基于kAS特征的目标识别新方法

Target Recognition Method Based on kAS Feature

提出一种复杂背景下目标识别的新方法,利用Canny算子和多边形分别提取轮廓和逼近轮廓曲线,计算k邻接轮廓线段组(kAS)特征,利用ISODATA聚类算法得到kAS码书。提取特征时采用分块加权的kAS直方图,识别过程中采用支持向量机进行训练和分类。实验结果表明,该方法在复杂场景下可以获得较高的识别率,具有平移和尺度不变性等特点。

A method for target recognition in cluttered images is presented. The Canny operator and polygon are used to calculate and approximate the contour curve, the k Adjacent Segments（kAS） feature is calculated, and the kAS codebook is obtained by using ISODATA clustering algorithm. Block-weighted kAS histogram is used in feature extraction, Support Vectorl Machine（SVM） is applied to the training process and the classification process. Experimental results show that this method can get higher recognition rate, with the property of translation and scaling invariance.