A computer vision based train boxcar transfer state real-time detection system

With economic development, the goods transport volume by railway runs up. The manual way of railway boxcar safe state checking cannot adapt to the development of railage more and more. This paper proposed a computer vision based boxcar safe state real-time monitoring system to substitute manual work. Line feature is extracted from the frame. By analyse the distribution of the line feature to differentiate the car type and the boxcar safe state. The little calculating amount of the algorithm and robustness to the noise and illumination disturbance makes it have good real-time quality and dependability. Experiment indicate that the system reach the design requirement.