基于双目视觉的连铸生产线铝锭垛定位技术研究

发布时间：2018-04-26 06:15

本文选题：双目立体视觉 + 摄像机标定　；参考：《兰州理工大学》2016年硕士论文

【摘要】：铝是需求量仅次于钢铁的有色金属材料,在国民经济中处于重要地位。近年来,随着电解铝产量的逐年递增,对高速铝锭连续铸造生产线的市场需求也日益迫切。在铝锭铸造生产线上,铝锭垛需要在打包、计量和打标工位启停,由于每垛铝锭重量高(重达1吨左右)、惯性大且非规则,生产线的链式传动导致铝锭垛在这些工位定位困难,目前基于到位信号检测控制铝锭垛停止的策略难以满足高速铸造生产需求;另一方面,在某些工位,铝锭垛定位不仅仅是测量位置信息,也需要测量铝锭垛的空间位姿信息。鉴于以上原因,研究铝锭垛静止状态的二次定位方法具有重要的应用价值。本论文针对铝锭垛的静止状态的二次定位需求,提出采用双目立体视觉测距定位的方法。第一,首先确定了双目视觉测量方案,采用Zhang摄像机标定法对摄像机进行标定,以求解摄像机内参数;根据摄像机光轴上点的坐标固定特性,摄像机外参进行左右摄像机分别独自标定,利用线性方程求解摄像机的其余参数,完成对两台摄像机的基线长度、光轴夹角和坐标系相对位置的标定。第二,针对采集到铝锭垛图像特征点提取与匹配算法,选择了比SIFT算法性能更好的SURF为基础算法,依据铝锭垛面高反光、纹理复杂的特点,提出把打包铝锭垛钢带作为特征点提取的主要依据,在改进SURF特征点提取的基础上提出采用KD-tree树索引,对提铝锭垛钢带上的特征点进行距离约束,最终得到特征点。特征点匹配基于Harris角点匹配方法,基于先验知识提出基于Harris角点和改进的SURF特征提取算法相结合的立体匹配算法。第三,为解决实际测量测距光轴可能不平行问题,基于两台摄像机坐标系之间的平移转换和两台摄像机光轴之间夹角转换,对平行式立体视觉模型的测距公式进行了修正。最后,通过实验验证了本文提出的方法的有效性,可以满足铝锭垛定位要求。
[Abstract]:Aluminum is a non-ferrous metal material in demand second only to iron and steel, and it plays an important role in the national economy. In recent years, with the increasing production of electrolytic aluminum, the market demand for high speed aluminum ingot continuous casting production line is increasingly urgent. In the aluminum ingot casting production line, the aluminum ingot stacking needs to start and stop at the packing, metering and marking stations. Because of its high weight (up to 1 ton weight, large inertia and irregular inertia), the chain transmission of the production line causes the aluminum ingot stacking to be difficult to locate in these stations. At present, the strategy of controlling the stop of aluminum ingot stack based on in-place signal detection is difficult to meet the demand of high-speed casting production. On the other hand, in some stations, the location of aluminum ingot stack is not only measured position information, but also needs to measure the space position information of aluminum ingot stack. In view of the above reasons, it is of great value to study the secondary positioning method for the static state of aluminum ingot stacking. In this paper, a binocular stereo vision location method is proposed to meet the secondary positioning requirement of aluminum ingot stacking. Firstly, the binocular vision measurement scheme is determined, and the camera is calibrated by Zhang camera calibration method to solve the internal parameters of the camera, according to the fixed characteristics of the coordinates of the points on the camera axis, the camera is calibrated by the Zhang camera calibration method. The external parameters of the camera are calibrated independently, and the other parameters of the camera are solved by linear equation. The baseline length, the angle of the optical axis and the relative position of the coordinate system are calibrated for the two cameras. Secondly, according to the feature point extraction and matching algorithm of aluminum ingot stacking image, the SURF algorithm, which has better performance than SIFT algorithm, is selected as the basic algorithm, according to the characteristics of high reflection and complex texture of aluminum ingot stack surface. Based on the improved SURF feature point extraction, a KD-tree index is proposed to restrict the distance between the feature points on the aluminum ingot stacking steel strip, and finally to obtain the feature points. Feature point matching is based on Harris corner matching method. Based on prior knowledge, a stereo matching algorithm based on Harris corner and improved SURF feature extraction algorithm is proposed. Thirdly, in order to solve the problem that the optical axis of measuring ranging may not be parallel, the ranging formula of parallel stereo vision model is modified based on the translation conversion between two camera coordinate systems and the angle conversion between the optical axes of two cameras. Finally, the effectiveness of the proposed method is verified by experiments, which can meet the requirements of aluminum ingot stacking location.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TG292;TP391.41

【相似文献】