果园环境下采摘机器人水果目标识别算法探究——以绿色阳光玫瑰葡萄为例

该文研究果园环境下绿色阳光玫瑰葡萄的果实图像分割和定位方法。在分析了顺光、逆光和夜间3种光照情况下的采集图像后,选取最能体现绿色葡萄果实的颜色分量作为分割算法的输入图像。利用最大类间方差法(OTSU)和支持向量机法(SVM)实现果实和背景区域的分割。实验结果对比表明,绿色葡萄在夜间的识别率高于晴天顺光和逆光的情况。比较2种算法的准确率,可以发现SVM算法在晴天顺光和逆光时的准确率更高,而OTSU算法在夜间情况时较高,达到了98.7%。

The method of fruit image segmentation and location of green sunshine rose grape(aka.Vitis Labrusca)in orchard environment was studied in this paper.After analyzing the captured images under the conditions of smooth light,backlight and night light,the color component which can best reflect the green grape fruit is selected as the input image of the segmentation algorithm.The maximum inter-class variance method OTSU and support vector machine method SVM are used to segment fruit and background regions.The experimental results show that the recognition rate of green grapes at night is higher than that of smooth light and backlight in sunny days.Comparing the accuracy of the two algorithms,we can find that the accuracy of SVM algorithm is higher in sunny days with smooth light and backlight,while OTSU algorithm is higher at night,reaching 98.7%.