Description: Plant based on digital image recognition is a hotspot of research on automatic classification.But with the increase of plant species, the traditional classification method by the extraction of characteristics or more single classifier structure is too simple, leading to a lower leaf recognition rate.To this end, this paper proposes using the texture characteristics in combination with characteristics of shape, which can identify the belief network architecture and using the depth as a classifier.Texture characteristics by local binary pattern, Gabor filter and gray level co-occurrence matrix method.And shape characteristic vector by Hu s invariant and the Fourier descriptor.In order to avoid over fitting phenomenon, dropout method is used to train deep belief networks.This belief network based on feature fusion depth plant identification method
To Search:
File list (Check if you may need any files):
基于多特征融合和深度信念网络的植物叶片识别_刘念.pdf
Download