—The effect of biochar on temperature and C-CO2 emission. Bioresour. Technol. 2016, 200, 921–927. [Google Scholar] [CrossRef]
  25. Żukowska, G.; Mazurkiewicz, J.; Myszura, M.; Czekała, W. Heat Energy and Gas Emissions during Composting of Sewage Sludge. Energies 2019, 12, 4782. [Google Scholar] [CrossRef] [Green Version]
  26. Czekała, W.; Janczak, D.; Pochwatka, P.; Nowak, M.; Dach, J. Gases Emissions during Composting Process of Agri-Food Industry Waste. Appl. Sci. 2022, 12, 9245. [Google Scholar] [CrossRef]
  27. Niedbała, G.; Mioduszewska, N.; Mueller, W.; Boniecki, P.; Wojcieszak, D.; Koszela, K.; Kujawa, S.; Kozłowski, R.J.; Przybył, K. Use of computer image analysis methods to evaluate the quality topping sugar beets with using artificial neural networks. In Proceedings of the SPIE—The International Society for Optical Engineering, Chengdu, China, 29 August 2016; Falco, C.M., Jiang, X., Eds.; SPIE: Bellingham, WA, USA, 2016; p. 100332M. [Google Scholar]
  28. Kozłowski, R.J.; Kozłowski, J.; Przybył, K.; Niedbała, G.; Mueller, W.; Okoń, P.; Wojcieszak, D.; Koszela, K.; Kujawa, S. Image analysis techniques in the study of slug behaviour. In Proceedings of the SPIE—The International Society for Optical Engineering, Chengdu, China, 29 August 2016; Falco, C.M., Jiang, X., Eds.; SPIE: Bellingham, WA, USA, 2016; p. 100332I. [Google Scholar]
  29. Wojciechowski, T.; Niedbala, G.; Czechlowski, M.; Nawrocka, J.R.; Piechnik, L.; Niemann, J. Rapeseed seeds quality classification with usage of VIS-NIR fiber optic probe and artificial neural networks. In Proceedings of the 2016 International Conference on Optoelectronics and Image Processing (ICOIP), Warsaw, Poland, 10–12 June 2016; IEEE: Warsaw, Poland, 2016; pp. 44–48. [Google Scholar]
  30. Gorzelany, J.; Belcar, J.; Kuźniar, P.; Niedbała, G.; Pentoś, K. Modelling of Mechanical Properties of Fresh and Stored Fruit of Large Cranberry Using Multiple Linear Regression and Machine Learning. Agriculture 2022, 12, 200. [Google Scholar] [CrossRef]
  31. Pentoś, K.; Pieczarka, K.; Serwata, K. The Relationship between Soil Electrical Parameters and Compaction of Sandy Clay Loam Soil. Agriculture 2021, 11, 114. [Google Scholar] [CrossRef]
  32. Kujawa, S.; Niedbała, G. Artificial Neural Networks in Agriculture. Agriculture 2021, 11, 497. [Google Scholar] [CrossRef]
  33. Jajja, A.I.; Abbas, A.; Khattak, H.A.; Niedbała, G.; Khalid, A.; Rauf, H.T.; Kujawa, S. Compact Convolutional Transformer (CCT)-Based Approach for Whitefly Attack Detection in Cotton Crops. Agriculture 2022, 12, 1529. [Google Scholar] [CrossRef]
  34. Kujawa, S.; Dach, J.; Kozłowski, R.J.; Przybył, K.; Niedbała, G.; Mueller, W.; Tomczak, R.J.; Zaborowicz, M.; Koszela, K. Maturity classification for sewage sludge composted with rapeseed straw using neural image analysis. In Proceedings of the SPIE—The International Society for Optical Engineering, Chengdu, China, 29 August 2016; Falco, C.M., Jiang, X., Eds.; SPIE: Bellingham, WA, USA, 2016; p. 100332H. [Google Scholar]
  35. Kujawa, S.; Mazurkiewicz, J.; Mueller, W.; Gierz, Ł.; Przybył, K.; Wojcieszak, D.; Zaborowicz, M.; Koszela, K.; Boniecki, P. Identification of co-substrate composted with sewage sludge using convolutional neural networks. In Proceedings of the SPIE-The International Society for Optical Engineering, Guangzhou, China, 10–13 May 2019; Volume 11179. [Google Scholar]
  36. Hemidat, S.; Jaar, M.; Nassour, A.; Nelles, M. Monitoring of Composting Process Parameters: A Case Study in Jordan. Waste Biomass Valorization 2018, 9, 2257–2274. [Google Scholar] [CrossRef]
