基于多尺度角点的种子破损特征检测算法

针对种子破损局部特征的检测,提出一种基于多尺度角点准确性检测的簇分类算法。该算法对多尺度破损特征角点分类并进行分类点空间位置和尺度上非最大值抑制确定破损局部结构的簇候选特征点,并以候选特征点来参与特征准确性精化的迭代计算。试验结果表明,该算法去除了常规多尺度角点算法检测破损特征存在的冗余点,有效提高了多尺度下种子破损特征定位准确度,降低了迭代计算的时间复杂度。

The algorithm of seed damage feature detection basing on multi-scale corner

The paper proposed a cluster classification algorithm for the detection of accuracy of multi-scale corner used for seed damage detection.The algorithm was used to classify multi-scale damage feature corner and to explore cluster-candidate feature points of damage via non-maximum suppression in space domain and scale domain.The candidate feature points were used for the computation to refine the accuracy of the features.Experiment result showed that the algorithm midht remove the redundant points of detection for damage feature of existing multi-scale corner algorithm.It could improve accuracy to locate seed damage feature of multi-scale and reduce time complexity of iterative computation.