针对选煤厂脱泥筛上杂物检测问题,提出了一种开放场景下筛上杂物目标检测方法,针对不同类型的筛上杂物图像,基于目前成熟的目标检测网络框架构建筛上杂物检测模型,并通过定量实验分析,验证了方法的有效性;在此基础上,通过定性分析,探究了不同光照条件对模型检测效果的影响,并针对图像中存在的局部亮斑,引入滤波锐化算法缓解由于图像过曝产生的模型误检和漏检等问题。

In order to solve the problems in impurity detection on desliming screen, an impurity detection method under open scene was proposed. Aiming at different types of impurity on the screen images, a impurity detection model was constructed based on the current mature object detection network framework, and the effectiveness of the method was verified through quantitative experimental analysis. On this basis, through qualitative analysis, the influence of different lighting conditions on the model detection effect was explored. In view of the local bright spots in the image, the filter sharpening algorithm was introduced to alleviate the problems of model false detection and missing detection caused by image overexposure.