  37. Wojciechowski, T.; Czechlowski, M.; Szaban, S. Wykorzystanie metod spektrometrii VIS-NIR do oceny wilgotności kompostów rolniczych. In Aktualne Problemy Inżynierii Biosystemów; Lipiński, M., Przybył, J., Eds.; Wydawnictwo Uniwersytetu Przyrodniczego: Poznań, Poland, 2014; pp. 86–93. ISBN 978-83-7160-731-8. [Google Scholar]
  38. Zaborowicz, M.; Wojcieszak, D.; Wojciechowski, T.; Mioduszewska, N.; Boniecki, P.; Przybył, J. Użycie metod komputerowej analizy obrazu oraz modelowania neuronowego do określania suchej masy organicznej kompostowanego materiału. In Aktualne Problemy Inżynierii Biosystemów; Lipiński, M., Przybył, J., Eds.; Wydawnictwo Uniwersytetu Przyrodniczego: Poznań, Poland, 2015; pp. 337–344. ISBN 978-83-7160-778-3. [Google Scholar]
  39. Wojcieszak, D.; Zaborowicz, M.; Przybył, J.; Boniecki, P.; Jędruś, A. Assessment of the Content of Dry Matter and Dry Organic Matter in Compost with Neural Modelling Methods. Agriculture 2021, 11, 307. [Google Scholar] [CrossRef]
  40. Wolna-Maruwka, A.; Dach, J. Effect of type and proportion of different structure-creating additions on the inactivation rate of pathogenic bacteria in sewage sludge composting in a cybernetic bioreactor. Arch. Environ. Prot. 2009, 35, 87–100. [Google Scholar]
  41. PN-EN 14346:2011. Charakteryzowanie Odpadów-OBLICZANIE Suchej Masy na Podstawie Oznaczania Suchej Pozostałości Lub Zawartości Wody. Available online: https://sklep.pkn.pl/pn-en-14346-2011p.html (accessed on 10 January 2023).
  42. PN-Z-15011-3:2001. Kompost z Odpadów Komunalnych-Oznaczanie: pH, Zawartości Substancji Organicznej, Węgla Organicznego, Azotu, Fosforu i Potasu. Available online: https://sklep.pkn.pl/pn-z-15011-3-2001p.html (accessed on 10 January 2023).
  43. Kujawa, S.; Tomczak, R.J.; Kluza, T.; Weres, J.; Boniecki, P. A Stand for the Image Acquisition of Composted Material Based on the Sewage Sludge; Othman, M., Senthilkumar, S., Yi, X., Eds.; SPIE: Bellingham, WA, USA, 2012; p. 83341R. [Google Scholar]
  44. Isabona, J.; Imoize, A.L.; Ojo, S.; Karunwi, O.; Kim, Y.; Lee, C.-C.; Li, C.-T. Development of a Multilayer Perceptron Neural Network for Optimal Predictive Modeling in Urban Microcellular Radio Environments. Appl. Sci. 2022, 12, 5713. [Google Scholar] [CrossRef]
  45. Dogan, H.; Aydin Temel, F.; Cagcag Yolcu, O.; Turan, N.G. Modelling and optimization of sewage sludge composting using biomass ash via deep neural network and genetic algorithm. Bioresour. Technol. 2023, 370, 128541. [Google Scholar] [CrossRef]
  46. Aycan Dümenci, N.; Cagcag Yolcu, O.; Aydin Temel, F.; Turan, N.G. Identifying the maturity of co-compost of olive mill waste and natural mineral materials: Modelling via ANN and multi-objective optimization. Bioresour. Technol. 2021, 338, 125516. [Google Scholar] [CrossRef]
  47. Higashikawa, F.S.; Silva, C.A.; Nunes, C.A.; Sánchez-Monedero, M.A. Fourier transform infrared spectroscopy and partial least square regression for the prediction of substrate maturity indexes. Sci. Total Environ. 2014, 470–471, 536–542. [Google Scholar] [CrossRef]
  48. Meissl, K.; Smidt, E.; Schwanninger, M.; Tintner, J. Determination of Humic Acids Content in Composts by Means of Near- and Mid-Infrared Spectroscopy and Partial Least Squares Regression Models. Appl. Spectrosc. 2008, 62, 873–880. [Google Scholar] [CrossRef]
  49. Lim, S.-S.; Park, H.-J.; Hao, X.; Lee, S.-I.; Jeon, B.-J.; Kwak, J.-H.; Choi, W.-J. Nitrogen, carbon, and dry matter losses during composting of livestock manure with two bulking agents as affected by co-amendments of phosphogypsum and zeolite. Ecol. Eng. 2017, 102, 280–290. [Google Scholar] [CrossRef]
Applsci 13 03363 g001 550
Figure 1. The photographic chambers used in the study.
