Sensor Image Classification via Consolidated Probability Values
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing classification algorithms for sensor data, such as images, face challenges in achieving high accuracy due to the unreliability of single probability values, and existing approaches to address this issue are either unsatisfactory in terms of accuracy or overly complex.
Innovation Solution
The method involves determining consolidated probability values by combining base probability values from a specific capture with context probability values from temporally close captures, where the context probability values are normalized to reduce uncertainty and improve classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single probability value is used for classification, then the processing complexity is low, but the reliability of the classification result is insufficient
Solution Approach 1:
The patent combines multiple probability values from different captures and classes into a single consolidated probability value. This merging process integrates information from temporal sequences and multiple class predictions, resolving the contradiction by achieving high reliability through combination while maintaining computational efficiency through a unified consolidation formula.
Solution Approach 2:
The patent transforms multiple probability parameters into a single consolidated probability parameter through a mathematical formula. This parameter change approach converts the multi-dimensional probability space (multiple captures × multiple classes) into a one-dimensional consolidated probability, improving reliability without proportionally increasing complexity.
2Measurement precision
If existing approaches like majority voting or normalisation are applied to improve accuracy, then the classification accuracy improves, but the processing complexity becomes too high
Solution Approach 1:
The patent extracts only the essential information needed for classification from the full probability distribution, consolidating multiple probability values into a single representative value. This extraction approach achieves accurate classification by focusing on the most relevant probability information while discarding redundant details, thereby reducing processing complexity.
Solution Approach 2:
The patent performs preliminary consolidation of probability values from temporal sequences before the final classification decision. By pre-processing the probability values through consolidation in advance, the system prepares optimized input for classification, improving accuracy while reducing the computational burden during real-time decision-making.
Data Source
Figure 1
Figure 2
AI summary
A method for classifying a capture taken by a sensor comprises the following steps: - receiving a sequence of a plurality of captures taken by a sensor, each of the captures comprising a plurality of elements; - generating, per capture, a plurality of raw probability values, each of the raw probability values being linked to a respective one of a plurality of predetermined classes and indicating the probability that the capture or an element of the capture is associated with the respective class; - determining, for a respective one of the captures, a plurality of consolidated probability values in dependence of a plurality of base probability values and a plurality of context probability values, wherein the base probability values represent the raw probability values of the respective capture and the context probability values represent the raw probability values of at least one further capture of the sequence other than the respective capture, the context probability values being normalised according to a normalisation rule; and - classifying the respective capture or an element thereof on the basis of the consolidated probability values.