Segmentation And Extraction Of Soil Part In Purple Soil Color Images

There are usually complex backgrounds on a purple soil image collected in the field,and they are normally crops,lichens,weeds,etc.The complex background maybe interferes with the recognition of soil local type based on machine vision.So it is necessary to extract the purple soil region image from the purple soil color image,which is the basic work to further identify soil local type and analyze the characteristics of purple soil.In this paper,segmentation of the purple soil color image acquired by machine vision is studied.The main tasks of this paper are as follows:(1)In order to realize the completed extraction of purple soil region,a segmentation algorithm of purple soil color image based on improved SLIC was proposed.Minkowski distance is first draw into redefining the distance measurement and SLIC algorithm is improved by it.The purple soil color image is initially segmented into many super-pixels with the improved SLIC algorithm.Then,a new measure,based on the ‘a’ component of color space,is reconstructed to stretch the difference between purple soil and background.And the measure is used in the similarity definition of super-pixels.An optimization model is established to optimize the similarity threshold of merging adaptively super-pixels according to the Maximum Inter-class Variance Principle.Starting from the super-pixel in the center of the initial segmentation image,the neighbors of the center super-pixel,which the similarity between it and the center super-pixel is less than the threshold,are merged into the center super-pixel until the center super-pixel can’t be reconstructed.The purple soil area is segmented and the reconstructed center super-pixel is it.Finally,the holes inside the purple soil area are filled with the filling algorithm in this paper,and the purple soil region image is obtained.The simulation experiments show that Jaccard index of our algorithms is higher compared with the threshold segmentation algorithm and the clustering segmentation algorithm and the existing improved SLIC algorithm that are cited in this paper.However,it needs to initialize the number of superpixels and color normalization parameters artificially,so the adaptive segmentation cannot be realized.(2)An improved FCM algorithm is proposed to provide a better quality for adaptive segmentation of the purple soil color image.Firstly,creating the maximum difference optimization model with the Gaussian distance mean between all pixels in the image and each peak of histogram,and solving this optimal model can obtain the optimal clustering number and clustering centers.The dispersion of each cluster is then defined to act as the weight of the distance between each pixel and each clustering center,which realizes the creation of the optimal model of purple soil segmentation based on FCM algorithm.Aiming at the problem of removing scattered impurities in the background and filling hollows in the purple soil region,the algorithm of extracting the boundary of the purple soil region and the algorithm of filling the purple soil region are proposed.Experimental results show that the average segmentation accuracy of our algorithms is higher than the contrast algorithms,and executing time of our algorithms is fast.When there is a circular connected domain with background in the purple soil,it will be regarded as a soil cavity to be filled by the post-processing algorithm in this paper,and we will further study it.