Investigation on Traffic Signs Recognition Based on BP Neural Network and Invariant Moments

According to the traffic sign recognition process converges slowly, easy to fall into local minimum of defects in general BP algorithm, a BP algorithm is discussed in the paper, which is based on the improved conjugate gradient method on the basis of Fletcher-Reeves linear search method. At the beginning, the basic principle of the algorithm is introduced; meanwhile, an in-depth analysis is discussed from the theoretical aspects in the paper, and then the trained BP neural network is applied into the function approximation. Experimental results show that the algorithm has many advantages, such as it has a good effect in solving the rotation and scale invariance with the image, the extracted features with translation, scale and rotation invariance. The fast convergence speed and less iteration of this algorithm can meet the requirements of unmanned vehicle autonomous navigation.