Research On Railway Foreign Object Intrusion Detection Method Based On Target Enhancement Fusion

With the rapid development of high-speed railway,ensuring traffic safety is the premise of railway transportation and image fusion technology has become one of the key technologies in the fields of computer vision,robot,target recognition and tracking.At present,most of the foreign object detection systems based on machine vision use a single source image to detect and recognize the visible image.At the same time,in order to solve the problem of the low accuracy of foreign object recognition in the foreign objects intruding railway clearance at unclear target,night and severe weather(rain,snow,fog,etc.),this paper proposes a new method based on target enhancement fusion.The method makes up for the problem that the detection effect of single visible image is not good at night or under the condition of low illumination.And the fusion image enriches the scene detail information,improves the accuracy of foreign object detection and the reliability of the system.Image registration is the premise of image fusion.Aiming at the problem of heterogeneous image registration in different field of view and different scales of railway scene,this paper proposes a novel infrared and visible image registration algorithm based on the rich linear features of railway scene.The polar projection method is firstly proposed to extract the straight rail lines from images.Then,we use feature descriptor to define the matching fraction formula to match the corresponding straight line accurately.Then we solve the transformation matrix and get the coarse registration results.We use the rough registration results to build candidate boxes to extract other plane lines,recalculate the parameters of the image transformation model,get the final geometric transformation matrix.Finally,the model transformation parameters are calculated according to all above corresponding lines to achieve fine registration.In order to ensure the real-time and accuracy requirements of moving object detection in railway scene,aiming at the false detection and missed detection caused by environmental noise in the process of Vibe algorithm detection,this paper designs a target enhancement algorithm based on improved Vibe,which successfully suppresses the ghost problem and environmental noise,and improves the target detection accuracy and robustness of the measurement.And then,according to the similarity of the moving target area,we adopt the corresponding target enhancement fusion strategies for the background area and the target with different similarity;Inorder to obtain the fusion image with both visible contour and color details and infrared temperature characteristics,this paper proposes a target level image fusion algorithm of the non subsampled contourlet transform(NSCT)based on the phase congruency and fourth-order correlation coefficient matching for same target areas.Firstly,the infrared image and visible image of the moving target are decomposed into high and low frequency by NSCT transform.For the low-frequency subband,this paper uses the fusion rule based on phase consistency,and for the high-frequency subband,this paper uses the fourth-order correlation coefficient matching fusion rule.Finally,the final fusion detection results are obtained by using the inverse transformation of NSCT.At last,we combine with the fusion images of other target regions and background regions,we obtain the enhanced fusion detection image of moving target.Through qualitative and quantitative comparison with other commonly related algorithms,the results of our proposed algorithmare obviously better than other algorithm,which not only realizes the accurate detection of moving objects,but also enhances the relevant information of the target image.And the method solves the problem of low target detection rate under the condition of low illumination and long distance.It has a very important application prospect for the research of railway foreign object intrusion detectionand operation safety guarantee system.