Object Detection Using Map Coordinates to Reduce Processing Load
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection technologies, such as those using stereo cameras, require high processing loads and calibration, making them inefficient for real-time object detection and distance calculation.
Innovation Solution
An object detection apparatus that uses a combination of an imager, an interested-object detector, a position detector, a map storage, and a calculator to associate image coordinates with three-dimensional coordinate points, eliminating the need for stereo camera-based parallax calculations and reducing processing load by utilizing map information and machine learning for object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo camera-based parallax calculation is used for object detection, then object position and distance can be detected, but processing load increases and calibration is required
Solution Approach 1:
The patent extracts and removes the complex parallax calculation process from the object detection system. Instead of using stereo camera parallax methods, the invention uses a single camera combined with map information to directly determine object positions, eliminating the need for complex image processing and calibration procedures while maintaining detection accuracy
Solution Approach 2:
The patent introduces map information as an intermediary element between the camera and the object detection result. By using pre-stored map data containing coordinate information of road features, the system can convert simple image coordinates into accurate real-world positions without requiring complex real-time calculations or multi-camera calibration
2Measurement precision
If stereo camera calibration is performed to ensure accurate parallax calculation, then measurement precision improves, but device complexity and preparation time increase
Solution Approach 1:
The patent performs the complex calibration and coordinate system setup work in advance by pre-storing map information with accurate coordinate data. This preliminary action eliminates the need for time-consuming calibration procedures before actual object detection, as the coordinate transformation relationships are already established in the map database
Solution Approach 2:
The patent uses a simplified copy of the real-world environment stored in map information rather than relying on complex multi-camera geometric relationships. The map data provides a pre-established coordinate framework that can be directly used for position calculation, avoiding the need for precise physical camera calibration
3Productivity
If map information is used to calculate object position, then processing load is reduced, but the system requires access to pre-stored map data
Solution Approach 1:
The patent makes the map storage component serve multiple functions: it stores not only coordinate information for position calculation but also provides reference data for distance estimation and object location verification. This multi-functional use of the map database reduces the need for separate calibration data structures and processing algorithms
Data Source
AI summary
An object detection apparatus is provided with: an imager configured to image surroundings of a subject vehicle and to obtain a surrounding image; an object detector configured to detect an interested-object from the surrounding image and to output first image coordinates, which indicate a position of the detected interested-object on the surrounding image; a calculator configured to associate the interested-object with one or more coordinate points out of a plurality of coordinate points, each of which indicates three-dimensional coordinates of respective one of a plurality of points on a road, on the basis of the first image coordinates and a position of the subject vehicle, and configured to calculate at least one of a position of the interested-object on a real space and a distance to the interested-object from the subject vehicle on the basis of the position of the subject vehicle and the one or more coordinate points associated.