Identification Of Grape Varieties Based On Leaves Image Analysis

Grape varieties are very complex which makes its identifying difficult. Relying on people to carry out a variety of identification is time-consuming and impractical, because it requires the operators have a wealth of knowledge about the grape classification and long-term practical experience which has subjectivity due to the large workload. Identificating the grape varieties based on image analysis skills which could help the operators reduce the time of identification, and at the same time help those operators who do not have the complete knowledge of the grape. In this paper, the image of the grape leaves was preprocessed firstly, and secondly features extraction was performed on the feature of artificial designing and deep learning. Finally, the recognition model was established by using support vector machine classifier, and the extracted feature vectors were used as input parameters which provided a new method for the identification of grape varieties. The main research contents of grape leaf variety recognition based on computer image analysis are as follows:(1) The method of preprocessing for grape leaves. To analyze the image of grape leaf samples obtained from scanning, and then according to the difference of pixel between target and background, the method would selecte the gray threshold to segment and use the minimum external rectangle to extract the image. We also used the method of morphology, gray scale and geometric variation to preprocessing which had a solid foundation for features extraction.(2) The artificial features and the convolutional neural network features for recognition of grape leaves. After analysising on the morphology and the structure of grape leaves, we introduced the artificial features of gradient histogram and the convolutional neural network features to recognize different grape leaves. The improved gradient histogram features could be efficient and effective to abstract the leaves of edge contour and texture characteristics. The convolutional neural network could learn the characteristics of grape leaves independently, and the information was complete and high dimension semantic expression.(3) According to the characteristics of grape leaves, the identification model was establised based on support vector machine. The experiments were carried out in Matlab 2010 b environment, and especially we did the experiments about the characteristics of the artificial features and neural network features. Experimental results showed that the recognition rate of orientation gradient histogram features was 86.67% and the recognition rate of deep learning features was 88.33%.In conclusion, through the analysis of the morphological characteristics of grape leaves, a method of using high dimensional artificial design features and deep learning characteristics to represent the grape leaves was proposed and using support vector machine model to identify the varieties of grape leaves had excellent performance on the algorithm which providing a reference method for the identification of different kinds of plants in the same subject.