Location And Detection Image Recognition Algorithm Based On The Structural Characteristics Of Wagon Bogie Failure

With China’s rapid economic development, the railway freight transport represents a high-speed, high-density, and large traffic trends. Because of some deficiencies like low efficiency, high risky and sensibility to the abilities of operators, traditional Parking and Inspecting Algorithm cannot meet the development of the railway trucks transport. Consequently, we designed and developed a dynamic image detection system for trucks running fault, eventually achieving the goal of automatic locating and detecting for these types of faults and thus presenting great practical significance to improve the efficiency, accuracy and safety for Freight Train Inspection.Based on the failure feature and structural characteristics of the bogie, this paper discusses investigates a new kind of localization and identification algorithm for bogie faults, including bolts (the center plate bolts and plywood bolts) loss fault, cross rod bending failure and safety chain lost failure. The contributions of this work are as follows:This paper designed a kind of failure region location algorithm based on the structural characteristics of bolts missing. Then, according to the property that large dark circular cavities would be formed in contrast with the surrounding gray scale after bolts miss, a Hough Transform holes detection method is proposed, which utilize the theory of the accumulation of binary image edge gradient direction and constraints of radius of the circle. This method first calculates the binary image and edge gradient direction for the region of interest, and then reduces the search space of the center of the parameters, according to the gradient direction and constraints of the radius of the circle. This method improves the processing speed of the Hough Transform detecting and solves the problems of bolts missing fault detection under the finite changes of geometry, rotation and gray-scale value, by mean of adopting the internal gray circular area change to detect whether bolts miss.To solve the problems like that cross rod bending failure is difficult to detect, location of the fault it not fixed and cross rod is susceptible to interference, this paper first locates cross rod through the geometry structural characters of bogie, and then proposes a cross edge linear detection method based on geometry deformation of cross rod, improving the robustness of fault detection under dislocation of fault image and change of gray-scale. In terms of the high stability of cross rod, this paper proposes a localization and binarization method for the interested mainstream safety chain areas and then exacts maximum connectivity area from binary chain area. By calculating the area is maximum connectivity area, this method could find out the safety chain and avoid the problem that the mainstream safety chain gets lost when the geometry change.In this project, this paper introduce three fault localization and detection software modules for lose fault of the center plate bolts and plywood bolts, cross rod bending fault and safety chain loss fault. These modules were tested on Northern Railway and two-month practical experiment results exacted from tests on41328cars in Wuhan Northern Railway prove that every false detection rate of the three fault types is below10%, and the rate of missing detection is below5%.These rates meet the standard requirements and also prove these algorithms are eligible.