Multi-kernel SVM based multi-feature fusion algorithm for analyzing wastewater membrane permeation in environmental pollution

Wastewater membrane permeation analysis in environmental pollution is of great significance because water pollution information contains rich semantics. Traditional algorithms based on one or several features of environmental pollution related factors are common ways of wastewater membrane permeation but with unsatisfied performance and disadvantages in complex environmental pollution. Algorithms based on deep machine learning seems better in analyzing wastewater membrane permeation with the cost of a great deal of calculation and training samples but shows unsatisfactory and disadvantaged at the case of small sample field. In this paper, a novel algorithm using multiple feature fusion of geometric features and high-level ones of environmental pollution factors based on support vector machine ( SVM) with multi-kernel functions is proposed. By using multiple feature fusion, pollution factors can be detected more effectively in complex scenes. Multi-kernel functions may avoid the limitations of single kernel function, which can improve the performance of the algorithm. Known as a statistical learning theory, SVM is suitable for small sample space classification. Experimental results showed that, the proposed algorithm performs well in small sample field and also has good generalization ability.