An Online Metro Train Bottom Monitoring System Based on Multicamera Fusion

The structure of the train bottom is relatively complex and has many small components. The failure of train bottom will threaten the safety of passengers, and train bottom monitoring is important for the safety of train operation. Thus, an online metro train bottom monitoring system based on multicamera fusion is developed. First, the linear array cameras are used to collect the images, effectively overcoming the problems of distortion and repeated captures. Then, an adaptive image correction method is introduced to correct the underexposed and overexposed images. The image-stitching method based on scale-invariant feature transform (SIFT) feature image registration is used to concatenate the train bottom images. Finally, the developed monitoring system is applied in Guangzhou Metro Line 21. The results show that the developed correction method effectively corrects the underexposed and overexposed images. The feature matching is performed after determining the overlap areas, which reduces the number of iterations and improves the stitching speed of the system. Compared with the existing method, the stitched images have higher quality in peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and difference of edge map (DoEM).