Vehicle Video Driving-Area Detection Using Bounding-Box Calibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing dashcams and integrated vehicle cameras face challenges in accurately estimating the driving cone due to intrinsic and extrinsic variations in camera installation and vehicle characteristics, leading to errors in advanced driving assistance systems and increased resource consumption.
Innovation Solution
A video system uses object recognition and classification techniques to detect vehicles within a video frame, determining bounding boxes and applying a linear fit model to estimate the driving cone, which can also be adjusted based on vehicle class and road conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional camera installation methods are used without correction, then device complexity is reduced, but measurement precision of the driving cone deteriorates
Solution Approach 1:
The system performs preliminary calibration by detecting vehicles in multiple video frames before actual driving assistance operations. Bounding boxes of detected vehicles are used to pre-calculate correction values for the driving cone estimation, so that when the system operates normally, accurate measurements can be obtained without real-time complex calculations.
Solution Approach 2:
The patent introduces an intermediary calibration process that uses detected vehicles as reference objects. These vehicles serve as mediators to establish the relationship between camera installation variations and actual driving cone geometry, enabling correction of measurement errors without directly measuring the driving cone itself.
2Measurement precision
If multiple video frames are processed for calibration, then measurement precision improves, but use of energy increases
Solution Approach 1:
The system processes multiple video frames during calibration to ensure sufficient data for accurate correction value calculation, but only performs this intensive processing once or periodically rather than continuously. During normal operation, the pre-calculated correction values are applied without requiring continuous heavy processing of multiple frames.
Solution Approach 2:
The calibration process using multiple video frames is performed in advance before actual driving assistance operations. This preliminary computation establishes correction values that can be applied efficiently during normal operation, avoiding repeated energy-intensive processing.
3Reliability
If camera installation variations are not corrected, then device complexity remains low, but reliability of ADAS deteriorates
Solution Approach 1:
The system performs self-calibration by automatically detecting vehicles in the environment and using them as reference objects to calculate correction values for its own camera installation variations. This self-service approach eliminates the need for manual calibration procedures or external calibration equipment, improving reliability without requiring complex external calibration systems.
Solution Approach 2:
Detected vehicles serve as intermediary reference objects that enable the system to self-calibrate. These vehicles mediate between the camera installation variations and the driving cone estimation, providing a practical reference framework for automatic correction.
4Manufacturing precision
If manual calibration procedures are used, then manufacturing precision can be achieved, but ease of operation deteriorates
Solution Approach 1:
The system automatically performs calibration by detecting vehicles in video frames and calculating correction values without human intervention. This eliminates complex manual calibration procedures while achieving high precision through algorithmic processing of environmental reference objects.
Solution Approach 2:
The patent replaces manual mechanical calibration procedures with an automated computer vision-based system. Instead of physically adjusting camera parameters or using manual measurement tools, the system uses algorithmic detection of vehicles and automated calculation of correction values to achieve precise calibration.
Data Source
AI summary
In some implementations, a video system may capture, from a camera mounted to a vehicle, a video of a portion of a road on which the vehicle is traveling. The video system may detect, in the video, a driving lane associated with the road on which the vehicle is traveling. The video system may detect, in the video, multiple other vehicles within the driving lane. The video system may determine, for each of the multiple other vehicles within the driving lane, a bounding box that substantially surrounds an image of the other vehicle, resulting in a plurality of bounding boxes. The video system may determine a region in the video corresponding to an area to be driven by the vehicle based on the plurality of bounding boxes.