基于机器视觉的杏鲍菇外观品质分级系统设计

目的:将机器视觉技术引入到杏鲍菇的外观检测中。方法:提出一种双边滤波代替高斯滤波作为图像平滑滤波器，Ostu最大类间方差法代替固定双阈值分割的改进型Canny算子，并作为边缘检测算法。利用HALCON算子和颜色空间转换，完成杏鲍菇的长度、直径、弯曲度、均匀度、色泽及菇帽缺损特征要素提取。使用HALCON 18.05联合C#在VS 2017开发环境下完成视觉软件功能模块开发设计。结果:随机获取200根杏鲍菇，对算法处理准确度和视觉软件工作性能进行检测。杏鲍菇直径分级精度为83%,其余特征要素可达95%以上，装置整体各规格杏鲍菇分级精度达90%以上。结论:通过算法的改进和视觉软件的设计可以完成杏鲍菇的外观品质的分级。

Design of appearance quality grading system for apricot mushroom based on machine vision

Objective: This study aimed to induce the machine vision technology into the appearance detection of Pleurotus eryngii. Methods: A bilateral filter was proposed to replace Gaussian filter as image smoothing filter, and Ostu maximum inter-class variance method was proposed to replace the improved Canny operator, based on fixed double threshold segmentation, and used as edge detection algorithm. HALCON operator and color space conversion were used to extract the length, diameter, curvature, evenness, color and cap defects of P. erynii, and the development and design of visual software function modules were completed, under the VS 2017 development environment with HALCON 18.05 and C#. Results: 200 pieces of P. eryngii were randomly obtained to test the accuracy of the algorithm and the performance of the visual software. The diameter grading accuracy of the Pleurotus eryngii was 83%, and the remaining characteristic elements could reach more than 95%, with the overall classification accuracy of all specifications more than 90%. Conclusion: The classification of appearance quality of P. eryngii can be completed through the improvement of algorithm and the design of visual software.