Vehicle Obstacle Recognition Using Dual-Dimension Ground Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing obstacle recognition systems for autonomous vehicles face challenges in accurately detecting curbs and obstacles due to limited sensor information, noise, and misclassification of low-height features as ground, leading to difficulties in recognizing distant or low-height obstacles.
Innovation Solution
A device and method utilizing a distance sensor, signal processing units for vertical and horizontal ground filtering, and a signal mixer to extract and mix sensing information, thereby improving obstacle recognition by distinguishing between ground and obstacle signals, reducing noise and misclassification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If only sensor information from side surfaces of curb is used for detection, then the detection method is simple, but the detection performance is poor and difficult to recognize distant curbs
Solution Approach 1:
The patent transitions from using only side surface information (one-dimensional approach) to incorporating both side surface information and top surface information (two-dimensional approach). This dimensional expansion allows the system to detect curbs at various distances by utilizing multiple sensing directions, thereby improving detection accuracy without significantly increasing system complexity.
Solution Approach 2:
The patent divides the curb detection task into two separate detection processes: one for side surfaces and one for top surfaces. Each process uses dedicated sensor information processed through specific filtering methods. This segmentation allows each detection path to be optimized independently, improving overall detection performance while maintaining manageable complexity.
2Reliability
If all distance information estimated as ground is removed from signal information, then ground filtering is effective, but low-height curbs and distant obstacles are misclassified as ground and information is lost
Solution Approach 1:
The patent implements separate filtering processes for vertical and horizontal directions. The vertical ground filtering processes distance information in the vertical direction, while the horizontal ground filtering processes information in the horizontal direction. This segmentation prevents the misclassification of low-height obstacles as ground because each filtering process operates independently with direction-specific criteria, preserving obstacle information that would otherwise be lost.
Solution Approach 2:
The patent applies different filtering characteristics to different spatial directions. Vertical ground filtering uses criteria appropriate for vertical distance variations, while horizontal ground filtering uses criteria suited for horizontal distance variations. This local quality approach ensures that filtering is adapted to the specific geometric characteristics of ground surfaces in each direction, improving reliability without causing information loss.
3Ease of operation
If sensor information from side surfaces is used for curb detection, then the method is straightforward, but the amount of information is insufficient for distant curbs
Solution Approach 1:
The patent merges the detection results from side surface sensing and top surface sensing into a unified obstacle detection output. The signal mixer combines information from both detection paths, increasing the total quantity of usable sensor information for distant curbs while maintaining the simplicity of the individual detection methods. This merging allows the system to leverage multiple information sources without complicating the overall operation.
Data Source
AI summary
Provided are a device and method for recognizing an obstacle around a vehicle. The device includes a distance sensor configured to generate sensing information regarding each of a plurality of signals by outputting the plurality of signals to an area spaced a predetermined distance from the front of the vehicle, a signal processing unit configured to extract the sensing information regarding signals determined as sensing an obstacle by individually performing vertical ground filtering and horizontal ground filtering on the sensing information regarding each of the plurality of signals, a signal mixer configured to mix the sensing information extracted by performing vertical ground filtering and the sensing information extracted by performing horizontal ground filtering, and an obstacle recognition unit configured to recognize the obstacle using a result of mixing the extracted sensing information.