Work Vehicle Road Recognition Using Two-Stage 3D Point Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing work vehicles face challenges in accurately distinguishing between roads and non-road areas, particularly when weeds grow on farm fields, leading to potential self-driving interruptions due to erroneous recognition.
Innovation Solution
An area recognition system with a first discrimination processor using image information and a second discrimination processor utilizing height, width, and three-dimensional point group data to re-discriminate areas, enhancing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If only image information from a camera is used for road discrimination, then the system is simple, but the discrimination accuracy deteriorates when weeds grow on farm fields
Solution Approach 1:
The patent transitions from two-dimensional image information to three-dimensional point group data by introducing a LiDAR sensor. This dimensional change enables the system to utilize height information to distinguish between roads and farm fields with weeds, resolving the discrimination accuracy problem while maintaining reasonable system complexity through modular sensor integration.
Solution Approach 2:
The patent changes the discrimination parameters from only visual features (image information) to include spatial features (height, width, depth from point group data). By adding height as a new parameter, the system can differentiate between low-lying weeds on farm fields and the road surface, significantly improving discrimination accuracy.
2Measurement precision
If multiple types of sensors and processing stages are added to improve discrimination accuracy, then the accuracy improves, but the device complexity increases
Solution Approach 1:
The patent divides the discrimination process into two sequential stages: first discrimination processing using image information, and second discrimination processing using point group data. This segmentation allows each processor to specialize in specific types of data analysis, improving overall accuracy while keeping individual processing modules relatively simple and manageable.
Solution Approach 2:
The patent introduces a controller as an intermediary that coordinates between the camera, LiDAR sensor, and discrimination processors. The controller manages data flow, synchronizes processing stages, and integrates results from both image and point group data analysis, enabling complex multi-sensor operation without proportionally increasing system complexity.
3Device complexity
If a single discrimination processor is used, then the system is simple, but self-driving interruptions occur due to erroneous recognition
Solution Approach 1:
The patent performs preliminary discrimination using image information before final confirmation with point group data. The first discrimination processor provides an initial assessment that guides the second discrimination processor to focus computational resources on borderline cases, preventing erroneous recognition and ensuring self-driving continuity through layered verification.
Solution Approach 2:
The patent implements a feedback mechanism where the results from first discrimination processing inform the second discrimination processing stage. The controller uses output from the initial image-based discrimination to adjust and refine the subsequent point group data analysis, creating a closed-loop system that reduces erroneous recognition and improves reliability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An area recognition system includes: a first discrimination processor configured to discriminate whether an area is a road based on first information obtained by setting the surrounding area of a work vehicle 10 as a detection target; and a second discrimination processor configured to discriminate between a road and a non-road area by using identification information on a height or a width based on second information obtained for a range overlapping a range discriminated by using the first information.