Soil porosity analysis using combined maximum entropy and class variance thresholding

Abstract

Soil is a type of a compound mixture. Computer tomography (CT) images have a greater potential in providing finer and thinner soil porous space of different medium. CT samples preserve all the information of solid and void spaces of soil in a nondestructive manner. Decision handling by a human is a negotiable duty in an automated soil management system. Image thresholding generally segments void and solid medium in CT image samples. In earlier versions, the most referred automated global thresholding is Otsu’s class-based and Kapur’s maximum entropy-based thresholding. In Otsu’s class-based thresholding, maximized interclass variances of objects between foreground and background of the object were used whereas in Kapur’s maximum entropy, it maximizes the entropy of self-dissimilar junction between foreground and background. In both these threshold-based segmentations, specific delimitations such as misinformation about pore and solid spaces in the object prevail. The above two historical methods are most excellent in finding macropores, but moreover, real-time sample images contain a number of micropores. To reimburse the drawback in identification of micropores and low resolution of gray scale images, a segmentation model that integrates both class variance and entropy of the signals was needed. The proposed combined maximum entropy-class variance thresholding (CME-CV) will be most useful in attaining accurate identification of pore structure. Finally, it was compared with the typical methods to validate the impact on porosity, void ratio, relative porosity, and misclassification error. Comparative analysis reveals that crossbreed optimal thresholding has been effective in detecting micropores.