Research on Recognition Method of Car Fuel Tank Cap Based on Residual Network

Abstract

Abstract Aiming at the problem that traditional algorithms have low recognition accuracy for different car fuel tank caps, a car fuel tank cap recognition algorithm based on improved Faster-RCNN is proposed. Firstly, the VGG16 network in the feature extraction network link is improved to the ResNet-101 residual network to improve the recognition accuracy and system speed. Secondly, the ROI pooling of the original pooling layer is changed to ROI align regional feature aggregation method makes the target rectangular frame more accurate. Finally, the fully connected network is used to perform the classification of the car fuel tank cap and the accurate regression of the frame. Through the detection of 200 actual images of car fuel tank caps, the experimental results show that the algorithm can achieve 98.0% accuracy and a detection speed of 0.139 seconds per image. Compared with traditional image algorithms, it can obtain faster speed and higher accuracy, and can provide technical support for unmanned gas stations, which has certain engineering application value.