摘要
本文研究了一种基于支持向量机(SVM)的车型图像识别算法。采用图像边缘检测方法,该方法首先基于邻域灰度极值提取边界候选图像,然后以边界候选像素及其邻域像素的二值模式作为样本集,进行运动目标分割并提取具有RST不变性的轮廓特征向量,输入支持向量机进行训练和识别。此外,该算法与传统的算法比较,使用核函数少,计算量小,能较好地解决小样本、非线性和局部极小点等问题。实验表明,基于支持向量机(SVM)的车型图像识别算法具有更好的性能。
A kind of image recognition algorithms for recognizing automobile type based on the supporting vector machine （SVM） is studied in the article. The image edge detection method is adopted, which picks the edge candidate image up based on the neighbourhood gradation extremum, then produces sample set using the bilevel patterns of these candidate pixels and their neighborhood pixels to finish the motion target segmentation, and gains the profile characteristic vector with RST invariance. After that, the vectors are imported into the supporting vector machine for training and distinguishing. Besides, comparing with the traditional algorithm, this algorithm utilizes less core functions, has smaller calculate amount,and can resolve such problems as minor sample book , nonlinearity and part fairly good minimal point well. The experiment indicates that the algorithm owing to the supporting vector machine （SVM） has much better functions.
出处
《电子测量技术》
2008年第7期22-25,共4页
Electronic Measurement Technology
