基于形变校正的形状识别方法

Shape recognition algorithm based on deformation correction

针对鱼类形状易发生形变导致识别困难的问题,提出了一种新的识别算法。该算法通过对弯曲的鱼类图像进行形变校正,得到接近标准形状的鱼类轮廓,再利用等间距极坐标映射方法得到映射图,提取映射图中的局部极值点大小和相邻极值点的间隔,最后使用匹配识别算法计算匹配度完成对鱼的识别。实验结果表明,该算法消除了弯曲形变给识别带来的影响,实现了准确的识别分类。

Aimed at shape recognition with non-rigid objects deformation, a new recognizing algorithm is presented. In this algorithm, correcting algorithm is applied to correct fish image, then fish profile is obtained, which is similar with standard shape. A mapping image is gotten by using even-grid-polar mapping algorithm, local extreme point＇ s value and distance of adjacent extreme points are extracted as features for recognition, at last the matching degree is calculated to complete recognition. Experimental results show that the algorithm eliminates the effects from bending deformation, and achieves the accurate recognition.