Abstract:According to the entrance and export of construction surveillance video, this paper discusses the low-resolution image recognition method of the safety helmet and deduces the relation of different features and classifiers with recognition rate in view of taking account of the low-resolution safety helmet recognition problem at the long distance. It first captured the head of video to obtain images of the size of 22×22, and then extracted the statistical features of each image, LBP and PCA features. Finally, the recognition rate of test sample was calculated by taking advantage of minimum distance classifier and BP artificial neural network. The experimental results show that the LBP statistical features in combination with BP artificial neural network can recognize the safety helmet effectively. The recognition rate reached 87.27%.