Automatic Parking Space Detection With Multi-View Confirmation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic parking systems inaccurately detect parking spaces, leading to potential collisions with obstacles due to misidentification of obstacles as parking spaces.
Innovation Solution
An automatic parking method and apparatus that utilizes a combination of environmental images and panoramic surround-view images to determine parking space information, enhancing detection accuracy by confirming the presence of parking spaces in multiple views and using confidence levels to select the target parking space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single environmental image is used to detect parking spaces, then the detection process is simple and fast, but the accuracy is low and obstacles may be misidentified as parking spaces
Solution Approach 1:
The patent combines multiple image sources (environmental images from different orientations and panoramic surround-view images) into a unified detection framework. By merging these different visual data streams, the system cross-validates parking space detections across multiple views, significantly reducing misidentification of obstacles as parking spaces while maintaining systematic complexity management through integrated processing.
Solution Approach 2:
The patent transitions from single-view detection to multi-dimensional detection by incorporating images from different orientations (front, rear, left, right) and different types (environmental vs. panoramic). This dimensional expansion allows the system to verify parking spaces from multiple perspectives, dramatically improving detection accuracy through spatial cross-validation.
2Measurement precision
If multiple images are used to confirm parking space presence, then detection accuracy is improved, but processing time and computational load increase
Solution Approach 1:
The system performs preliminary detection using environmental images during the cruise phase to identify potential parking spaces before the parking phase begins. This preliminary action allows the system to pre-process and prioritize candidate parking spaces, reducing the computational burden during the actual parking execution while maintaining high detection accuracy through subsequent verification with panoramic images.
Solution Approach 2:
The patent implements a tiered detection approach where environmental images provide initial parking space identification and panoramic surround-view images provide verification. This partial action strategy processes only the necessary subset of images at each stage, avoiding full processing of all available images and thereby reducing overall processing time while maintaining high accuracy through selective verification.
Data Source
AI summary
An automatic parking method includes determining parking space information of a plurality of candidate parking spaces based on a first image and a second image of an ego vehicle, where the parking space information indicates locations of the parking spaces in a stereo garage, the first image is an environmental image in a first orientation of the ego vehicle, and the second image is a panoramic surround-view image of the ego vehicle; determining a target parking space from the plurality of candidate parking spaces based on the parking space information of the plurality of candidate parking spaces; and controlling the ego vehicle to be parked in the target parking space for more accurate recognition.