Nonparallel SVM Construction for Hyperspectral Image Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional support vector machines (SVM) and twin support vector machines (TWSVM) face challenges in hyperspectral image classification due to the assumption of parallel class boundaries, which affects generalization performance and classification accuracy, particularly in remote sensing applications.
Innovation Solution
A nonparallel support vector machine model is constructed by adding a least square term for empirical risk minimization, involving preprocessing, solving Lagrange multipliers for positive and negative-class hyperplanes, and determining class membership based on Euclidean distances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a traditional support vector machine (SVM) with parallel class boundaries is used, then the model has solid theoretical basis and desirable generalization, but the assumption of parallel boundaries is difficult to establish for actual remote sensing data, affecting model generalization ability
Solution Approach 1:
The patent transitions from the symmetric parallel boundary assumption in traditional SVM to asymmetric nonparallel boundaries in TWSVM, allowing each class to have its own hyperplane that is not constrained to be parallel. This asymmetry enables the model to better adapt to the actual distribution characteristics of remote sensing data while maintaining theoretical rigor through dual optimization problems.
Solution Approach 2:
The patent introduces dynamic adjustment capability by allowing the boundary structure to adapt based on data characteristics. The dual hyperplane system in TWSVM can dynamically adjust their orientations and positions independently, unlike the fixed parallel relationship in traditional SVM, enabling the model to flexibly respond to varying data distributions in remote sensing applications.
2Adaptability or versatility
If a twin support vector machine (TWSVM) with nonparallel hyperplanes is used, then the model can handle nonparallel class boundaries, but it only considers empirical risk without structural risk, affecting generalization performance
Solution Approach 1:
The patent modifies the optimization parameters and objective functions of TWSVM by incorporating regularization terms that account for structural risk. This involves changing the parameter space to include constraints that penalize model complexity, thereby balancing the empirical risk minimization with structural risk control to improve generalization performance while maintaining the nonparallel boundary advantage.
3Adaptability or versatility
If TWSVM is used for hyperspectral image classification, then nonparallel boundaries can be achieved, but the classification effect is not as good as traditional SVM in many cases due to lack of structural risk consideration
Solution Approach 1:
The patent creates a composite optimization framework that combines elements of both traditional SVM (structural risk minimization) and TWSVM (nonparallel dual hyperplanes). This composite approach integrates the regularization principles of SVM with the flexible boundary structure of TWSVM, resulting in a hybrid model that achieves both good generalization and high classification accuracy for hyperspectral images.
Data Source
AI summary
A method for constructing a support vector machine of a nonparallel structure is provided. On the basis of a traditional parallel support vector machine (SVM), a least square term of samples is added, an additional empirical risk minimization term and an offset constraint term are added to an original optimization problem, so as to obtain two nonparallel hyperplanes respectively, and a new nonparallel support vector machine with additional empirical risk minimization is formed. The method includes: preprocessing data, solving a Lagrange multiplier of a positive-class hyperplane, solving a Lagrange multiplier of a negative-class hyperplane, solving parameters of positive-class and negative-class hyperplanes, and determining a class of a new data point. Through the new method, a new nonparallel vector machine algorithm is proposed to further improve classification accuracy of hyperspectral images on the basis of the algorithm itself, so as to obtain better classification performance.