An Image Retrieval Algorithm Based on SURF for Embedded System

Abstract

To study the problem of real-time and accuracy of the image retrieval algorithm of embedded system with scale, rotation and illumination. Improved SURF algorithm, which is applied to the binary image feature extraction, improve the real-time performance of the image, the use of LSH algorithm to establish the index of the image feature database, to avoid duplicate feature extraction, the use of LSH algorithm to approximate the search image features, and the similarity into line sort, select the best matching image. When the image has scale, rotation, illumination, this algorithm has a higher retrieval accuracy than the SURF algorithm and ORB algorithm, and has a better robustness than the SURF algorithm. The experiments show that the algorithm has scale, rotation and illumination conditions of the embedded system of image retrieval, has good real-time performance and achieved good application effect.