Research On Root Image Analysis Method In Multi-scenario

The root is one of the three major organs of the plant,the root not only firmly fixes the plant on the ground,but also plays a key role in the healthy growth of the whole plant.The morphological parameters of roots are important basis for botanists to study plant health and the role of root in the ecological environment.Therefore,with the continuous development of science and technology in recent years,the research of root system has become more and more in-depth.This thesis takes the root image with different characteristics obtained in three different scenarios as the research object.The target root system is extracted from different backgrounds by image processing,and the root skeleton is extracted by using the thinning algorithm.The morphological calculation of the skeleton is used to obtain the relevant parameters of the root system.The main work of the thesis is as follows:1.The root growth environment is special,the background image of the obtained root image is complex,and the shape of the root system itself is relatively slender.In this thesis,the specific performance of various filters in the root image is studied through experiments,and the appropriate filter is selected as the preprocessing method.2.For the specific root segmentation,this thesis attempts to segment the root image that obtained by different scenes by the three methods of threshold segmentation,edge detection and region growing.The experimental results are analyzed to obtain the specific performance of the traditional algorithm which applied to the root image.3.Based on the comparison of traditional segmentation algorithms,this thesis introduces a pulse coupled neural network with good performance in the field of image processing.A method suitable for root image is proposed by combining pulse coupled neural network and Sobel operator.Finally,the method is compared with the traditional segmentation method from three aspects: manual interaction,time consumption and mis-segmentation rate.4.Finally,in order to obtain the root morphology related parameters,this thesis improves the Zhang-Suen refinement algorithm and introduces the chain domain to optimize the root tip counting algorithm.Then we implements the root image analysis system on the Matlab2018 b platform based on the root image processing flow and the related algorithms involved in this thesis.