Research On Plant Leaf Image Recognition Method Based On Jaccard

Plants are closely related to our life.The classification of plants plays an important role in the exploitation and protection of plant resources.Due to the variety of plants,it is difficult to classify and identify them quickly and effectively by traditional classification methods.With the rapid development of digital image processing and pattern recognition,more and more researchers pay attention to classification and identification of plant species.Compared with plant organs such as flowers,fruits,and stems,leaves have a longer survival time,they are easy to collect and have a more stable shape.Therefore,leaves are usually used as the main reference organs for plant identification.The identification of plant leaf is of great significance to plant protection,ecological environment and agricultural production.This thesis studies the robustness of plant leaf identification features,and proposes an effective plant leaf identification scheme based on Jaccard coefficient and BOW model.The main work is as follows:1.The plant leaf recognition system is introduced from two aspects of feature extraction and classification recognition,and common feature extraction technology and classification technology are introduced respectively.The performance of the main feature extraction and classification techniques is evaluated.2.Evaluate the robustness of common leaf recognition features in non-ideal environments.The non-ideal conditions of the experiment include adding Gaussian noise,multiplicative noise and salt and pepper noise of different densities,light sources of different radius,intensity and position,and occlusion of different proportions and positions to the leaf image.And analyze the robustness of the features based on the experimental results.3.The texture feature BOW?Laws based on the BOW model and Laws texture energy measurement,and the contour feature BOW?Sobel based on the BOW model and Sobel operator are proposed,which describe the local features of the leaf image well.Compared with other existing features based on the BOW model,they have better information description capabilities.4.A two-stage plant leaf identification method based on Jaccard coefficient and BOW model is proposed.First,the Jaccard coefficient is used to calculate the similarity between the test image and the training image,and the candidate species with higher similarity are selected.Secondly,use the proposed texture feature BOW?Laws and contour feature BOW?Sobel to extract features from the leaf image and build a dictionary.Finally,support vector machines are used for classification and the final result is obtained.5.The parameter setting and noise robustness experiments of the proposed plant leaf identification method are performed,and it has better robustness in the salt and pepper noise environment.Experiments were performed on the Flavia,Swedish,ICL,MEW,and LZU datasets.The method proposed in this paper obtained good results on all datasets,and the recognition accuracy rates were99.7%,99.3%,94.5%,95.2% and 99.3%,respectively.Compared with existing algorithms,the method proposed in this paper has certain advantages.