A Study Of Image Processing Methods Related To Plant Leaf Recognition

Plant leaf is one of the important characteristics of plant.And most plants can be identified by their leaves.The plant leaf automatic recognition through computer image processing technology is a convenient and fast plant recognition method,which can help people to distinguish and identify plants.In this paper,some image processing technologies related to plant leaf automatic recognition are studied.The main research contents as followings:(1)Extracting edge information of plant leaves.Consider that the global character of two high and low global thresholds is deficiency of traditional Canny edge detection algorithm.,The adaptive global segmentation threshold of the gradient magnitude image is obtained by using the Otsu method,and then the ratio of the number of pixels that is larger than the threshold to the total number of all pixels in the image are calculated as the global proportion.Secondly,In the vertical and transverse directions of the image,one pixel is chosen as the calculation point at the same interval.The local high and low thresholds of the point are calculated to ensure that the ratio of the number of points larger than the local high threshold in the local area to the total number of points in the region is equal to the global proportion,and low threshold equals half of high threshold.The experimental results show that the algorithm can detect more detail edges while avoiding false edges.(2)Extracting plant leaf veins based on local fuzzy c-means.When using global fuzzy c-means to extract leaf veins automatically,it is sensitive to light and has a high misclassification rate.To overcome this shortcoming,an algorithm for extracting plant veins based on local fuzzy c-means is proposed.The experimental results show that the proposed algorithm can not only extract more complete leaf veins,but also avoid the influence of illumination effectively.(3)Research on plant leaf recognition based on VGGNet-16 convolution neural network.A new network model for plant leaf recognition is constructed through modifying the VGGNet-16 network model appropriately.Using this network model,the color image,edge image and vein image of plant leaves are trained and tested respectively.The experimental results show that the accuracy of color image recognition is 91.48%,that of edge image recognition is 94.92%,and that of leaf vein image recognition is 92.81%.