基于机器视觉的水果尺寸检测误差分析

Analysis of errors in fruit size inspecting based on machine vision

介绍了当前应用机器视觉进行水果尺寸检测的现状。根据水果成像时水果、摄像机透镜、水果图像三者之间的相互关系 ,运用几何光学理论分析了尺寸检测中的各种误差及其原因。水果成像时 ,由于水果表面各点的高度变化 ,水果图像上各点所代表的实际长度不尽一致 ,形成标定误差 ;水果与摄像机透镜光心之间的距离不可能无穷远 ,成像后 ,水果图像的边缘点到形心的距离并不能真正代表水果的半径 ,形成半径误差 ;水果中心与摄像机光心偏离后 ,得到的图像存在形状误差。给出了标定误差的计算公式和半径的估算公式。

Research advancements on fruit size inspecting with machine vision was introduced. The relationship among fruit, camera lens and fruit image was studied based on geometric optic theory. Pixels in the fruit image stand for varied actual size because of varied positions of points on the fruit surface, which gave rise to calibration errors. The distance between camera and fruit was limited and not long enough to obtain the actual radius from the image, which caused radius errors. Since the location of the fruit relates to the core of the camera, the shape on the image was varied, which resulted in shape errors. Three formulas were set up to calculate the errors respectively.