面向危化品仓储堆垛三维几何重建的单目视觉标定算法研究

为了解决安全生产部门对危化品仓存储缺乏有效技术手段的问题，构建了五元组参数单目视觉成像模型并提出一种相机标定方法。该方法无需已知相机内参数信息，仅通过2个标识点成像关系就可实现参数标定，避免了传统算法实现过程复杂、运算量大等问题。标定得到的参数作为堆垛三维几何重建运算的基本条件，重建的效果则客观评价了标定参数的精确性。试验结果表明:提出的标定算法易于实现，其结果在堆垛几何重建过程中有较强的鲁棒性，重建结果与实际情况比较误差很小，适用于危化品仓储五距的监管。

Research on monocular vision calibration algorithm for 3D geometrical reconstruction of hazardous chemicals storage stacking

To solve the problem about lacking of effective technical means on storage of hazardous chemicals for the work safety departments, a monocular vision imaging model with quintuple parameters was established, and a camera calibration method was proposed. This method can achieve the parameters calibration only through the imaging relationship of two identification points, without the known intrinsic parameters information of the camera, so it avoids the complex operation process, the large amount of computation and other problems in the traditional algorithm. The parameters obtained from the calibration were taken as the basic conditions of the 3D geometrical reconstruction computation on stacking, and the effect of reconstruction evaluated the accuracy of the calibration parameters objectively. The experimental results showed that the proposed algorithm was easy to implement, and the results were robust in the process of geometrical reconstruction. The reconstruction results were very close to the real situation. It is suitable for the supervision on the five distances in storage of hazardous chemicals.