摘要
建立了基于修正不变矩的车标识别方法.该方法通过模板匹配定位车标大致区域,利用边缘检测对车标进行精确定位,计算待识车标的修正不变矩特征值,利用不变矩的最小距离进行车标识别.该法不受待识别车标图像大小以及变形等因素的影响,可以自适应地进行车标定位和车标识别.实验结果表明,与Hu不变矩相比较,修正Hu不变矩可以提高图像目标的识别率,有利于车标的识别.
The vehicle-logo recognition method was built based on modified invariant moments. Template matching was applied to the rough location of vehicle-log, then, edge detection wae used to locate vehicle- logo precisely, the modified invariant moments eigenvalue was calculated for the recognition vehicle-logo, at last the vehicle-logo recognition was achieved by invariant moment minimum distance. The approach can adapt to complete object location and recognition, it can be unacted on the image size and deformation. The experimental results showed that compared with Hu invariant moments, the method based on modified in- variant moments has higher accuracy of recognition, which is conducive to the vehicle-logo recognition.
出处
《郑州轻工业学院学报（自然科学版）》
CAS
2013年第2期91-94,共4页
Journal of Zhengzhou University of Light Industry:Natural Science
关键词
不变矩
特征提取
车标识别
最小距离
invariant moments
feature extraction
vehicle-logo recognition
minimum distance
