Design Of Rail Surface Defects Detection Based On Image Processing

With the rapid development of railway in our country,the traditional railway track detection methods cannot meet the requirements of high efficiency and accuracy.The track images have been treated as research objects.Combined with the related techniques such as image processing and pattern recognition,this paper is proposed based on a set of rail surface defects detection system of image processing.This system can recognize the rail surface defects intelligently.This paper mainly includes the following aspects.Firstly,with the comparison of several traditional image filtering algorithms,this paper chose a fast median filtering method which combined the mean,and used it in preliminary image processing.Combined with morphological top hat operation,this paper shows a method based on projection gradient which is used for positioning track surface area.In addition,the effectiveness of this method has been verified by the experimental data.Secondly,by comparing the differences between the former and the latter image,images can be judged by whether they include defects.According to the human visual attention mechanism,the defect region is segmented by the method of visual saliency computation,and the results are compared with the SR algorithm and Itti algorithm;thirdly,four features,shape feature,geometric features,gray features,and texture features,are extracted from the defect areas.Afterwards,all these data will be analyzed and used for constructing a SVM classifier.After completing the optimization of SVM parameters,the defects and fake defects(rail connections)can be classified precisely.Lastly,the defect detection software is designed to implement the algorithm,and the stability and practicability of the software have been tested.The experimental results show that the method in this paper has good stability,and the final detection result is 91.67% which has reference value in practice.