Parking Space Recognition Using Multi-View Reference Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing parking assistance systems face accuracy issues in recognizing target parking spaces due to images that are not suitable for recognition processing, such as those lacking distinct features.
Innovation Solution
The system utilizes multiple reference images captured from different viewpoints and angles to verify the target parking space, enhancing accuracy by using a combination of imaging devices and image processing to generate overhead views and perform pattern matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single overhead image is generated and used for parking space recognition, then the device complexity is reduced, but the accuracy in recognizing target parking space deteriorates when the image lacks suitable features
Solution Approach 1:
The patent segments the parking space recognition task by dividing it into multiple verification stages using different reference images. Instead of relying on a single overhead image, the system divides the verification process into multiple independent checks using reference images from different viewpoints (front, rear, side images), thereby improving recognition accuracy without proportionally increasing overall system complexity.
Solution Approach 2:
The patent transitions from a two-dimensional overhead image to a multi-dimensional verification approach by incorporating reference images from multiple viewpoints and directions. This adds temporal and spatial dimensions to the verification process, allowing the system to verify parking spaces from front, rear, and side perspectives, thereby compensating for the limitations of any single image.
2Measurement precision
If multiple different reference images are used to execute verification process, then the accuracy in recognizing target parking space is improved, but the device complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-registering multiple reference images (front images, rear images, side images) in the memory before the actual parking verification is needed. This allows the system to have verification images ready in advance from different viewpoints, so when verification is required, the system can quickly retrieve and compare these pre-prepared images without real-time processing delays or excessive computational complexity.
Solution Approach 2:
The patent changes the parameter of image viewpoint by using reference images captured from different positions (front, rear, side) and directions. This parameter diversification allows the verification process to be more robust against poor viewing angles, as the system can switch between different reference images depending on which provides the best verification match for the current parking situation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A plurality of different reference images (IM31, IM41, IM42) for specifying a target parking space (Pt) is preliminarily stored. The different reference images are generated by using an imaging device (20) that captures an image around a subject vehicle (V1). After that, when the subject vehicle (V1) is parked in the target parking space (Pt), the current image captured around the subject vehicle is verified with at least one reference image of the plurality of different reference images to recognize the target parking space, and the travel operation of the subject vehicle to the target parking space is autonomously controlled based on positional information of the recognized target parking space.