Abstract: The calculation amount of the current scraper detection method for coal preparation plants is large. Moreover, the method runs on the low-power, low-cost embedded Jetson Nano and has problems such as low execution efficiency and poor real-time performance. In order to solve the above problems, this paper proposes an edge computing-oriented scraper detection method that can parallelize the Hough transform.Firstly, the scraper image is pre-processed.Then the Hough transform algorithm is used to detect the scraper and calculate the scraper angle. If the scraper angle was less than the set threshold value, the alert would besent. If the scraper angle was within the threshold value, the display would be normal.In the scraper detection, the Hough transform is parallelized and the data is transmitted with zero copy. At the same time,the overall process of the scraper detection is designed into a CPU and GPU cooperative working mode. The scraper image pre-processing runs on the CPU side and the parallelized Hough transform runs on the GPU side.In this mode, the hardware resources in Jetson Nano can be fully utilized to realize real-time detection of the scraper in Jetson Nano.The experimental results show that by using the parallelized Hough transform algorithm, the scraper image in Jetson Nano with resolution of 960×540 can be detected 10 times faster than the original Hough transform. The detection frame rate can reach 17 frame/s and the scraper angle accuracy rate can reach 96.3%, which meets real-time requirements.