Research on Automobiles Tire Detection Method Based on Deep Learning
Aiming at the problems of small scale damaged tires and the difficulty in detection due to the similarity between features and background,a deep learning tire damage detection method based on improved YOLOX(YOLOX-O)is proposed.Using the YOLOX model as the framework,downsampling layer pruning is performed on the original YOLOX network to increase the minimum feature map scale for detecting small-scale damaged targets.To solve the difficult damage caused by similar features and backgrounds,a max pooling layer and 1×1 convolutional layer are added to the CSP structure of the backbone network to preserve the most prominent target features.To balance the problem of imbalanced target samples,the BCE-loss function used to calculate the loss function in the original network is replaced with Focal loss.The model is compressed and optimized using the Aidlux platform and deployed to mobile devices to achieve target detection on the mobile terminal.Experimental results show that the proposed method can not only detect tires of small damage scales and similar features and backgrounds with a detection average accuracy of up to 90.8％,but also achieve fast detection on mobile devices.It is suitable for promotion and application.
万方数据
