Remote Sensing Classification Using KL Divergence Time Series
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing remote sensing classification methods face challenges in accurately identifying multiple ground object types due to reliance on normal distribution assumptions, morphological traits, and sensitivity to weather conditions, leading to low spatial resolution and classification accuracy.
Innovation Solution
A remote sensing classification method utilizing relative entropy (Kullback-Leibler divergence) integrates temporal information with KL divergence to classify ground objects without mandatory distribution assumptions, using time series vegetation index data and calculating KL values to determine the minimum distance between pixel distributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional parameterized classification methods are used, then the classification process is simplified, but classification accuracy deteriorates due to unrealistic normal distribution assumptions for high-dimensional data
Solution Approach 1:
The patent changes the fundamental parameter assumption from normal distribution to arbitrary distribution, enabling the use of KL divergence instead of traditional statistical methods. This parameter change allows the system to handle high-dimensional remote sensing data without making unrealistic distributional assumptions, thereby improving classification accuracy while maintaining computational feasibility
Solution Approach 2:
The patent substitutes traditional mechanical classification approaches (parameterized methods relying on normal distribution) with an information-theoretic approach using KL divergence. This substitution replaces the need for distributional assumptions with a more flexible mathematical framework that better captures the complexity of high-dimensional remote sensing data
2Productivity
If morphological trait-based similarity matching methods are used, then classification can be performed, but the process becomes relatively complex and less efficient
Solution Approach 1:
The patent replaces complex morphological trait analysis with KL divergence calculation, substituting a detailed shape-based comparison approach with a more efficient information-theoretic distance measure. This substitution maintains classification capability while reducing computational complexity and improving processing efficiency
3Measurement precision
If KL divergence is applied to single-crop identification, then classification accuracy is improved, but the method cannot identify multiple ground object types simultaneously
Solution Approach 1:
The patent extends the KL divergence method from single-crop identification to multi-type ground object classification by formulating a unified classification framework. The same KL divergence metric is applied universally across multiple reference distributions (one for each ground object type), enabling the system to identify multiple types simultaneously while maintaining the accuracy benefits of the information-theoretic approach
Solution Approach 2:
The patent segments the classification problem into multiple reference distributions, each representing a different ground object type. By calculating KL divergence from the input data to each segmented reference distribution, the system can identify multiple object types through comparison, thereby achieving both high accuracy and multi-type capability
Data Source
AI summary
A remote sensing classification method based on relative entropy includes: determining sample points of n types of ground objects in a study area and determining series remote sensing parameters; extracting, based on the sample points, remote sensing parameter values to form standard time series plots as a first distribution; taking remote sensing parameter values of to-be-classified pixels as a second distribution, determining, based on the second distribution and the first distribution, KL values of the to-be-classified pixels by using a KL-divergence formula, then obtaining n KL layers; and comparing n KL values of each to-be-classified pixel to classify it to be a type of ground objects with a minimum KL value. The method utilizes variation characteristics of ground objects in series and tightly combines with KL-divergence possessing obvious advantages in measuring probability distribution similarity, thereby achieving better classification and recognition on ground object types and improving classification accuracy.