基于编码点阵结构光的苹果果梗/花萼在线识别

On-line Identification of Apple Stem-end/Calyx Based on Coded Spot-array Structured Light

为解决苹果机器视觉自动分级时果梗/花萼识别的难题,提出一种基于位置变化的点阵结构光编码方法,并用于苹果果梗/花萼的在线检测。通过分析投射在物体和参考平面上光斑的成像规律,提出将光斑的位置变化作为编码基元;在二元域中,利用编码基元生成M阵列,将其作为近红外点阵结构光的编码模式;通过分析匹配后的差值矩阵,识别果梗/花萼的位置。在线实验结果表明:该方法可以有效地实现果梗/花萼的在线识别,在满足实时性要求前提下,平均识别正确率可达到93.17%。

Automatic detection of apple defects using a computer vision system is difficult due to the similarity between stem-end / calyx and true defects. Identification of stem-end / calyx is always a challenging project in automatic apple grading. This paper presents an encoding method based on position change for spot-array structured light,and the proposed method was used in distinguishing defects from stem-end / calyx of apple images in real time. By analyzing imaging process of the spots projected onto the object surface and reference plane,the position change of spot was chosen as coded primitive. Over the field of two elements,the M-array was generated by using the primitives,and was used as the coded pattern of near-infrared（ NIR） spot-array structured light. Analysis of difference matrix made it possible to identify the location of stem-end / calyx regions after matching. The on-line experimental results demonstrated that the proposed method could realize the detection of stem-end / calyx efficiently,and achieve an average of 93. 17% recognition accuracy with real-time performance. The results indicated that the proposed method was effective in identification of apple stem-end / calyx.