Figure 1. The photographic chambers used in the study.
Applsci 13 03363 g001
Applsci 13 03363 g002a 550Applsci 13 03363 g002b 550
Figure 2. Example images taken at the beginning (left) and the end (right) of each composting experiment ((a)—experiment 1; (b)—experiment 2; (c)—experiment 3; (d)—experiment 4; (e)—experiment 5; (f)—experiment 6; (g)—experiment 7; (h)—experiment 8).
Figure 2. Example images taken at the beginning (left) and the end (right) of each composting experiment ((a)—experiment 1; (b)—experiment 2; (c)—experiment 3; (d)—experiment 4; (e)—experiment 5; (f)—experiment 6; (g)—experiment 7; (h)—experiment 8).
Applsci 13 03363 g002aApplsci 13 03363 g002b
Applsci 13 03363 g003 550
Figure 3. Results of the linear regression (a) and error histogram (b) for the test dataset and the MLP 33-10-1 TS10 model for the prediction of the total solids content.
Figure 3. Results of the linear regression (a) and error histogram (b) for the test dataset and the MLP 33-10-1 TS10 model for the prediction of the total solids content.
Applsci 13 03363 g003
Applsci 13 03363 g004 550
Figure 4. Results of the linear regression (a) and error histogram (b) for the test dataset and the MLP 33-10-1 VS10 model for the prediction of the volatile solids content.
Figure 4. Results of the linear regression (a) and error histogram (b) for the test dataset and the MLP 33-10-1 VS10 model for the prediction of the volatile solids content.
Applsci 13 03363 g004
Table
Table 1. Plan of experiments.
Table 1. Plan of experiments.
Experiment NumberSludge
Content (%)		Straw
Content (%)		Straw TypeExperiment Duration (Days)Sampling Days
13070untreated 361, 10, 20, 36
24555untreated361, 10, 20, 36
36040untreated 361, 10, 20, 36
43070ensiled291, 6, 14, 29
54060ensiled291, 6, 14, 29
65050ensiled291, 6, 14, 29
74555ensiled391, 7, 14, 25
85545ensiled391, 7, 14, 25
Table
Table 2. Colour and texture parameters extracted from the images.
Table 2. Colour and texture parameters extracted from the images.
No.Parameter CategoryColour ModelNumber of
Parameters		Description of Parameters
1colourRGB (original)9mean, median, and standard deviation of the R, G, and B components
2colourgreyscale3mean, median, and standard deviation of a pixel’s brightness
3colourbinary4the percentage of white colour in an image binarised using the assumed threshold values
4colourHSV9mean, median, and standard deviation of the H, S, and V components
5texturegreyscale5entropy, brightness, contrast, energy, and homogeneity
Table
Table 3. Input parameters used in the datasets.
Table 3. Input parameters used in the datasets.
No.Name of
Dataset for TS		Name of Dataset for VSNumber of
Parameters		Description
1TS1VS19colour parameters for the RGB model only
2TS2VS23colour parameters for greyscale only
3TS3VS34colour parameters for binary scale only
4TS4VS49colour parameters for HSV model only
5TS5VS525all colour parameters
6TS6VS65texture parameters only
7TS7VS730all image parameters
8TS8VS831all image parameters + SLUD_PER a
9TS9VS932all image parameters + SLUD_PER a + STR_TYPE b
10TS10VS1033all image parameters + SLUD_PER a + STR_TYPE b + IS_YCOMP c
a Sewage sludge percentage; b maize straw type (0—untreated, 1—ensiled); c compost maturity stage (0—immature compost, 1—compost at early maturity).
