Research On Image Positioning And Feature Recognition Algorithm For Track Fasteners

Track safety has always been a important research topic in China,and fasteners are one of the important components of the track.Its state determines the safety of the track.Once the fasteners are missing or incomplete,there is a significant risk to the safety of the railways,even would cause traffic accident.At present,China is researching on computer vision technology and image processing.Track fastener detection which is based on computer vision technology has become the future development trend,this technology has the advantages of fast detection speed and high accuracy.Based on the present research situation at home and abroad,this paper proposes a method of fastener detection based on computer vision.It mainly uses the image processing algorithm to identify the fasteners for the track images collected in the laboratory.The specific method is as follows:In order to realize the extraction of the regional target of the fastener in the image,an improved cross-location method is proposed.First,the left and right sides of the fastener are segmented according to the prior knowledge,and then the median filter is used to divide the segmented image.And then used Canny edge detection to the image,the binary image is obtained,and then the gray value of the upper and lower parts of the fastener is used to make the horizontal projection,and finally the threshold segmentation is used to obtain the image of the fastener part.The method is proved by experiments,and the preliminary extraction of the fastener part can be realized very well,and the extraction speed is fast and the detection is accurate.Firstly,the image is transformed by wavelet transform,the image information dimension is reduced,the image noise is reduced,the useful image data is enhanced,and then the image is obtained.Then,the image is extracted by using the wavelet transform and the local binary combination.The local binary algorithm is used to extract the features of the fasteners,and the image features at different scales are obtained.Then all the features are fused to form the fastener feature vector,which is identified by the nearest neighbor classifier.The 2DPCA algorithm can preserve the image structure information,and the PCA algorithm can reduce the image dimension and reduce the image data.The 2DPCA algorithm can extract the deduction of the image data and feature.Finally,it is judged by the nearest neighbor classifier.The above two methods have proved that the fastening part has a good recognition effect.In this paper,the fastener extraction algorithm proposed in the matlab experimental demonstration,are able to accurately distinguish the fasteners.