An Improved Pedestrian Detection Method Based on YOLOv5 Network

In pedestrian detection research,, occlusion has always been an important challenge that requires immediate attention. In order to solve this predicament, an improved pedestrian detection method based on YOLOv5 was proposed. Firstly, the Mosaic9 data enhancement method was used in the data processing stage to enhance the training effect of the model and make the network more robust. Secondly, the Selective kernel attention mechanism is added to the Neck part of YOLOv5 to boost the characteristic extraction extraction in relation to occluded goals. Finally, the hierarchical pooling structure layer, Spatial Pyramid Pooling Fast(SPPF) within the foundational YOLOv5 framework algorithm was substituted by atrous space convolutional pooling pyramid Atrous Spatial Pyramid Pooling(ASPP) to enhance the model's perception scope the model. Experiments show that contrasted against the initial YOLOv5, average precision and recall rate have been greatly increased, so that the occlusion target can be effectively detected.