Stereo Camera Road Surface Disparity Segmentation for Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems face challenges in achieving high detection performance due to factors such as road shape, structures, and the presence of other objects on the road.
Innovation Solution
A road surface detection device and object detection system that utilize a stereo camera to generate a disparity map, where the processor approximates the relationship between road surface coordinates and disparities using two straight lines, enabling accurate object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object detection is performed using a disparity map from a stereo camera, then distance measurement and object detection capability are improved, but detection accuracy deteriorates due to interference from road surface disparities caused by road shape and structures
Solution Approach 1:
The disparity map is segmented into road surface regions and non-road surface regions by detecting edges and classifying pixels based on their disparity characteristics. This segmentation separates the harmful road surface disparities from the target object disparities, allowing accurate object detection by processing only the non-road surface regions.
Solution Approach 2:
The road surface disparity information is extracted and removed from the disparity map by identifying and masking out regions corresponding to road surfaces based on edge detection and disparity classification. This extraction eliminates the harmful interference while preserving the object information for accurate detection.
2Reliability
If the entire disparity map is processed for object detection, then comprehensive object detection is achieved, but processing time increases due to the large amount of data including road surface information
Solution Approach 1:
The disparity map is divided into road surface and non-road surface regions through edge detection and pixel classification. By segmenting the data, the system processes only the relevant non-road surface regions for object detection, significantly reducing processing time while maintaining detection completeness.
Solution Approach 2:
Instead of processing the entire disparity map, the system performs partial processing by focusing only on non-road surface regions identified through classification. This partial action approach reduces computational load and processing time while maintaining sufficient detection coverage for safety-critical applications.
3Area of stationary object
If disparity information from all regions is used for object detection, then detection coverage is improved, but detection precision deteriorates due to mixed road surface and object disparities
Solution Approach 1:
The system segments the disparity map into road surface and non-road surface regions using edge detection and disparity-based pixel classification. This segmentation enables the system to maintain wide detection coverage by monitoring all regions while achieving high precision by analyzing only the non-road surface regions for object detection.
Solution Approach 2:
Different processing qualities are applied to different regions: all regions are scanned for coverage, but only non-road surface regions undergo detailed object detection processing. This local quality approach ensures both comprehensive coverage and high detection precision by concentrating computational resources where they are most needed.
Data Source
AI summary
An object detection device includes a processor. The processor acquires or generates a first disparity image generated on the basis of an output of a stereo camera mounted in a mobile object. In the first disparity image, pixels representing disparities are arranged on a two-dimensional plane formed by a first direction corresponding to a base-length direction of the stereo camera and a second direction intersecting the first direction. The processor approximates a relationship between a coordinate, in the second direction, of a road surface in a direction of travel of the mobile object and a disparity representing the road surface with two straight lines, the relationship being included in the first disparity image.