Vehicle Door Camera Sensing for 3D Collision Warning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies for detecting obstacles around vehicle doors using cameras are either costly or inaccurate, failing to precisely estimate depth and predict collision risks, leading to potential accidents.
Innovation Solution
A vehicle control apparatus utilizing a camera, processor, and storage to transform depth information into 3D point information, set collision sensing zones, and warn of potential collisions by analyzing the distribution of 3D points and variations in their numbers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If radar sensor or LiDAR sensor is employed to detect obstacle, then measurement precision of depth information is improved, but device cost increases
Solution Approach 1:
The patent uses a camera to capture images and generates depth information through image processing algorithms, creating a virtual copy of the depth sensing capability without using expensive physical depth sensors like LiDAR or radar. This allows the system to achieve depth estimation functionality at a lower cost while maintaining sufficient measurement precision for collision risk assessment.
2Ease of manufacture
If SFM algorithm is employed to estimate depth, then device cost is reduced, but measurement precision of depth information deteriorates
Solution Approach 1:
The patent merges multiple algorithms including SFM (Structure from Motion) with additional image processing techniques and machine learning methods to compensate for the limitations of individual algorithms. By combining these approaches, the system achieves both cost-effectiveness (using only camera) and improved depth estimation accuracy beyond what SFM alone can provide.
Solution Approach 2:
The patent uses a composite approach by integrating multiple algorithmic components (SFM, depth estimation algorithms, collision risk assessment algorithms) to create a hybrid system that leverages the strengths of each component while mitigating their individual weaknesses, achieving both low cost and high precision.
3Ease of operation
If object detecting algorithm based on deep-learning is employed, then ease of operation is improved, but measurement precision of distance recognition deteriorates
Solution Approach 1:
The patent incorporates feedback mechanisms where the system continuously monitors detection results and adjusts its processing parameters. The collision risk assessment algorithm receives feedback from both the object detection algorithm and SFM algorithm, allowing it to compensate for distance recognition inaccuracies and provide more reliable collision risk warnings.
4Ease of manufacture
If camera is used instead of radar or LiDAR, then device cost is reduced, but measurement precision of depth information deteriorates
Solution Approach 1:
The patent changes the processing parameters by applying multiple image processing algorithms and adjusting processing conditions to extract maximum depth information from the camera data. By transforming the same input data through different algorithmic lenses and combining results, the system achieves higher precision depth estimation from a low-cost camera than would be expected from a single algorithm approach.
Data Source
AI summary
An apparatus of controlling a vehicle and a method for controlling the same includes a camera, a processor operatively connected to the camera, and a storage operatively connected to the processor to store instructions executed by the processor. The processor obtains an image of a surrounding of a door of the vehicle, which is captured through the camera, obtains depth information from the obtained image, transforms the depth information into three dimensional (3D) point information, determines collision possibility of the door with an obstacle, based on distribution of the 3D point information to determine collision possibility, and warns the collision of the door when the collision possibility is present, when executing the instructions.