基于双目立体视觉技术的成熟番茄识别与定位

Recognizing and locating ripe tomatoes based on binocular stereovision technology

研究了基于双目立体视觉技术的成熟番茄的识别与定位方法,获取了成熟番茄的位置信息,用于指导温室内成熟番茄的自动化采摘作业。该方法利用成熟番茄与背景之间颜色特征的差异进行图像分割来识别成熟番茄;根据图像分割的结果,用形心匹配方法获取番茄中心的位置;然后根据相邻区域像素点灰度的相关性,利用区域匹配方法计算番茄表面各点的深度信息。使用限制候选区域和两次阈值分割的方法减少计算量,提高了计算精度。实验结果表明,工作距离小于550mm时,番茄深度值的误差约为±15mm。利用该测量方法可以实现成熟番茄位置信息的获取,测量精度较高。

A method for recognizing and locating ripe tomatoes based on binocular stereovision technology was developed for tomato harvesting in greenhouse. The color features were used to recognize ripe tomatoes by segmenting target objects from background. According to the results, the position of the fruit centroid was acquired with centroid-based matching. Then in terms of grey correlation of neighborhood regions, depths of the tomato surface points were calculated with area-based matching. And the method of limited candidate region and twice thresholds was used to reduce computation load and improve precision. The error ranged within ±15 mm when the distance was less than 550 mm. It＇s feasible to obtain position information of ripe tomatoes with this measurement, and the measurement precision is high.