Product Recognition Algorithm Based on HOG and Bag of Words Model

Abstract

The rapid detection and identification of products based on computer vision has important applications in the fields of unmanned retail and goods sorting. At present, the recognition rate of traditional product identification methods is not high, and the deep learning recognition method requires large-scale training and cannot meet real-time requirements. This paper proposes a product identification algorithm that combines traditional HOG detection with the SIFT feature-based bag of words model for the needs of product identification. Compared with the traditional product identification method for feature matching, the algorithm has the advantages of higher recognition rate and shorter time. The test results show that the real-time recognition rate can reach 98%. At the same time, the algorithm has the advantages of light weight and easy portability, and can be applied to many occasions such as unmanned retail or express picking.