针对长方体纸箱三维尺寸测量问题提出一种单目视觉方法, 利用虚拟像平面相机模型, 恢复出长方体在支撑平面上的虚拟投影图像, 利用支持向量机的方法从投影图像上判别出纸箱区域, 再利用一维Hough变换方法获取每条边缘线段的样本点集, 利用投影图像上的长方体边缘间存在的平行或垂直等几何约束关系, 通过最小二乘方法提取边缘直线方程, 利用长方体固有的几何约束关系计算其三维尺寸.实验结果表明, 该方法对于长方体三维测量结果误差在毫米量级, 平均相对误差在2％以内, 其测量精度不低于当前的最好方法, 在测量系统结构的简单性、自动化程度、鲁棒性等方面都具有优势.%A monocular visual method is presented to measure box dimensions. Using a virtual image plane camera model, a virtual projective image of the box on its support plane is firstly recovered from original image. Then the box region in the virtual image is extracted by Support Vector Machine (SVM) method. The candidate point set of each edge segment of the box is constructed by one dimensional Hough transform method. In the virtual projective image, edge segments in the box keep their parallel and perpendicular properties. Their line functions are extracted by Least Squared Method from the corresponding sample points.Then the box dimensions can be calculated in accordance with the geometric constraints. The simulation and real experiment results showed the measurement errors achieved millimeter level and the average relative error is less than 2％. Its measurement accuracy is not less than the known best methods. The method has the advantages of simple system structure, high degree of automation and robustness.
展开▼
