In order to solve the problem of current pedestrian detection algorithms for infrared images in terms of low accuracy and missing detection,an infrared pedestrian target detection method based on improved YOLOv5s was proposed.Firstly,the Trans-former coding structure was utilized to replace the Bottleneck structure in the C3 module in order to strengthen the feature fusion capability of the detection network.Secondly,the recursive gated convolution gnConv was utilized to improve the visual sensory field module RFB,and the improved RF-gnConv module was added in front of the YOLOv5s head detection network,leading to the improvements of the model's resilience to pedestrian detection in various complex scenes.Finally,the algorithm model was vali-dated using the OTCBVS dataset.The results show that the improved algorithm model achieves an average accuracy of 97.3% ,and the detection speed is 63 frames/s,indicating the effectiveness of improved algorithm,mentioned in this paper,for the detection of pedestrians in infrared images.
infrared imagepedestrian detectiondeep learningreceptive field
NETL
NSTL
万方数据
