基于图像处理和双BP神经网络的电气化铁路接触网立柱标识牌识别算法研究

故障位置点定位是实现轨道维护及保养的前提,利用接触网立柱标识牌实现定位是一种常用的轨道定位方法,但常规的识别方法存在识别率低且速度慢的缺点。针对该问题,提出一种基于图像处理和双神经网络的接触网立柱标识牌识别算法。首先利用Hough变换提取出接触网支柱区域,减小识别区域,其次通过形态学方法实现标识牌的定位与裁剪,再采用水平投影方法对字符进行分割,最后对字符中的字母和数字分别进行特征提取,利用两路并行的反向传播神经网络进行识别。通过实验验证了该算法的有效性,结果表明:该方法精度可达98.3%,相较于传统识别方法速度提高了17%。因此该识别算法能够实现轨道故障位置的快速精确定位,可用于轨道智能巡检系统。

Research on the Identification Algorithm of Electric Railway Catenary Pillar Signage Based on Image Processing and Double BP Neural Network

Fault positioning is the premise of track maintenance. The identification of catenary pillar signage to position faults is a commonly used track positioning method, but the conventional identification method has the disadvantage of low recognition rate and slow speed. Aiming at this problem, an identification algorithm of catenary pillar signage identification based on image processing and neural network is proposed. Firstly, Hough transform is used to extract the area of catenary pillar to reduce the recognition area. Secondly, morphological method is employed to position and cut the area of the signboard. Then the characters are segmented by the horizontal projection method. Finally, the features of letters and numbers in the characters separately are extracted and two parallel back propagation neural networks are used to identify letters and numbers in the characters respectively. The effectiveness of the proposed algorithm is verified by experiments. The results show the accuracy of the method is up to 98.3%, which is 17% faster than the traditional identification method. Therefore, the identification algorithm can realize fast and accurate positioning of the track fault location, and it is applicable to the track intelligent inspection system.