Parking Space Detection Using Corner Verification and Frame Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional parking space detection methods in intelligent driving systems suffer from insufficient accuracy and high costs, failing to effectively identify and track parking spaces using sensors like cameras and lidar.
Innovation Solution
A method and apparatus for detecting parking spaces using consecutive image frames, which involves recognizing parking space corners, verifying valid spaces, tracking them using a list, and outputting semantic information only when spaces are consistently visible, thereby improving accuracy and resource efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods detect only general area of parking space, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The method segments the parking space detection process into distinct stages: initial general area detection, corner point identification, and verification phase. This segmentation allows the system to first identify the general parking space area using simpler methods, then focus computational resources on detecting corner points and verifying specific spaces, thereby improving precision without proportionally increasing overall system complexity
Solution Approach 2:
The method performs preliminary detection of the general parking space area before conducting detailed corner point detection and verification. By first identifying the rough boundaries and then refining the detection through corner point analysis, the system prepares data structures and filters candidates in advance, reducing the computational burden during the precision verification phase
2Reliability
If multiple consecutive frames are processed for verification, then reliability is improved, but use of energy increases
Solution Approach 1:
The method applies partial verification by processing only the necessary number of consecutive frames required to establish reliability, rather than continuously processing all available frames. The verification mechanism uses a threshold-based approach where detection is confirmed after a predetermined number of consistent detections across consecutive frames, avoiding excessive computational energy consumption while maintaining reliable detection
Solution Approach 2:
The system implements feedback mechanisms where detection results from previous frames inform processing in subsequent frames. Once a parking space is reliably detected through verification across multiple frames, the system adjusts its processing behavior, reducing redundant verification computations in later frames while maintaining detection reliability through continuous monitoring
3Loss of information
If tracking list maintains all verified parking spaces, then loss of information is reduced, but device complexity increases
Solution Approach 1:
The method extracts and maintains only the essential tracking information for verified parking spaces in a structured list, separating critical data (position, verification status, frame count) from unnecessary details. This extraction approach ensures that all necessary information for tracking and verification is preserved while keeping the data structure compact and manageable, reducing the complexity of tracking management
4Measurement precision
If verification based on parking space corners is performed, then measurement precision is improved, but difficulty of detecting and measuring increases
Solution Approach 1:
The method transitions from two-dimensional area-based detection to three-dimensional corner point detection by incorporating depth information and spatial coordinates. By detecting corner points as distinct features with precise x, y, and z coordinates rather than merely identifying bounding boxes, the system achieves higher boundary precision while using dimensionality to simplify the verification process through spatial relationship analysis
Data Source
AI summary
A parking space detection method, apparatus, device, and storage medium are provided. The method includes: obtaining image frames of a region where a current vehicle is located; for each image frame, recognizing one or more parking spaces and parking space corners of the recognized parking space; determining, based on the parking space corners, a verified parking space; tracking the verified parking space to record in a parking space tracking list a quantity of consecutive visible frames where the verified parking space is recognized, and a quantity of consecutive missing frames where the verified parking space is not recognized, and delete the verified parking space if the quantity of the consecutive missing frames reaches a first threshold; and determining and outputting, based on the parking space corners of the verified parking space, semantic information of the verified parking space if the quantity of consecutive visible frames reaches a second threshold.