摘要
为了防治储粮害虫带来的危害,借助计算机对储粮害虫进行有效的图像识别是具有重要意义的。针对基于图像的储粮害虫多分类识别问题,引入了基于深度卷积神经网络的储粮害虫图像识别方法。该方法与传统的储粮害虫识别方法相比,大幅度简化了数据预处理过程,同时该方法在识别精确度方面达97.61%,也明显优于传统方法。因此,基于深度卷积神经网络的储粮害虫识别方法具有较高的实用性,且具有进一步研究和推广的意义。
In order to prevent and control the damage caused by stored grain pests, it is of great significance tofind an effective way of recognizing stored grain pests with the help of computer. Focusing on the problems ofmulti-class recognition of stored grain pests based on images, a new image recognition method of stored grainpests based on deep convolutional neural network was proposed. Compared with the traditional ways of storedgrain pests recognition, this new method greatly simplified the data preprocessing process, and the accuracy ofthis new method achieved 97.61%, which was significantly superior to the traditional methods. Therefore, themethod of recognizing stored grain pests based on deep convolutional neural network has high practicability aswell as the significance of further research and promotion.
出处
《中国农学通报》
2018年第1期154-158,共5页
Chinese Agricultural Science Bulletin
基金
安徽省自然科学基金项目"基于基因表达式编程的作物生长建模方法研究"(1508085MF110)
茶树生物学与资源利用国家重点实验室开放基金"基于茶树转录组装优化后的中国变种与阿萨姆变种SSR与SNP的特征比较研究"(SKLTOF20150103)
