Research Of Extraction And Classification Of Plant Leaves Based On Active Contour Model In Complex Background

In this paper,the image segmentation and recognition algorithm of plant leaves with complex background collected in the natural environment are studied systematically and deeply.Plant leaves were obtained by photographing directly on an outdoor plant.The illumination,surrounding environment and overlapping of the leaves had an effect on the target leaves in the image.Therefore,the focus and difficulty work of image preprocessing were image denoising and image segmentation.In this paper,a geometric template for the initial contour curve of plant leaves is proposed for the improved snake model.The experimental results show that it is feasible and effective to apply the template to snake model for image segmentation.In order to highlight the effect of the template,this paper compares the accuracy of the recognition images with the improved snake model and the corresponding white background images.The experimental results show that the image segmentation effect of the snake model is better.Finally,the image segmentation and recognition system is designed.Details as follows:1)We studied the active contour model deeply in this paper.Considering the features of the image we collected,an elliptic parametric curve was proposed as the initial contour curve of the active contour model of the target leaf image,then image segmentation was run,using the improved snake model based on the traditional snake model.2)The shape,texture and invariant feature extraction were extracted from the leaf image which segment from nature background leaf image.According to the characteristics of the initial contour curve to generate a new feature attributes and then added to the shape feature.And then the decision tree,support vector machine were used to classify and identify,finally compared the classification accuracy.3)The system of plant leaf image segmentation and recognition based on matlab and java web was designed for the process of image processing.Due to the numerous of the shape of plant leaves,it was difficult to make the parameter activity contour model apply to the image segmentation of plant leaves with complex background.Although the shapes can be divided into several classes,and a parametric active contour curve template was proposed for different classes,due to the characteristics of snake model,only the initial contour curve can be located in the target boundary neighborhood to make the curve eventually evolve to the target boundary.Therefore,how to apply the contour model to the plant leaf image segmentation well will to be studied.And it is also the direction of future research to select the appropriate segmentation method for the characteristics of plant leaves in complex background.