Computer Vision Technology Based on Color Quantized Image Global Matching Algorithm

The indoor positioning service of computer vision is increasingly becoming an important demand in people's lives, and with the popularization of intelligent technology and equipment, indoor positioning technology has become a research hotspot in visual positioning. A visual localization system based on 2D to 3D conversion feature point matching was designed using color space quantization descriptors and image matching algorithms. Firstly, for the global image matching algorithm, color quantization descriptors are used to perform global matching on the image database, in order to select candidate matching images with similar scenes. Using the improved eight point method to solve the basic matrix, two matching images and user images are obtained for localization estimation. Finally, a matching image with high similarity is obtained by using 2D to 3D conversion feature point matching, and the positioning error is compared with the matching image. The experiment outcomes denoted that the matching error of the optimized eight point method is below 0.1, and the average positioning error of solving the basic matrix is 1.385, while the average positioning error of the 2D to 3D conversion feature point matching method is 0.463. Therefore, it has been proven that the visual positioning system based on 2D to 3D conversion feature point matching method has high accuracy.