Table
Table 4. Neural models for the prediction of the total solids content in composted material.
Table 4. Neural models for the prediction of the total solids content in composted material.
No.ModelDatasetNumber of
Training Epochs		MSE
Training DatasetValidation
Dataset		Test Dataset
1MLP 9-10-1 TS1TS112219.52223.83723.393
2MLP 3-10-1 TS2TS211749.27449.72152.211
3MLP 4-10-1 TS3TS310348.29153.38951.965
4MLP 9-10-1 TS4TS45920.16323.92924.284
5MLP 25-10-1 TS5TS54912.61917.92420.346
6MLP 5-10-1 TS6TS67538.19445.00342.737
7MLP 30-10-1 TS7TS74011.39017.18519.359
8MLP 31-10-1 TS8TS82142.9958.10210.012
9MLP 32-10-1 TS9TS9492.0183.5114.759
10MLP 33-10-1 TS10TS10361.0492.4542.470
Table
Table 5. Neural models for the prediction of the volatile solids content in the composted material.
Table 5. Neural models for the prediction of the volatile solids content in the composted material.
No.ModelDatasetNumber of
Training Epochs		MSE
Training DatasetValidation
Dataset		Test Dataset
1MLP 9-10-1 VS1VS11008.4569.17010.902
2MLP 3-10-1 VS2VS25114.43913.32413.927
3MLP 4-10-1 VS3VS314314.44213.76614.418
4MLP 9-10-1 VS4VS4407.3429.7649.769
5MLP 25-10-1 VS5VS5525.2418.3928.582
6MLP 5-10-1 VS6VS65813.37213.08913.308
7MLP 30-10-1 VS7VS7914.7158.0629.315
8MLP 31-10-1 VS8VS8751.5002.9784.109
9MLP 32-10-1 VS9VS9400.6271.4961.570
10MLP 33-10-1 VS10VS10580.1410.3770.494
Table
Table 6. Regression statistics and error values for the 33-10-1 TS10 model.
Table 6. Regression statistics and error values for the 33-10-1 TS10 model.
Regression Statistics and
Error Values		Training DatasetValidation DatasetTest Dataset
Data Mean36.221337.012435.7915
Data SD9.75059.87459.6006
Error Mean−0.0176−0.0658−0.0094
Error SD1.02481.56711.5737
SD Ratio0.10510.15870.1639
Correlation0.99450.98730.9865
Coefficient of determination0.98890.97480.9731
RAE0.02730.04090.0424
RMSE1.02431.56641.5717
MAE0.66830.96721.0054
MAPE1.88472.71852.8802
Table
Table 7. Regression statistics and error values for the 33-10-1 VS10 model.
Table 7. Regression statistics and error values for the 33-10-1 VS10 model.
Regression Statistics and
Error Values		Training DatasetValidation DatasetTest Dataset
Data Mean76.072276.507376.3556
Data SD4.57094.56104.4766
Error Mean0.00140.04190.0324
Error SD0.37590.61300.7030
SD Ratio0.08220.13440.1570
Correlation0.99660.99100.9876
Coefficient of determination0.99320.98190.9753
RAE0.00490.00800.0092
RMSE0.37570.61360.7028
MAE0.25180.37760.4444
MAPE0.33500.49620.5896
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Kujawa, S.; Niedbała, G.; Czekała, W.; Pentoś, K. Neural Image Analysis for the Determination of Total and Volatile Solids in a Composted Sewage Sludge and Maize Straw Mixture. Appl. Sci. 2023, 13, 3363. https://doi.org/10.3390/app13053363

AMA Style

Kujawa S, Niedbała G, Czekała W, Pentoś K. Neural Image Analysis for the Determination of Total and Volatile Solids in a Composted Sewage Sludge and Maize Straw Mixture. Applied Sciences. 2023; 13(5):3363. https://doi.org/10.3390/app13053363

Chicago/Turabian Style

Kujawa, Sebastian, Gniewko Niedbała, Wojciech Czekała, and Katarzyna Pentoś. 2023. "Neural Image Analysis for the Determination of Total and Volatile Solids in a Composted Sewage Sludge and Maize Straw Mixture" Applied Sciences 13, no. 5: 3363. https://doi.org/10.3390/app13053363

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop