果实摘取机械手的果实识别方法与摘取路径规划

Fruit recognition method and path planning for fruit picking manipulator

摘要:

通过图形分割处理技术从采集到的原始图形中分割出目标果实图形,再利用由遗传算法优化后的神经网络算法构建果实图形识别模型,完成对果实图形的识别。使用图像识别技术识别出机械手当前视线中的果实数量和位置,利用遗传算法对机械手的运行路径进行优化,并将其结果与随机路径规划和人工路径规划的结果进行对比。研究结果表明：随着样本中的完整果实比例逐渐减少,果实识别模型的识别准确率均有所下降,在仅有60%完整果实比例的样本中所研究的识别模型仍保持较高的识别准确率。使用遗传优化算法得到的机械手行进路径相比随机路径规划和人工路径规划所消耗的成本更低。并且随着果实数量的增加,遗传优化算法得到的机械手行进路径消耗的成本低的优势更加突出。

Abstract:

Through graphic segmentation target graphics graphics processing technology from the original fruit collected in the construction of fruit pattern recognition model by neural network algorithm genetic algorithm optimization, to complete the identification of fruit pattern. The use of image recognition technology to identify the number and location of manipulator in the current view, using the genetic algorithm running path of the manipulator is optimized, and compared the results with the results of random path planning and path planning manual. The results show that with the samples of intact fruits decreased fruit identification recognition model of the accuracy rate declined, with only 60% full fruit percentage of the sample identification model of this study still maintain a high recognition accuracy rate. Compared with random path planning and artificial path planning, the cost of running path of robot manipulator obtained by genetic algorithm is lower. And with the increase of the number of fruits, the advantage of the genetic algorithm for the path cost of the manipulator is more prominent.