Autonomous Parking Area Recognition via Deep Learning and Ultrasonic Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous parking technologies face challenges in recognizing parking areas without objects, such as vehicles or pillars, and struggle to accurately detect parking lines due to factors like light reflection and shadow effects, especially when auxiliary lines or merchandise loading marks are present.
Innovation Solution
A vehicle parking assistance system incorporating an image sensing device, artificial intelligence learning device, and controller that uses deep learning to detect parking lines, entrance points, and generate parking information for autonomous parking, overcoming obstacles like light reflection and shadow effects by analyzing pixel value changes and classifying entrance points with high confidence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ultrasonic waves are used to recognize parking area, then the system can detect objects around the vehicle, but it cannot recognize parking areas where there are no objects around the space
Solution Approach 1:
The patent combines ultrasonic wave detection with image recognition technology to create a hybrid system. The image sensing device captures visual information of the parking area including empty spaces, while ultrasonic sensors detect objects. By merging these two detection methods, the system overcomes the limitation of ultrasonic-only systems that cannot identify parking areas without objects, achieving both accurate object detection and empty space recognition.
2Adaptability or versatility
If image recognition is used to detect parking area, then the system can recognize spaces without objects, but it fails to accurately distinguish parking lines due to light reflection, shadow effects, auxiliary lines, or merchandise loading marks
Solution Approach 1:
The patent introduces an intermediary processing layer that analyzes image data through multiple stages. First, the system detects candidate parking lines in the image, then uses these detected lines as intermediaries to identify parking areas. This intermediary approach allows the system to overcome direct detection failures caused by light reflection, shadows, or misleading marks, by using the detected lines as reference points to confirm actual parking boundaries.
Solution Approach 2:
The image processing is divided into multiple segmentation stages: detecting candidate lines, identifying parking line pairs, determining parking areas, and locating entrance points. This segmentation allows the system to handle complex visual conditions by breaking down the detection task into manageable steps, where each stage refines the output of the previous stage, improving overall accuracy despite challenging lighting and marking conditions.
3Productivity
If conventional image processing is used to detect parking lines, then the system can identify lines in the image, but it cannot reliably distinguish actual parking boundaries from auxiliary lines or merchandise loading marks
Solution Approach 1:
The system performs preliminary detection of candidate parking lines before final identification. It first detects all potential line segments in the image, then uses these as candidates for further verification. This preliminary action allows the system to maintain high processing speed while enabling subsequent stages to filter and verify which candidates represent actual parking boundaries versus auxiliary lines or marks.
Solution Approach 2:
The system employs feedback mechanisms where detected parking lines are used to verify and refine parking area identification. The detected lines provide feedback to confirm whether identified parking areas are valid, and vice versa. This mutual verification process improves boundary identification accuracy by cross-checking results from different detection stages, reducing false positives from auxiliary lines or merchandise marks.
Data Source
AI summary
A vehicle parking assistance device includes an image sensing device, an artificial intelligence learning device, and a controller connected with the image sensing device and the artificial intelligence learning device. The controller is configured to obtain an image using the image sensing device, detect at least one parking line pair in the obtained image, detect a parking slot based on deep learning, detect a parking area based on the detected parking slot and the at least one detected parking pair, detect an entrance point for the parking area, and generate parking information based on the parking area and the entrance point.