Detail Enhancement And Scene Reconstruction Of Catenary Inspection Tour Images

With the significant increase in the operating mileage of railways and the running speed of trains,the problems,such as increased wear and short life of bow nets,have brought huge challenges to rail traffic catenary safety detection.The traditional catenary security monitoring system is difficult to meet the needs of the times.The security monitoring of the c catenary system based on intelligent video analysis has become a key technology to be solved,which will promote the essence of the safety guarantee mode from passive safety to active safety,contact monitoring to non-contact monitoring.The research work of this paper is based on the high speed railway power supply safety detection and monitoring system(6C system).The inspection video of the catenary,which was previously taken to the airborne camera as the experimental data,uses the image enhancement filtering algorithm based on the semantics to obtain the high quality of the catenary observation image,so as to improve the monitoring image of the catenary under the bad environment.The macroscopic structure of key components of catenary is strengthened and extracted.Through the motion estimation and scene reconstruction of catenary,the 3D state reappearance of catenary pillar is realized,which lays the foundation for later catenary safety monitoring.Finally,the practicability and effectiveness of the algorithm is verified by comparative experiments.The main contents of this paper are as follows:(1)In the process of image pre-processing,the camera in the driver's room has serious reflections on the patrolling image of the catenary captured through the windshield.This article uses the single-image defocusing blur reflex suppression algorithm to weaken the reflection in the image.According to the gradient sparse prior knowledge of the image,the objective function is obtained combining with the gradient information of the two image layers to establish the corresponding probability model.The objective function corresponding to the two image layers obtained is a non-convex function,and the function needs to be further optimized so that a clear image layer and a reflective layer are separated.Compared to the Retinex method,the one-dimensional image entropy value and mean-square error of a clear layer obtained by the single-map defocusing blurring reflex suppression algorithm is smaller,which verifies that the single-image defocus blur reflexion algorithm is more suitable for the work of this paper.(2)For the problem of unclear video images on the catenary,this paper proposes two methods to visually enhance the details of the image.In the forward vehicle video image enhancement method based on the semantic of the catenary,the edge of the catenary is acquired by using the edge detection network of the catenary and the template matching method;and the visual reinforcement of the contact network is further implemented by using the image AlphaBend hybrid method.In the image detail enhancement method based on structured forest edge detection,the structured forest edge detection algorithm is used to extract the detail edges in the image,and then the image details are visually enhanced by combining the image AlphaBend blending method.Both image enhancement methods have clear catenary monitoring images in harsh environments.Finally,these two methods are compared with other image enhancement methods to verify the effectiveness of the proposed method.(3)This paper studies the scene reconstruction of the catenary,including the catenary motion estimation and panorama synthesis.First,the FOE(Focus of Expansion)in the video sequence is obtained by the epipolar method.Then the image is divided into regions to extract the scene area of the catenary pole by the FOE.The selected line segment detector was used to detect the position of the catenary pole,and the catenary pole area was obtained,and the perspective transformation was performed to correct the catenary pole area.Then,the synthesis based on the gray-scale related panorama is selected.Finally,the gray-scale related features of the image are extracted,and the panorama is synthesized.The direct average image fusion method is used to smooth the traces at the synthesizing and stitching,thereby realizing the synthesis of multiple image panoramas and reducing the video storage.This paper tests the existing patrolling image datasets of the catenary.The results show that clear patrolling images of the catenary can be obtained by using this algorithm,and the panorama of the catenary pillars is obtained,which lays a foundation for the detection of anomalous catenary.It also verifies that this algorithm has a certain engineering application value.