Seat Belt Detection Using Convolutional Neural Network BN-AlexNet

To tackle the problems of the dependence on source image clarity, the underutilization of source image information and the dependence on human-designed features in existing seat belt detection methods, a seat belt detection method using convolutional neural network (CNN) is proposed. In this paper, an improved convolutional neural network (called the BN-AlexNet) which adds the Batch Normalization (BN) module to the traditional convolutional neural network AlexNet is built to further enhance the classification ability of the convolutional neural network and greatly reduce the training difficulty. Later the confidence of detection results is analyzed, and the 95% confidence interval is used to set the rejection area. The result shows that the method achieves 92.51% correct detection rate by rejecting 6.50% test samples. Compared with the traditional methods based on image processing, the proposed method has higher correct detection rate.