Drives & Controls April 2022

32 n MACHINE VISION April 2022 www.drivesncontrols.com 3D imaging helps to spot dangerous defects in train rails T he surfaces of train rails are subjected to severe stresses caused by heavy freight vehicles, high-speed trains, and the constant braking and acceleration of all types of rail traffic. In addition to natural wear, these stresses become evident in the form of a variety of defects including spalling and flaking. Newly manufactured rails also need to be inspected before they are put into service. Using imaging systems to detect rail defects is a challenging task because of uneven light intensity levels, low contrast and noise. Two of the most common methods for detecting rail defects are 3D profile reconstruction, and using structured light and line-scanning stereo matching to obtain 3D information. Both of these are affected by the problem of encoding and decoding structured light fringes, and their reconstruction accuracy depends on their resolution. Imaging information can be affected by illumination levels and by shooting angles. Also, when detecting profiles, local distortion or loss of depth information can lead to false detection. To overcome these problems, the researchers from Northeastern University in Shenyang, China, devised a stereoscopic system capable of acquiring 2D images and 3D profiles of rail surfaces at high speeds and with high resolutions. Using a Chromasens 3DPIXA camera coupled with a Chromasens Corona II linescan light, combined with depth outlier detection, they were able to exclude non-uniform information acquisition between 2D and 3D during the inspection process, and gainmore effective information for detecting surface defects. A disparity map of the camera's left and right sensors was obtained by stereo matching. A corresponding depth map was then calculated based on the disparity map by combining the internal and external parameters of the sensors. To accelerate this calculation, a simple, fast stereo matching algorithmwas applied, so that precise depth information could be achieved with a resolution of 14μm and at a range of 52mm. The optical resolution of the camera used by the researchers can reach 70μm/pixel, with an acquisition speed of up to 1.4m/s and a maximum frame rate of 21kHz. It has a resolution of 7,142 pixels and a three-channel RGB sensor. Test results showed that the system based on the linescan camera outperformed 15 of the most advanced methods available for inspecting rail surfaces. To check the system’s applicability to grey images, a different dataset was tested, comparing the new systemwith other visual inspection techniques. This resulted in similarly good results. The Chinese researchers now plan to develop and enhance their system to detect shallow internal defects and defect attributes more effectively. n Chinese researchers have developed a stereoscopic vision system for locating defects in railway rails. Tests have shown that it out-performs other inspection technologies. The Chinese rail-monitoring application uses a camera mounted above the rail, with lighting being provided from either side Pairs of sensors are used to achieve stereo triangulation of the rail surface