设计了一种基于平面镜的低成本多视角投影成像结构，用于获取鲜食葡萄果穗不同侧面信息来判断果穗形状、颜色是否合格。采用两片前表面平面镜来延伸单目相机的拍摄视野，以最大限度地获取果穗全表面信息。通过图像处理方法分割果穗的实像虚像区域，并借助悬挂果穗的高度不变属性对虚像区域进行放大，得到三个每隔120°的同样高度果穗。提取果穗区域、轮廓宽度曲线参数来对果穗外观形状进行评价，与人工分级对比，对果穗形状分级的准确率为95.5%；将彩色图像RGB转换到HSI颜色空间，从色度图像(Hue)获取果穗成熟时的典型颜色区域，计算果面着色面积比例，并按照着色面积比例的大小进行颜色分类，准确率为81.1%。结果表明多视角同时成像的方法可用于葡萄外观的分级，为在线检测提供参考。

A low-cost multi-perspective projection imaging system was developed based on a plane mirror, to study table grape clusters and determine the quality of the cluster shape and color. In this system, the field of view of the monocular camera was expanded using two front-surface plane mirrors and optical information of the whole surface of the grape cluster was collected. Through image processing methods, the virtual/real image regions of the cluster were segmented, the virtual regions were enlarged based on the same height of the hanging grape clusters, and three clusters at the same height were obtained at 120° intervals. The parameters of cluster regions and contour widths were extracted to assess the external shape of grape clusters, and 95.5% accuracy was achieved for grading of the grape clusters compared to manual grading. The hue, saturation, and intensity (HSI) color space was transformed from the red, green, blue (RGB)-colored image, the typical color region of maturity was obtained from Hue channel image, and the ratio of colored surface area of grape was calculated. The color of the grape cluster was classified according to the ratio of colored cluster area, with an accuracy rate of 81%. The results indicate that this multi-perspective approach to simultaneously capture images can be applied to grade the external quality of grapes and can be applied to on-line shape and color measurement.

“十二五”国家科技支撑计划项目(2012BAD29B04-4)；江苏省普通高校研究生科研创新计划(KYLX_1068)；江苏大学高级人才科研启动基金(15JDG056)