一种基于机器视觉的交通标识符自动识别方法

Auto-Recognizing Method for the Traffic Sign Based on Machine Vision

在交通标志实时识别时,由于参考图像与实测图像非同时获取,摄像机与被摄交通标志之间的位置难以保证完全相同,致使所获取的交通标志参考图像与实测图像之间可能产生几何失真。几何失真对交通标志图像的识别结果有较大影响,因此寻找一种具有旋转和比例不变性的图像识别方法势在必行。基于不变矩的交通标志识别算法,首先对交通标志图像进行归一化处理,然后求其Hu矩,最后利用相关性测度算法进行比较。结果表明,算法针对旋转或缩放后的交通标志图像具有100%的识别率。

When recognizing the traffic sign,because consult vision and survey vision couldn＇t be obtained at the same time,so gamma camera and position that is feared between the traffic signs are difficult to guarantee self-samly.Then reference traffic sign vision obtained and survey traffic sign might produce geometry distortion,geometry distortion will cause great influence to discernment result.So we need to look for a rotating and proportion-variant way.The recognition algorithms foe the traffic sign on the basis of distance-variant will first naturalize the signs,then calculate its HU,and then analyze the interdependences.The results indicated that the discernment algorithms put forward have 100% recognition capability to the figures.