基于深度学习的目标橘子识别方法研究

Research on the method of identifying target orange with deep learning

为了更好地解决自然条件下目标橘子的遮挡、重果问题,采用深度学习的方法对目标橘子进行识别,并用传统的目标识别算法与Faster-RCNN两种方法进行对比实验。根据大量的数据对比可知,传统的目标识别方法对自然光照敏感,对遮挡、重果的识别效果不佳,泛化能力及鲁棒性较差。而Faster-RCNN算法对光照及枝叶遮挡的识别更友好,更符合采橘机器人实际采摘的需要。深度学习方法有望在采橘机器人目标识别中得到更广泛的应用。

In order to better solve the problem that the target orange is covered by something or overlapped in natural condition,this paper uses the deep learning method to recognize the target orange,and makes the comparative experiment between the traditional target recognition algorithms and the Faster-RCNN method.According to a large number of data comparison,the traditional target recognition methods are sensitive to natural light,the recognition effect on covered or overlapped fruit is not so good,and have poor generalization ability and robustness.The Faster-RCNN algorithm is more suitable to recognize the light and branch covered fruit,which is more in line with the actual needs of orange picking robot.Deep learning methods are expected to be more widely used in the target identification of orange picking robot.