基于机器视觉的电铲斗齿脱落检测算法研究

发布时间：2018-10-26 06:39

【摘要】：电铲是一种广泛应用于露天矿产的重要采掘设备,其铲斗的斗齿在工作时因受到矿石物料较大的反作用力而容易发生脱落现象,脱落的斗齿连同矿石物料一同进入碎石机中会导致碎石机损坏,进而引发整条采掘——破碎生产线的停工,造成重大的经济损失。针对这一问题,本文利用红外机器视觉系统对电铲斗齿的工作状态进行监控,并结合图像处理技术对斗齿脱落的视觉检测算法进行了研究。主要内容如下:通过分析电铲的工作过程,利用电铲斗齿在工作时与矿石物料不断摩擦并向外辐射红外线的机理,采用红外热像仪对斗齿进行成像,以满足电铲昼夜不停的工作要求,同时能获得较高质量的监控图像。在检测算法上,先使用梯度方向直方图(HOG)特征结合支持向量机(SVM)的目标检测框架对红外图像中的斗齿对象进行目标检测;在此传统目标检测方法的基础上,通过分析斗齿图像在形状上的相似性关系,利用形状上下文(Shape-Context)算法提取斗齿的形状特征,并使用形状匹配方法向斗齿的目标检测过程添加形状特征约束,以提高目标检测的准确率。在准确检测出图像中的斗齿目标后,通过分析电铲铲斗与红外相机之间的相对运动关系,确定了斗齿在图像中空间位置关系的不变性,并以此为依据对斗齿脱落与否进行判断,最后,通过实验对整套检测算法进行了验证并分析了误差产生的原因,提出了一些改进思路和方法。
[Abstract]:Electric shovel is a kind of important mining equipment which is widely used in opencast mine. The bucket tooth of its bucket is easy to fall off because of the reaction force of ore material. The lost bucket teeth along with the ore material entering the stone crusher will lead to the damage of the stone crusher, which will lead to the stoppage of the whole excavation-crushing production line, resulting in great economic losses. In order to solve this problem, the infrared machine vision system is used to monitor the working state of bucket teeth, and the visual detection algorithm of bucket tooth loss is studied in combination with image processing technology. The main contents are as follows: by analyzing the working process of the electric shovel, using the mechanism that the bucket teeth of the electric shovel continuously rub with the ore material and radiate infrared ray, the infrared thermal imager is used to image the bucket tooth. In order to meet the working requirements of the shovel day and night, at the same time can obtain a higher quality of monitoring images. In the detection algorithm, the target detection framework based on gradient direction histogram (HOG) feature and support vector machine (SVM) is used to detect the object in infrared image. On the basis of the traditional object detection method, the shape feature of bucket tooth is extracted by using shape context (Shape-Context) algorithm by analyzing the shape similarity of bucket tooth image. The shape matching method is used to add shape feature constraints to the object detection process in order to improve the accuracy of target detection. After accurately detecting the bucket tooth target in the image, by analyzing the relative motion relationship between the shovel bucket and the infrared camera, the invariance of the spatial position relation of the bucket tooth in the image is determined. Finally, the whole detection algorithm is verified by experiments, the causes of error are analyzed, and some improved ideas and methods are put forward.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD422.21;TP391.41

【参考文献】