Driveable Space Boundary Detection Using Entropy and Color Bins
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current driver aids and autonomous vehicle sensors struggle to accurately identify and differentiate between driveable and non-driveable surfaces, leading to potential hazards such as driving onto unsafe terrain or objects, and require detailed mapping or multiple sensors.
Innovation Solution
A method involving capturing color images, computing histograms of oriented gradients, assigning entropy and color characteristic values to image bins, and using a classifier to determine driveable space, with spatial and temporal filtering to smooth outputs and provide continuous boundary detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detailed mapping and multiple sensors are used to identify driveable surfaces, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The image is divided into multiple bins corresponding to a rectangular grid in the real world. Each bin is processed independently to calculate entropy and color characteristics, enabling localized analysis of driveable surfaces without requiring complex global mapping
Solution Approach 2:
The patent replaces complex mechanical mapping systems and multiple physical sensors with image processing algorithms. By analyzing color images through entropy and color characteristic calculations, the system achieves accurate driveable space identification without physical contact or complex hardware
2Reliability
If ultrasound sensors are used to detect objects, then reliability is improved, but adaptability deteriorates as they cannot detect non-reflective surfaces
Solution Approach 1:
The image processing system serves multiple functions: detecting objects, identifying surface types (driveable vs. non-driveable), and providing spatial awareness. By analyzing both entropy and color characteristics, a single system achieves what previously required multiple specialized sensors
Solution Approach 2:
The system changes the detection parameters from ultrasound reflection strength to image-based entropy and color characteristics. This allows detection of surfaces that do not reflect ultrasound well, such as soft sand or heavily rutted terrain, by analyzing their visual texture and color properties instead
Data Source
AI summary
A method of determining the boundary of a driveable space in a scene around a vehicle, comprises capturing a first colour image of the scene, computing a set of histograms of oriented gradients, for instance using a HOG algorithm, each histogram corresponding to a cell in a set of cells, assigning an entropy value to the cell by computing the entropy of the histogram for the cell, dividing the image into bins in a way that corresponds to a rectangular grid in the real world, and calculating an overall entropy value for each bin from the entropy values for the cells in the bin and an overall colour characteristic value for each bin. The entropy value and colour characteristic value for each bin are fed into a classifier that is configured to classify regions of the image corresponding to each bin as regions likely to be driveable space or that are not likely to be driveable space, from which the boundary is derived.