Multi-Camera Object Detection With Overlap Depth Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems in autonomous vehicles face inaccuracies due to deviations in depth values and feature vectors between images from different cameras, particularly in overlapping regions, leading to degraded detection performance.
Innovation Solution
An apparatus and method that utilize a stereo matching device to correct depth value deviations and a feature vector determination device to reduce discrepancies between overlapping and non-overlapping regions, employing adversarial learning to enhance object detection accuracy and speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo cameras are used to obtain depth information, then object detection capability is improved, but depth value deviation between overlapping regions of adjacent images occurs
Solution Approach 1:
The patent introduces a depth correction module that acts as an intermediary between stereo camera depth extraction and object detection. This module uses feature matching between overlapping regions of adjacent images to calculate correction values, which are then applied to correct depth values. The intermediary processing ensures depth consistency across multiple camera views while maintaining the benefits of stereo vision for object detection.
Solution Approach 2:
The patent implements a feedback mechanism where depth correction values are calculated based on the comparison of overlapping region features from adjacent images. These correction values are fed back to adjust the depth information, creating a closed-loop system that continuously improves depth accuracy. The feedback process ensures that depth deviations are systematically corrected based on actual observed discrepancies.
2Area of stationary object
If feature vectors from overlapping regions are used for object detection, then detection coverage is improved, but feature vector deviation occurs leading to reduced detection accuracy
Solution Approach 1:
The patent introduces a feature correction module as an intermediary that processes feature vectors from overlapping regions. This module calculates correction values based on comparisons between adjacent image features and applies these corrections to the feature vectors before they are used for object detection. The intermediary ensures feature consistency across overlapping regions while maintaining comprehensive detection coverage.
Solution Approach 2:
The patent modifies feature vector parameters by applying correction values derived from overlapping region analysis. The correction process adjusts key parameters of the feature vectors to reduce deviations caused by pixel data variations in adjacent images. This parameter adjustment maintains the beneficial coverage of overlapping regions while improving feature consistency for accurate detection.
3Area of stationary object
If multiple cameras with overlapping fields of view are used, then detection coverage is improved, but system complexity increases
Solution Approach 1:
The patent extracts and processes only the overlapping regions from multiple camera images, rather than processing entire images from all cameras. By isolating and correcting only the overlapping portions, the system reduces computational complexity while maintaining the detection coverage benefits of multiple cameras. The extraction approach focuses resources on the critical overlapping areas where corrections are most needed.
Data Source
AI summary
An apparatus and method for detecting an object and a computer readable recording medium therefor are disclosed. The apparatus includes: a first camera that obtains a first image of a first view angle range; a second camera that obtains a second image of a second view angle range having an overlapping region with the first view angle range; an object detection network that detects an object in the first and second images; a stereo matching device that corrects a deviation of a depth value between the first and second images in the overlapping region based on a feature value output by a backbone network of the object detection network; and a feature vector determination device that performs adversarial learning to reduce a deviation of a feature vector between objects in the overlapping region and a non-overlapping region while the object detection network detects the object.