Stereo Image Object Detection Using Cost Volume Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional stereo-based object detection methods face challenges with inaccurate depth determination using local methods and high computational requirements of global methods, making them unsuitable for real-time applications, and region of interest generation techniques often result in missed detections.
Innovation Solution
The method computes a cost volume from stereo images and applies object detection directly using features obtained from this volume, leveraging T decision tree classifiers learned from training features to avoid the complex depth map estimation step and improve detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If local methods are used for depth determination, then computational resources are reduced, but detection accuracy deteriorates due to inaccurate depth determination
Solution Approach 1:
The patent extracts depth information directly from the cost volume without performing complete depth map estimation. By taking out only the necessary depth cues from the cost volume data structure, the system avoids computationally intensive global optimization while maintaining sufficient accuracy for object detection.
Solution Approach 2:
The cost volume serves as an intermediary data structure that contains depth information in a compressed form. Instead of computing a full depth map through complex optimization, the system uses the cost volume as a mediator to extract depth cues directly, reducing computational complexity while preserving measurement accuracy.
2Measurement precision
If global methods are used for depth map estimation, then depth determination accuracy is improved, but computational resources and processing time increase significantly
Solution Approach 1:
The patent extracts only the necessary depth information from the cost volume without performing complete depth map estimation. By taking out only the necessary depth cues from the cost volume data structure, the system avoids computationally intensive global optimization while maintaining sufficient accuracy for object detection.
Solution Approach 2:
Instead of performing complete global depth map estimation, the system applies partial action by extracting depth information only from the cost volume where it is most relevant for object detection. This partial extraction approach achieves sufficient accuracy without the full computational burden of global methods.
3Device complexity
If region of interest generation techniques are used, then computational complexity is reduced, but detection reliability deteriorates due to certain miss detection when area of interest does not cover the object
Solution Approach 1:
The cost volume-based object detection system is universally applicable to all regions of the image without requiring predefined regions of interest. The system can detect objects anywhere in the image by searching through the entire cost volume, making it multi-functional and reliable across different scenarios without the limitations of region-based approaches.
Data Source
AI summary
A method detects an object in a pair of stereo images acquired of a scene, by first generating a cost volume from the pair of stereo images, wherein the cost volume includes matching costs for a range of disparity values, for each pixel the stereo images, between the stereo images in the pair. Feature vectors are determined from sub-images in the cost volume using a feature function of the disparity values with a minimal accumulated cost within regions inside the sub-images. Then, a classifier is appled to the feature vectors to detect whether the sub-image includes the object.