Study On Detecting Methods For Foreign Body Of Locomotive Roof

At present both passenger and cargo transport system are in a stage of rapid development in China, especially the mileage of high-speed railway has expanded rapidly, which accompanied by tremendous pressure for safe operation of the railway. The presence of foreign bodies on locomotive roof will bring potential danger to the safe operation of the train, such as the roof device damage, causing the device does not work properly. But the locomotive roof foreign body alarm detection has always been a difficult problem. Until now, there is not an complete visual equipment that can recognize the locomotive roof foreign body automatically at home and abroad. For roof foreign body recognition and detection, a foreign body recognition method is presented based on invariant moment features and feature matching in this paper. The method provides an important reference solution for the roof foreign body alarm detection system, which has important research significance.Firstly, the main functions and characteristics of the locomotive roof foreign body detection system is described, the system can detect the object outside the roof device in normal weather and light conditions. But at night or dark environment, the system needs to open additional lighting to work normally. Then the whole system frame, the video image acquisition and the installation of cameras are discussed, mainly composed by the image acquisition subsystem, an image transmission system and image processing subsystem. The key point is the image processing subsystem, whose performance determines the accuracy and timeliness of the foreign body detection analysis system.Secondly, pre-process the roof image to improve the timeliness and accuracy of the detection system. Mainly includes the sample processing of video image, which improves the speed of image analysis; using the gamma correction method to enhance the roof image details at night or dark light shooting situations and the median filtering method to reduce the noise produced in the process of image capture or transmission.Finally, the establishment of no foreign body reference model and foreign body detection model is described detailed. Because of the active detection methods can reduce the storage space of features, shorten the treatment time and more timeliness, so this paper chose active foreign body detection model. Roof images of the inherent devices (without foreign bodies) and those of the running train are processed respectively with the cubic B-spline wavelet edge detection to get multi-scale edge images and segment objects. And thus the invariant moments of each closed area in the edge images are extracted. Then, after calculating the similarity measure by Euclidean distance, we can judge whether the objects are foreign bodies.The roof foreign body recognition algorithm is written in the MFC platform. Through the experimental simulation results, this method can identify the locomotive roof foreign body effectively. In the end of the paper, improved algorithm about the foreign body detection is discussed for perfecting the system, and applied to the railway safety and security system.