Image matching method of pest image based on improved ORB

Tian Zhuo;Li Chengxuan;College of Information Technology, Jilin Agricultural University;

crop pest identification;;feature point detection;;image matching;;scale invariance;;ORB;;BRRB

This paper introduces the idea of BRISK algorithm and proposes an improved algorithm: BRRB algorithm(BRISK and ORB). Firstly, the feature points in the image are detected by the feature detection algorithm oFAST in the ORB algorithm, and the scale information is added to the feature points by the improved Harris corner response function. Finally, the feature points are uniformly sampled by the BRISK algorithm and the scale invariance is generated. Binary feature descriptor. The collected data of 200 pest samples were divided into 50 groups, and image registration experiments were performed separately. The experimental results show that the average matching accuracy of the BRRB algorithm is about 95%, which is about 73% higher than the original algorithm; the average calculation speed is 47.8 ms; in the comprehensive performance experiment, the average matching accuracy of the improved algorithm is higher than the traditional algorithm. Out of 0.6 percentage points, the illumination invariance is 1.9 percentage points higher than the traditional algorithm. The improved algorithm effectively solves the defect that ORB does not have scale invariance, and retains the high efficiency of the original algorithm in calculation speed and the invariance of rotation and illumination, so that the matching work of pest images is more accurate, and provides technical support for crop pest identification and prevention.