融合图像去雾与Tiny-YOLOv3的护帮板状态检测研究

Research on the Status of the Hydraulic Support Guard Plate Based on Image Dehazing and Tiny-YOLOv3 Algorithm

为解决液压支架工长时间作业过程中,因身体疲劳不能及时发现护帮板未护帮的问题,采用实时性高的Tiny-YOLOv3算法检测护帮板状态,但检测任务会受到综采工作面尘雾的影响。因此,提出一种融合图像去雾与Tiny-YOLOv3的目标检测算法,并在此基础上优化图像去雾算法的CUDA实现,首先将暗通道图像用RGB单通道图像代替,然后按列分组求大气光值,合并初始透射率的kernel函数并优化精细化透射率计算方式,提升图像去雾速度,保证算法的实时性。实验结果表明,在煤矿护帮板状态检测场景中,融合算法比Tiny-YOLOv3算法的准确率提高了22.8%,且满足实时检测的要求。

In order to solve the problem that the hydraulic support workers can not find the hydraulic support guard plate in time due to physical fatigue, a real-time Tiny-YOLOv3 algorithm is used to detect the status of the hydraulic support guard, but the detection task will be subject to the impact of surface dust and fog oncomprehensive mining work.Therefore, an object detection algorithm combining image dehazing and Tiny-YOLOv3 is proposed, and the CUDA implementation of the image defogging algorithm based on this is optimized.First, the dark channel image is replaced with a single RGB image, and then is grouped by column atmospheric light value, the kernel function of the initial transmittance is combined and the refined transmittance calculation method is optimized to improve the image defogging speed and ensure the real-time performance of the algorithm.The experimental results show that the accuracy of the fusion algorithm algorithm is 22.8% higher than that of the Tiny-YOLOv3 algorithm in the state detection scenario of coal mine guard board, which meets the requirements of real-time detection.