基于機器視覺的葡萄采摘域計算方法研究

摘要 鮮食葡萄采摘依靠人工，存在作業強度大、勞動力緊缺、生產成本高等問題，制約了產業規模化發展，亟需研制智能采摘設備提高生產效率。葡萄果實串采摘域的準確定位，是采摘設備視覺系統研究的重要內容。基于顏色空間和霍夫直線檢測算法融合的方法，研究了棚架式葡萄圖像目標提取和采摘域計算。分析葡萄果實串圖像顏色特征后分割出目標區域，對目標區域開展形態學算法處理，計算出圖像質心和最小外接矩形，以矩形與質心線的交點、矩形長寬等為參數構造出果梗感興趣區域，分割提取果梗目標，采用霍夫直線檢測算法計算區域內直線，以最長的直線區作為果梗采摘區域。根據測試結果，果梗感興趣區域圖像分割正確率為96%，果梗提取正確率為92%，采摘域計算正確率為92%，采用的算法數據處理量小、計算速度快，可以作為研制葡萄采摘機器人、產量測算、長勢監測等設備的理論基礎。

關鍵詞 機器視覺；葡萄果梗；顏色空間；采摘域

中圖分類號 S225 文獻標識碼 A 文章編號 0517-6611（2024）17-0224-04

doi：10.3969/j.issn.0517-6611.2024.17.051

Research on Grape Harvesting Domain Calculation Method Based on Machine Vision

MA Cong， CHEN Xue-dong

（Institute of Agricultural Economy and Information Technology， Ningxia Academy of Agriculture and Forestry Sciences， Yinchuan， Ningxia 750002）

Abstract Fresh grape picking relies on manual labor， which has problems of high workload， labor shortage and high production costs， which restricts the large-scale development of the industry. It is urgent to develop intelligent picking equipment to improve production efficiency. The accurate positioning of the picking domain for grape fruit clusters is an important aspect of the visual system research for picking equipment. Based on the fusion of color space and Hough line detection algorithm， we studied the target extraction and picking domain calculation of trellis grape images. After analyzing the color characteristics of grape fruit string images， the target area was segmented. Morphological algorithms were applied to the target area to calculate the centroid and minimum bounding rectangle of the image. The region of interest for the fruit stem was constructed using the intersection point of the rectangle and the centroid line， as well as the length and width of the rectangle. The fruit stem target was segmented and extracted. The Hough line detection algorithm was used to calculate the line within the region， and the longest line area was used as the fruit stem picking area. According to the test results， the accuracy rate of image segmentation in the region of interest of the fruit stem was 96%， the accuracy rate of fruit stem extraction was 92%， and the accuracy rate of picking domain calculation was 92%. The algorithm used had small data processing capacity and fast calculation speed， which could serve as the theoretical basis for developing grape picking robots， yield calculation， growth monitoring and other equipment.

Key words Machine vision;Grape stem;Color space;Picking domain

基金項目 寧夏農林科學院科技創新引導項目“寧夏釀酒葡萄智慧種植關鍵技術研究與示范”（NKYG-23-02）；寧夏自然科學基金項目“基于機器視覺的葡萄免碰采摘路徑規劃方法研究”（2023AAC03408）。

作者簡介 馬聰（1987—），女，回族，寧夏青銅峽人，助理研究員，碩士，從事農業信息化研究。

收稿日期 2023-09-21

葡萄味道鮮美、富含維生素，副產品也頗受市場消費者認可，是常見的水果種類之一。由于葡萄果樹產量高、適應性強、收益好，因此我國葡萄果樹的種植面積逐年上升。……