Occupancy Map Generation Using Optical Flow and Velocity Sensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating occupancy maps around vehicles rely on expensive and difficult-to-handle sensors like stereo-cameras, LIDAR, or require large labeled databases for image-based semantic segmentation, which struggle to detect free spaces like lawns effectively, posing challenges for collision avoidance systems.
Innovation Solution
A method using a single forward-looking camera and a simple velocity sensor to capture consecutive images, estimate 3D points via structure from motion, and generate occupancy probabilities through log-odds ratios, eliminating the need for extensive training and leveraging conventional techniques like ray tracing and optical flow analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo-camera, LIDAR or radar based sensors are used, then measurement precision of occupancy map is improved, but device complexity and cost increase
Solution Approach 1:
The patent combines a camera and velocity sensor data through optical flow analysis to achieve occupancy map generation without requiring complex stereo-cameras, LIDAR, or radar systems. The optical flow vectors from the single camera are integrated with velocity information to estimate 3D point positions, merging multiple data sources to compensate for the simpler sensor configuration.
Solution Approach 2:
The patent introduces optical flow vectors as an intermediary element that connects the single camera images with 3D space estimation. By computing optical flow between consecutive images and using it to estimate 3D point positions, the system bridges the gap between simple 2D imaging and 3D occupancy mapping without requiring complex sensors.
2Device complexity
If image-based semantic segmentation is used, then device complexity is reduced, but reliability of free space detection deteriorates
Solution Approach 1:
The patent transitions from 2D image-based semantic segmentation to 3D occupancy map generation by estimating 3D point positions from optical flow vectors. This dimensional transformation allows the system to detect free spaces in three-dimensional space, improving reliability for collision avoidance by considering vertical clearance and spatial depth that 2D segmentation cannot capture.
Solution Approach 2:
The patent changes the fundamental parameters from 2D image pixel classification to 3D spatial occupancy probability. By computing occupancy probabilities for 3D grid cells based on optical flow-derived 3D points and ray-tracing, the system transforms the problem from texture-based classification to geometry-based spatial reasoning, improving detection of lawns and other free spaces.
3Productivity
If semantic segmentation with large labeled database training is used, then occupancy map generation is achieved, but loss of time for training and data preparation increases
Solution Approach 1:
The patent implements a self-service approach where the system generates occupancy maps in real-time using optical flow computation from consecutive camera images and velocity sensor data, without requiring external training databases or pre-processing. The method autonomously computes 3D point estimates and occupancy probabilities on-demand, eliminating time-consuming training phases while maintaining productivity in occupancy map generation.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach provides an efficient and effective occupancy map generation without the need for expensive sensors or extensive training, enabling accurate detection of free and occupied spaces for driver assistance and autonomous driving applications, including path planning and collision avoidance.
Implementation Method 1
two consecutive images are captured by a camera mounted on the vehicle
Data Source
Figure 1~2
Figure 3A~3B
Figure 4~5
AI summary
A method of generating an occupancy map representing free and occupied space around a vehicle, wherein the occupancy map is divided into a plurality of cells Mx,y, the method comprising: capturing two consecutive images by a camera mounted on the vehicle; generating optical flow vectors from the two consecutive images; estimating 3D points in the space around the vehicle from the optical flow vectors; generating rays from the camera to each of the estimated 3D points, wherein intersection points of the rays with the cells Mx,y defining further 3D points; determining for each of the cells Mx,y a function Lx,yt for a time step t; and determining an occupancy probability from the function Lx,yt for each of the cells Mx,y.