基于质心高度增量特征的目标识别算法  

Target Recognition Algorithm Based on Centroid Height Increments Feature

作　　者：于洋 郑伟 宋建辉 刘砚菊 YU Yang;ZHENG Wei;SONG Jian-hui;LIU Yan-ju(College of Automation and Electrical Engineering,Shenyang University of Science and Technology,Shenyang Liaoning 110159,China)

机构地区：[1]沈阳理工大学自动化与电气工程学院,辽宁沈阳110159

出　　处：《计算机仿真》2023年第4期213-218,共6页Computer Simulation

基　　金：国家重点研发计划基金资助项目(2017YFC0821001);辽宁省自然科学基金指导计划项目(2019-ZD-0252)。

摘　　要：针对传统目标识别算法识别准确率低、复杂度高等问题,提出基于质心高度增量特征的目标识别算法。在提取轮廓特征阶段,以轮廓质心为参考点,对于任意采样点,根据其它采样点相对于该点的高度关系构建质心高度增量描述符。描述符不仅计算简单,对旋转、平移和缩放等几何变换具有不变性,而且引入轮廓顺序这一全局特征,提升了描述符的鲁棒性和区分能力。在特征匹配阶段,利用轮廓顺序已知这一优势,采用动态规划算法计算质心高度增量描述符的相似度,最后引入形状复杂度分析,优化识别效果。MPEG-7测试集和Kimia99测试集的实验结果表明,上述算法能够有效的对目标图像进行匹配识别,而且对于噪声的干扰具良好的鲁棒性。Aiming at the problems of low recognition accuracy and high complexity of traditional target recognition algorithms,a target recognition algorithm based on centroid height increments was proposed.In the stage of contour feature extraction,the centroid of the contour was taken as the reference point.For any sampling point,the centroid height increment descriptor was constructed according to the height relationship of other sampling points relative to the point.The descriptor is not only simple in computation and invariant to geometric transformations such as rotation,translation and scaling,but also introduces the global feature of contour order,which improves the robustness and discrimination ability of the descriptor.In the feature matching stage,dynamic programming algorithm was used to calculate the similarity of centroid height increment descriptors,and shape complexity analysis was introduced to optimize the recognition effect.The experimental results of MPEG-7 test set and kimia99 test set show that the proposed algorithm can effectively match and recognize the target image,the proposed algorithm has good robustness to noise interference.

关 键 词：轮廓 质心高度增量 动态规划 形状复杂度 目标识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]