Roadside Obstacle Detection Jitter Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Roadside perception systems face inaccuracies due to camera jitter, especially when installed on tall poles, leading to invalid external parameters and inaccurate obstacle detection, as conventional methods like optical and mechanical image stabilization are costly and electronic stabilization has errors.
Innovation Solution
A method that acquires current images from cameras, inputs them into a pre-trained detection model to determine obstacle positions, calculates offsets between current and template images, and converts pixel coordinates to world coordinates, enhancing robustness without additional sensors and reducing computational costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If the camera is installed on a tall pole to increase field-of-view, then the roadside perception ability is improved, but the camera jitters due to wind or heavy vehicles causing inaccurate obstacle detection
Solution Approach 1:
The system performs offline external parameter calibration in advance to establish the initial coordinate transformation relationship between the camera and world coordinate system. This pre-calibrated parameter serves as a baseline that can be quickly adjusted using image registration when jitter occurs, avoiding the need for real-time complex calibration while maintaining detection accuracy.
Solution Approach 2:
The system uses image registration technology to calculate the offset between the current image and template image, then feeds this offset information back to adjust the pre-calibrated external parameters. This feedback mechanism compensates for camera jitter caused by wind or heavy vehicles, maintaining accurate obstacle detection despite camera movement.
2Reliability
If optical image stabilization or mechanical image stabilization is used to solve camera jitter, then the detection accuracy is improved, but the cost increases significantly
Solution Approach 1:
The system replaces mechanical image stabilization methods with a software-based image registration approach. By calculating the offset between current and template images and adjusting external parameters accordingly, the system achieves jitter compensation without requiring additional mechanical stabilization hardware, thereby reducing costs while maintaining detection accuracy.
3Ease of manufacture
If ordinary electronic image stabilization is used to reduce cost, then the cost is reduced, but large errors occur and feature point extraction becomes unrobust
Solution Approach 1:
The system performs comprehensive external parameter calibration offline before deployment, establishing a robust baseline transformation relationship. This pre-calibration ensures that even when using cost-effective electronic image stabilization methods, the system maintains high measurement precision because the stabilization operates on top of an already accurate calibrated framework rather than attempting to correct all errors in real-time.
4Reliability
If online calibration of external parameters is performed to maintain accuracy in real-time, then the detection accuracy is improved, but the computational cost and hardware requirements increase
Solution Approach 1:
The system performs the computationally intensive external parameter calibration process offline in advance, storing the calibrated parameters for quick retrieval and adjustment. During real-time operation, only lightweight image registration and offset calculation are needed to compensate for jitter, dramatically reducing computational cost and energy consumption while maintaining detection accuracy.
Solution Approach 2:
The system introduces image registration technology as an intermediary between the pre-calibrated external parameters and real-time detection. This intermediary efficiently handles the jitter compensation by calculating simple image offsets, avoiding the need for computationally expensive real-time full calibration while maintaining accuracy.
Data Source
AI summary
The present disclosure discloses a method and apparatus for detecting an obstacle, and relates to the technical field of intelligent transportation. A specific implementation plan is: acquiring a current image acquired by a camera; inputting the current image into a pre-trained detection model to obtain a position of a detection frame of an obstacle and determine a first pixel coordinate of a grounding point in the current image; determining an offset between the current image and a template image; converting the first pixel coordinate into a world coordinate of the grounding point based on the offset; and outputting the world coordinate of the grounding point as a position of the obstacle in a world coordinate system. This embodiment solves the problem of camera jitter from an image perspective, greatly improves the robustness of the roadside perception system, and saves computing resources.