基于深度学习的作物杂草识别研究

Deep learning based crop weed identification research

本设计基于深度学习针对作物杂草进行识别,首先是杂草的数据集的采集及预处理。数据集归一化为像素256*256的RGB图像,然后对这些数据进行分割和数据增强处理。经过交叉验证调整模型的参数后,模型评估的准确率达到了93%以上。这些技术的应用将会使农业生产率大幅度地提高,进一步实现农业地机械化和智能化水平,对于我国尚不发达地农业产生重大地意义。

This design is based on deep learning for crop weed recognition,starting with the acquisition and pre-processing of the weed dataset.The dataset was normalized to RGB images with 256*256 pixels,and then segmentation and data enhancement were performed on these data.After cross-validation to adjust the parameters of the model,the accuracy of the model evaluation reached more than 93%.The application of these technologies will lead to a significant increase in agricultural productivity,further mechanization and intelligence of agriculture,and will have a significant impact on the underdeveloped agriculture in China.