Aiming at the problem of repeated detection of YOLOV3 algorithm in vehicle detection, the ADE-YOLOV3 vehicle detection algorithm is proposed. The algorithm uses K-means clustering algorithm to determine the number of target candidate frames and aspect ratio according to the inherent width and height characteristics of the vehicle. Then, according to the results obtained by clustering, the anchor parameters are reset, which makes the ADE-YOLOV3 network have certain pertinence in vehicle detection. Finally, the migration learning method is used to improve the network structure, and the optimal weight model is obtained, which improves the training precision of the model. The experimental results show that compared with the original YOLOV3 method, the mAP is increased from 91.4% to 95%, the repeated detection rate is reduced from 5.6% to 2.1%,and the detection speed by 50fps. The detection accuracy is improved and the problem of repeated detection is effectively avoided. Keywords: Vehicle Detection; Deep Learning; YOLOv3; K-means; Migration Learning ; Applied Science and Computer Science Publications.
Vehicle Detection Method Based on ADE-YOLOV3 Algorithm
2020-06-01
oai:zenodo.org:4261431
Journal of Information and Communication Engineering Volume 6(Issue 1) 347-352
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Vehicle Detection Based on Improved Yolov3 Algorithm
IEEE | 2020
|Vehicle detection method based on improved YOLOv3
IEEE | 2022
|Non-motor vehicle helmet wearing detection method based on improved YOLOv3 algorithm
Europäisches Patentamt | 2021
|Improving Performance of YOLOv3 for Vehicle Detection
IEEE | 2019
|Vehicle Detection in Aerial Images Based on YOLOv3
Springer Verlag | 2021
|