论文部分内容阅读
为了解决复杂环境中采集的交通标志出现不同程度的几何失真现象,将不变矩具有的平移、旋转及比例缩放不变性特征用于图像识别中。首先对图像进行预处理,然后分别提取图像的Zernike和Hu不变矩特征,建立了相对应的feature Data数据集,最后将数据集输入支持向量机进行了目标分类。对德国公开的交通标志标准数据库(GTSRB)中的识别图库及实时采集的图像进行了测试。试验结果表明:与Hu不变矩比较,提取图像Zernike不变矩与支持向量机的识别方法对复杂环境中的交通标志识别具有更高的识别率和实时性。
In order to solve the geometric distortions of traffic signs collected in complex environment, the invariance, translation, rotation and scaling invariant features of invariant moments are used in image recognition. Firstly, the image is preprocessed, then the Zernike and Hu moments invariant features of the image are extracted respectively, and the corresponding feature data datasets are established. Finally, the dataset is input into SVM to classify the target. The identification gallery of the publicly available German Traffic Mark Standard Database (GTSRB) and the images captured in real time were tested. The experimental results show that compared with Hu moment invariants, the recognition method of Zernike moment invariants and SVM can improve the recognition rate and real-time of traffic sign recognition in complex environment.