Resumen
Road intelligence monitoring is an inevitable trend of urban intelligence, and clothing information is the main factor to identify pedestrians. Therefore, this paper establishes a multi-information clothing recognition model and proposes a data augmentation method based on road monitoring. First, we use Mask R-CNN to detect the clothing category information in the monitoring; then, we transfer the mask to the k-means cluster to obtain the color and finally obtain the clothing color and category. However, the monitoring scene and dataset are quite different, so a data augmentation method suitable for road monitoring is designed to improve the recognition ability of small targets and occluded targets. The small target mAP (mean average precision) recognition ability is improved by 12.37% (from 30.37%). The method of this study can help find relevant passers-by in the actual monitoring scene, which is conducive to the intelligent development of the city.