Coal mine underground people counting method based on bipartite graph

The invention relates to a coal mine underground people counting method based on a bipartite graph, and belongs to the field of artificial intelligence. The method comprises the following steps: taking images of a coal mine well entering mouth or an underground key place as input data, taking yov5 as a backbone network for target detection, taking interval time T to obtain the images, forming a coal mine well entering area personnel bipartite graph, performing personnel matching at T1 and T2 moments by using a Hopkroft-Carp algorithm, and counting the number of people entering and exiting the well; according to the method, on the basis of a yov5 lightweight personnel detection model, rapid iteration and deployment implementation can be realized in field application under the condition that the prediction precision is met. 本发明涉及一种基于二部图的煤矿井下人数统计方法，属于人工智能领域。以煤矿入井口或井下关键地点图像作为输入数据，yolov5作为目标检测的主干网络，取间隔时间T获取图像，形成煤矿入井区域人员二部图，使用霍普克洛夫特-卡普算法进行T1和T2时刻的人员匹配，统计出入井人数；本发明基于yolov5轻量型人员检测模型，可在满足预测精度的条件下，现场应用中实现快速迭代和实施部署。