球形水果表面斑块成像失真度控制与实验

Image Distortion Control and Experiment of Spherical Fruit Surface Plaque

针对机器视觉在线分选时球形水果表面缺陷斑块在图像中形状、尺寸失真的问题,依据射影几何原理和通用成像配置,建立了斑块的成像普遍方程及斑块成像面积的通用算法,分析了斑块图像的面积失真规律;利用Matlab建立了满足特定分选准确度要求下有效采像区域与成像系统参数的配置关系,提出了一种有效采像区域的简化划定方法;分别以半球块和埃及甜橙为实验对象,利用NI Vision Assistant对工业相机采集到的、位于划定区域内的缺陷斑块进行了面积失真度校验。实验结果表明,特定采像条件下,当失真度阈值 A 为20%时,求得最小球心角λ 0 min 为79.32°来划定有效采像区域,采集到的5个半球块位于不同采像工位及其对应的采像姿态下的图像,有效采像区域内各缺陷斑块面积失真度均小于20%,满足失真度控制要求;同样地,5个埃及甜橙样本中,满足失真度要求的缺陷斑块比例分别达:97.38%、97.33%、94.22%、96.44%、97.78%。可见,利用球形水果表面缺陷斑块成像普遍方程预测结果进行图像的有效采像区域分割,即可实现对分选准确率的有效控制;通过测定有效采像区域中缺陷斑块面积进行分选阈值判断,有利于简化图像处理计算,为球形水果成像系统成像参数配置提供理论指导.

The surface defect plaque of spherical fruit is an important criterion for grading and sorting. Given the problem of the shape and size distortion of the plaques in the image by machine vision online sorting, according to the principle of projective geometry and general imaging configuration, the general imaging equation and the universal imaging area algorithm for plaque were established, and the area distortion law of plaque image was analyzed. Based on these above, the relationship between the effective image region and the parameters of the imaging system was calculated by Matlab to meet the requirements of the specific sorting accuracy, and a simple method for defining the effective image region was proposed. Using hemispheric blocks and Egyptian sweet oranges as the test subjects, the area distortion of the defective plaques located in the demarcated area captured by industrial cameras were verified with NI Vision Assistant. Assuming the distortion threshold ( A ) was20%, the minimum spherical center angle (λ 0 min ) could be calculated and the result was 79.32°, and then the effective image area was designated. The experiment showed that the distortion degree of the defect plaques in the delineated area on the five hemisphere blocks were all less than the setting threshold under the specific image conditions. Similarly, for the five Egyptian sweet orange samples, the proportion of defective plaques that satisfied the distortion requirement were orderly 97.38%, 97.33%, 94.22%, 96.44% and 97.78%. Thus it s available that the predicted results based on the general equation of spherical fruit surface plaque imaging can be used for the effective region segmentation to realize the effective control of sorting accuracy. The sorting threshold judgment by determining the area of defective plaques in the effective image region was conducive to simplify image processing, which can provide theoretical guidance for the imaging parameter configuration of the spherical fruit imaging system.