In the process of annual vehicle inspection, strict inspection of vehicle light failure is of great significance to ensuring road traffic safety. Aiming at the problems of uneven lighting conditions in vehicle inspection tasks, different sizes of lights, and difficulty in detection, a vehicle light detection algorithm based on the improved CenterNet algorithm is proposed. The algorithm uses the ResNet18 network as the backbone network, uses DilatedEncoder expansion encoder and feature fusion enhancement module to strengthen the algorithm’s learning and processing capabilities for different scales of car light feature information; adds GIoU loss and distance deviation loss to constrain the position of the prediction center point and adjustment the size of the prediction box. The experimental results show that compared with the original algorithm, the improved algorithm increases 4.2% in the self-built Vehicle inspection vehicle light detection dataset, and 1.8% in the target detection COCO dataset. The improved algorithm has good accuracy and real-time performance for vehicle light detection, and basically meets the actual needs of motor vehicle detection stations.
Vehicle Light Detection Method Based on Improved CenterNet
Lect. Notes Electrical Eng.
International Conference on Computing, Control and Industrial Engineering ; 2021 ; Hangzhou, China October 16, 2021 - October 17, 2021
6th International Technical Conference on Advances in Computing, Control and Industrial Engineering (CCIE 2021) ; Chapter : 47 ; 470-477
2022-07-06
8 pages
Article/Chapter (Book)
Electronic Resource
English