基于深度学习的白菜田杂草分割

田间除草在农业生产中具有重要意义,传统杂草识别方式具有效率低或者局限性大的缺点。为此提出将Mask R-CNN算法应用到自然光照下杂草幼苗和白菜幼苗的图像识别。实验选取常见杂草幼苗和白菜幼苗图像作为训练集训练网络,经测试集测试得到81%的合格率。与传统阈值分割算法对比,Mask R-CNN在不同环境下都能精确地识别出杂草幼苗,解决了传统图像算法在复杂光照和叶片遮挡环境下图像难以分割的问题,并且避免了分类器设置工作。

Weed Identification in Cabbage Field Based on Deep Learning

Field weeding is of great significance in agricultural production,and traditional weed identification methods have the disadvantage of low efficiency or large limitations.To this end,deep learning is applied to the visual recognition system,and Mask R-CNN algorithm is applied to the image recognition of weed seedlings and cabbage seedlings under natural light.The experiments selected common weed seedlings and cabbage seedling images as the training set to train the network,and the test set test yielded a pass rate of 81%.Compared with the traditional threshold segmentation algorithm,Mask R-CNN can accurately identify weed seedlings in different environments,solving the problem that traditional image algorithms have difficulty segmenting images under complex lighting and leaf occlusion environments,and avoiding classifier settings jobs.