Aiming at the problem of low accuracy of vehicle and pedestrian target detection under complex background and large scale changes,and the problem of insensitive detection and easy occurrence of false alarms and missed detection,the study proposes a vehicle and pedestrian target detection algorithm based on improved DAB-DETR algorithm.SIoU loss function is used to optimize the predicted bounding box,and the detection error of small targets is amplified to better deal with the problem of small target detection.GELU activation function is introduced to enhance the algorithm's generalization ability and improve its detection performance under complex background and occlusion.Experimental results show that the improved algorithm has increased the recall rate and average precision by 2.6％and 8.3％respectively compared with the original DAB-DETR algorithm,which verifies the effectiveness of the improvement method and provides algorithm support for vehicle and pedestrian target detection in complex scenarios.
关键词
车辆行人检测/DAB-DETR算法/GELU激活函数/Siou损失函数
Key words
Vehicular pedestrain detection/DAB-DETR algorithm/GELU activation function/SIoU loss function
维普
万方数据