基于改进卷积网络的葡萄剪枝关键算法研究

Research on the key algorithm of grape pruning based on improved convolution network

为进一步对葡萄的剪枝过程进行优化,基于改进的卷积神经网络,对葡萄剪枝的关键算法进行研究。其中,主要对葡萄剪枝过程的剪枝点识别进行研究,采用卷积神经网络中的YOLOv3模型作为基本的识别方法,并通过对模型的网络结构以及损失函数进行改进,以提升模型的剪枝点识别率。实验结果表明,与其他识别方法相比,本研究提出的识别算法在进行剪枝点识别时能够取得更好的识别效果,在漏检率LPR、误检率FPR、检测率TPR上分别为10.5%、5.8%、89.5%;与改进前的YOLOv3模型相比,改进后的模型的图片处理过程更加优秀,F1值提高了5.43%。以上结果验证了本研究对模型的改进的有效性,同时也表明改进的剪枝方法适用于葡萄的剪枝。

To further optimize the grape pruning process,the key algorithms for grape pruning were studied based on the im-proved convolutional neural network.Among them,this study mainly studies the pruning point identification of grape pruning process,using the YOLOv3 model in the convolutional neural network as the basic identification method,and improving the network structure and loss function to improve the pruning point recognition rate of the model.The experimental results show that compared with other identification methods,the identification algorithm proposed in this study can achieve better identification results,with 10.5%,5.8%,and 89.5%respectively;compared,the image processing of the improved model is better,and the F1 value is increased by 5.43%.The above results verify the effectiveness of the present study on the model improvement,and also show that the improved pruning method is suitable for the pruning of grapes.