基于手持移动设备贴近摄影获取排土场物料粒度分布

Obtain particle size distribution of dump material based on handheld mobile device close-range photography

采用手持移动设备贴近摄影获取露天煤矿排土场物料影像,基于深度学习关键点检测算法识别并提取影像中的物料粒径,据此分析排土场物料粒度分布规律。改进算法精度优于形态学与深度学习分割算法,在测量研究区排土场砾岩、黏土物料粒度时分别可达2.10%和2.84%的误差率,提升测量精度的同时大大减少了分割算法造成的标注成本。基于手持移动设备贴近摄影的排土场物料粒度数据采集方法粒径检出限约为20~30mm之间,便携易用的同时规避了其他摄影测量方法的环境与安全限制。应用此算法和摄影测量手段,探讨了不同材质、高度下研究区排土场物料粒度组成特征;总结出不同材质、高度下研究区排土场物料粒度分布规律,且R-R分布函数可以很好地描述排土场物料粒度分布,一元三次多项式能够表明排土场物料粒度R-R分布函数参数随物料高度的变化关系。

The handheld mobile device is used to obtain the material images of the open-pit coal mine dump by close-range photography,and the material particle size in the images is identified and extracted based on the deep learning keypoint detection algorithm,according to which the material particle size distribution of the dump is analyzed. The accuracy of the improved algorithm is better than morphological and deep learning segmentation algorithms,and the error rates of2. 10% and 2. 84% can be achieved in measuring the particle size of gravel and clay materials in the dump of the study area,respectively,which improves the measurement accuracy and dramatically reduces the labeling cost caused by the segmentation algorithm. The handheld mobile close-range photography method with a particle size detection limit of 20-30mm is portable and easy to use while avoiding other photogrammetry methods' environmental and safety restrictions. By applying this algorithm and photogrammetry,the particle size composition characteristics of the material in the study area are explored under the influence of different materials and heights. In addition,the R-R distribution function can describe the material size distribution of the dump well,and the unitary cubic polynomial can show the relationship between the parameters of the R-R distribution function of the material size of the dump and the material height.