基于透视变换和SNc-YOLOv5的大块煤识别方法应用  

Application of large block coal recognition method based on perspective transformation and SNc-YOLOv5

作　　者：于涛 YU Tao(National Energy Group Xinjiang Energy Co.,Ltd.,Xinjiang Urumqi 830011,China)

机构地区：[1]国家能源集团新疆能源有限责任公司,新疆乌鲁木齐830011

出　　处：《西安科技大学学报》2024年第1期54-63,共10页Journal of Xi’an University of Science and Technology

基　　金：国家自然科学基金项目(51874231)。

摘　　要：为了提高带式输送机大块煤识别的准确率,避免大块煤堆积对皮带造成的磨损,延长输送机的使用寿命,提出了基于透视变换和SNc-YOLOv5的大块煤识别方法,该方法首先利用透视变换技术对原始图像进行处理,有效地将图像中的非输送机区域滤除,并对输送机区域进行坐标矫正;然后,采用SNc-YOLOv5模型对经过标准化处理的图像进行深度学习建模,得到大块煤识别模型;最后,通过试验分析和现场应用对该方法进行有效性验证。结果表明:该方法在1号煤矿数据集的试验分析精确率为94.8%,召回率为83.2%,在2号煤矿数据集的试验分析精确率为92.8%,召回率为85.3%,现场应用置信度达到0.9,与其他方法进行比较,精确率和召回率指标均优于其他方法;该方法在图像预处理阶段提取带式输送机区域对图像进行标准化,仅对感兴趣区域进行处理,提高了大块煤识别的准确率。该算法部署到某煤矿现场,能够实现实时监测,为带式输送机的安全运行和延长使用寿命提供了有力保障。To enhance the precision of identifying large coal chunks in belt conveyors,mitigate belt wear from coal accumulation and increase the longevity of belt conveyors,a novel method rooted in perspective transformation and SNc-YOLOv5 was proposed.Firstly,the perspective transformation was utilized to refine the original image,successfully omitting areas unrelated to the conveyor and adjusting the conveyor’s coordinates.Then,the SNc-YOLOv5 model was undertaken to establish deep learning model on these standardized images,yielding a model adept at recognizing large block coal.The efficacy of this approach was confirmed through both test analysis and on-site application.The results show that the precision rate on the dataset of Mine 1 is 94.8%,with a recall rate of 83.2%,while on Mine 2’s dataset,the precision rate is 92.8%,with a recall rate of 85.3%.In field applications,the confidence level reaches 0.9,compared with alternative techniques,this method outperforms in terms of both accuracy and recall.By focusing on the conveyor region during image preprocessing,the method not only standardizes the image but also elevates the precision of large block coal identification,which thus continuous real-time surveillance,fortifying the safety and longevity of belt conveyors.

关 键 词：大块煤识别 带式输送机 透视变换 图像标准化 

分 类 号：TD712[矿业工程—矿井通风与安全]