Plant Identification Based On Linear Features Of Leaves

Plant identification is of great significance in the fields of modern agriculture,Chinese herbal medicine,and plant classification.Leaves are an important feature of plant classification and morphological differences.Therefore,plant identification through leaf image recognition is an efficient and meticulous method.The current leaf recognition is usually based on one or two of the three linear features of leaf contour,texture and leaf vein in the leaf image,which greatly limits their recognition performance and application range.Based on the above problems,this paper introduces a new counting-based leaf recognition method that can directly and effectively combine three important features in leaf images to perform plant classification.At the same time,based on this,a more effective feature extraction algorithm:an elliptical half Gabor filter and a better linear structure descriptor:MGLLDP(Maximum Gap Local Line Direction Pattern)are proposed to solve the above problems.The main research work and results of the paper are as follows:(1)In order to obtain stable and independent local line responses from leaf contours,leaf textures,and leaf veins,this paper introduces an elliptical half Gabor filter and convolves with the original gray leaf image to effectively extract the various linear characteristics of leaves and the multi-directional response of curves.The experimental results show that the modified elliptical half Gabor filter has greatly improved the recognition rate compared with the traditional Gabor filter.(2)Extract the maximum gap local line pattern from the local line response through MGLLDP and perform a normalized circular right shift process until it shifts to the maximum bit plane with a value of 1 on the leftmost bit plane in the direction.After that,the normalized histogram is calculated and used as a local structure pattern based on counting,which more accurately expresses the structural characteristics of the linear direction of the leaves.It is verified by experiments that MGLLDP has a better accuracy than the traditional gradient operator in recognition rate.(3)After combining the elliptical half Gabor and MGLLDP algorithms,they are normalized,and then use support vector machines as classifiers in the three most commonly used databases in Sweden,Flavia,and ICL(Intelligent Computing Laboratory)conduct a comparative experiment.The recognition rate of our proposed algorithm on these three commonly used leaf libraries reaches 98.40%,97.83%and 97.37%respectively,which is superior to the current mainstream leaf-based plant recognition algorithms.In summary,in order to solve the actual plant classification problem based on the leaf image,combined with the traditional Gabor filter,local descriptor algorithm,and counting-based leaf recognition method.The author proposes several keys to solve the linear feature problem of leaves algorithm.These new algorithms have achieved good application results through experimental verification.Through experimental analysis,it is concluded that the success of the leaf classification recognition algorithm mainly depends on the feature extraction algorithm.Among them,comprehensive extraction of linear features,accurate expression of linear structure expression and how to avoid the robustness caused by point features are our focus.The core problem.We focus on the new algorithms proposed by these core problems to properly solve these difficulties and achieve good experimental results.