Image Region Segmentation for Low-Visibility Object Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distant objects and poor line-of-sight conditions pose challenges for reliable object detection, which is crucial for ensuring safety in driving scenarios such as expressway driving or adverse weather conditions.
Innovation Solution
An information processing device that includes an object detection unit and an object determination unit, which segments image data into regions representing different classes of objects and determines confidence levels to accurately classify objects, even in overlapping regions, using machine learning models to enhance detection reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object detection methods are used, then processing speed is maintained at acceptable levels, but detection accuracy deteriorates for distant objects and objects in poor line-of-sight conditions
Solution Approach 1:
The patent segments the image into multiple regions (first region for object recognition, second regions for region recognition) and processes each region separately using different recognition models. This segmentation allows the system to handle distant objects and poor line-of-sight conditions by dividing the complex detection task into manageable parts, each optimized for specific detection challenges.
Solution Approach 2:
The patent introduces a confidence level dimension to the detection process. By determining confidence levels for object recognition results and using these to select between different recognition models, the system adds a decision-making dimension that improves detection accuracy for challenging cases without fundamentally changing the core detection architecture.
2Measurement precision
If region recognition segments image into multiple regions, then object classification accuracy improves, but processing time increases
Solution Approach 1:
The patent applies partial region recognition only to areas where object recognition confidence levels are insufficient, rather than applying full region recognition to the entire image. This selective approach maintains high classification accuracy for challenging objects while reducing overall processing time by avoiding unnecessary computation in high-confidence regions.
Solution Approach 2:
The system uses confidence levels from object recognition as feedback to determine whether region recognition is needed for specific regions. This feedback mechanism allows the system to dynamically adjust processing depth based on initial results, optimizing the balance between accuracy and processing time.
3Reliability
If confidence level threshold is set high, then object detection reliability improves, but detection coverage decreases for uncertain regions
Solution Approach 1:
The patent dynamically adjusts the detection strategy based on confidence levels. Instead of using a fixed threshold, the system adapts its behavior by selecting different recognition models (first or second recognition model) based on the confidence level of object recognition results, allowing it to maintain both reliability and coverage across different detection scenarios.
Solution Approach 2:
The system changes the parameter of confidence level thresholding dynamically. By using confidence levels to determine which recognition model to apply, the system effectively adjusts detection sensitivity and specificity in real-time, maintaining high reliability for clear detections while preserving coverage for uncertain regions through model selection.
Data Source
AI summary
An object recognition unit detects, from image data that represents an image using respective signal values of a plurality of pixels, a first region that is a region representing the object in the image. The object recognition unit determines a first confidence level that is a class confidence level for a first class, the first class being a class of the object represented in the first region. A region recognition unit segments the image of the image data into second regions representing different classes of an object, and determines, for each of second regions, a second class being a class of an object in the second region. An object determination unit determines, as the second class, a class of an object in a non-overlapping region that is the second regions that do not overlap with the first region.