关键词: 茶叶, 溯源鉴定, 特征点匹配, 尺度不变特征变换, 随机抽样一致性

Abstract: In the feature point fine matching of tea cake images, manual selection of threshold will lead to false matching and missing matching problems, a method based on F1-Score maximization is proposed to automatically select the Random Sample Consensus (RANSAC) algorithm of distance threshold for feature point pair screening. In this paper, the Scale Invariant Feature Transform (SIFT) algorithm is used to extract the feature points of the tea cake image, and the Fast Library for Approximate Nearest Neighbors (FLANN) algorithm is used to coarsely match the feature points extracted from the heterogeneous image, and then the improved RANSAC algorithm is used to optimize the feature point matching. By comparing the matching accuracy and rms error of different algorithms, it is proved that the proposed algorithm can comprehensively consider the accuracy and recall rate of tea cake images after rotation, viewing angle and brightness transformation, and adaptively determine a distance threshold, and the improved RANSAC algorithm can increase its accuracy by up to 18.9%, and reduce the rms error by 0.706 pixel on average. Studies have proved that the proposed algorithm can achieve better matching effect.

Key words: tea, traceability identification, feature point matching, scale invariant feature transform, random sample consensus