Camera Road Environment Modeling Using Gamma Maps and Homography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle environment modeling systems using cameras face challenges in accurately detecting road surfaces and hazards in varying weather and lighting conditions, leading to noise interference and increased processing time and power consumption.
Innovation Solution
The system employs a deep neural network trained on unrectified images to produce a gamma map, which represents the height of pixels above a road plane and their distance from the sensor, allowing for efficient road surface modeling without the need for pre-processing, and uses a selected homography to create a model of the road environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional road surface detection methods are used with cameras, then object detection and recognition can be performed, but processing time and power consumption increase significantly
Solution Approach 1:
The system performs preliminary rectification of image coordinates to a road coordinate system before hazard detection. By pre-establishing the geometric transformation relationship between image coordinates and road coordinates, the system avoids complex real-time calculations during hazard detection, thereby reducing processing time while maintaining detection accuracy
Solution Approach 2:
The patent replaces traditional mechanical/geometric road surface modeling methods with a neural network-based approach. The neural network directly predicts hazard locations from rectified image coordinates, substituting complex geometric calculations with learned patterns, which significantly reduces computational burden and processing time
2Measurement precision
If traditional road surface detection methods are used with cameras, then road surface modeling can be achieved, but power consumption increases
Solution Approach 1:
The system performs preliminary rectification of image coordinates to road coordinates before hazard detection. By pre-establishing the geometric transformation relationship, the system avoids complex real-time calculations during hazard detection, thereby reducing processing time while maintaining detection accuracy
Solution Approach 2:
The patent replaces traditional mechanical/geometric road surface modeling methods with a neural network-based approach. The neural network directly predicts hazard locations from rectified image coordinates, substituting complex geometric calculations with learned patterns, which significantly reduces computational burden and power consumption
3Adaptability or versatility
If camera-based systems operate in varying weather and road conditions, then environmental adaptability is improved, but noise interference from shadows, lights, and reflective surfaces increases
Solution Approach 1:
The patent introduces an intermediary coordinate transformation step that maps image coordinates to road coordinates through rectification. This intermediary representation separates the hazard detection task from the complexities of varying lighting and weather conditions in the original image space, allowing the neural network to focus on predicting hazard locations without being distracted by noise from shadows, lights, and reflective surfaces
Data Source
AI summary
System and techniques for vehicle environment modeling with a camera are described herein. A device for modeling an environment comprises: a hardware sensor interface to obtain a sequence of unrectified images representative of a road environment, the sequence of unrectified images including a first unrectified image, a previous unrectified image, and a previous-previous unrectified image; and processing circuitry to: provide the first unrectified image, the previous unrectified image, and the previous-previous unrectified image to an artificial neural network (ANN) to produce a three-dimensional structure of a scene; determine a selected homography; and apply the selected homography to the three-dimensional structure of the scene to create a model of the road environment.