Drivable Area Recognition on Unpaved Roads Using Frame Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
General image recognition systems for vehicles fail to provide effective driving assistance functions in unpaved or unimproved roads, as they cannot accurately identify drivable areas and issue timely warnings.
Innovation Solution
An image recognition system and method that utilizes an image sensor and processor to analyze sensed images, determine valid frames, segment drivable areas, and generate danger warnings when the proportion of drivable area in sub-image regions falls below a threshold, using a determination model and deep learning model for accurate identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a general image recognition system is used for vehicle driving assistance, then lane line and lane shift detection functions can be provided, but the system fails to work effectively in unpaved or unimproved road environments
Solution Approach 1:
The image processing system segments the sensed image into multiple sub-image regions and independently analyzes the drivable area proportion in each region. This segmentation approach allows the system to adapt to unpaved roads by detecting drivable areas locally rather than relying on global road structure assumptions, thereby improving reliability in diverse road environments
Solution Approach 2:
The system changes the detection parameters by evaluating the proportion of drivable area in each sub-image region and comparing it against a preset threshold. This parameter-based approach enables the system to adapt to different road types (paved vs. unpaved) by dynamically assessing local drivable area proportions rather than relying on fixed road structure expectations
2Measurement precision
If the system divides the image into multiple sub-image regions for analysis, then the detection accuracy of drivable areas improves, but the computational complexity and processing time increase
Solution Approach 1:
The system divides the sensed image into multiple sub-image regions to improve detection precision by analyzing drivable area proportions locally. While this increases computational complexity, the segmentation enables parallel processing and focused analysis of critical regions, making the complexity manageable and the precision gain worthwhile for safety-critical applications
3Reliability
If the system processes multiple continuous frames and accumulates danger marks, then the reliability of danger warning increases, but the processing time and computational load increase
Solution Approach 1:
The system performs preliminary analysis on each frame by determining whether it contains a road and identifying valid frames before accumulating danger marks. This preliminary action filters out irrelevant frames early in the processing pipeline, reducing the computational load for subsequent danger mark accumulation while maintaining reliable warning detection across multiple frames
Data Source
AI summary
The image recognition system includes an image sensor and an image processor. The image sensor acquires a plurality of sensed images of continuous frames. The image processor performs image recognition on the plurality of sensed images. The image processor determines on the plurality of sensed images respectively whether the sensed image is a valid frame not containing a road, to determine a plurality of valid sensed images. The image processor determines from the plurality of valid sensed images respectively whether a proportion of an area of a drivable area in a plurality of sub-image regions of the valid sensed image is less than a preset proportion, to perform danger marking. The image processor calculates the number of danger marks of the sub-image regions for each of the valid sensed images. When the number of danger marks is greater than a preset threshold, the image processor generates a danger warning.