Pedestrian and vehicle detection via port surveillance video

In view of the diversity and complexity of the landscape environment and spatial layout of the port, the behaviors of relevant personnel and vehicles are relatively hidden, which leads to the problem that they can't be accurately identified in the surveillance video. In this paper, we used an improved object detection algorithm for personnel and vehicles applied in port environment, which combines the feature pyramid networks for feature extraction based on ResNet-101 and Faster R-CNN object detection network. The experimental results show that the algorithm has good real-time performance, and effectively improves the accuracy of related personnel and vehicle location detection and recognition in the port environment.