基于不变矩和神经网络的交通标志识别方法研究

在交通标志实时识别过程中，由于参考图像与实测图像不是同时获取的，因此摄像机与被摄交通标志之间的位置难以保证完全相同。于是，所获取的参考交通标志图像与实测交通标志图像之间就可能产生几何失真。几何失真将对于图像识别的结果带来很大的影响。因此，需要寻找一种具有旋转和比例不变性的图像识别方法，以满足实际应用中的需要。针对上述问题，提出了一种基于不变矩和神经网络的交通标志识别算法。实验结果表明，所提出的识别算法具有很好的识别能力。

Study on Method of Traffic Signs Recognition Based on Neural Network and Invariant Moments

During the traffic signs real-time recognition process,test images are not obtained together with reference images,so the situation of camcorder (photo camera) is hard to be completely identical with that of traffic signs.It's to say that geometric distortion may be produced in test images,which would bring great effect on recognition results.So a method of image recognition with rotation and scale invariance is required to satisfy the practical application.In accordance with above problem,algorithm of traffic signs recognition based invariant moments and Neural Network was presented in the paper.The experiment results show that the algorithm presented by the paper has a good ability to recognize traffic signs.