Monocular Camera Lane Detection via Single-View Geometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automotive systems rely on expensive sensors like Radar and Lidar for object detection, and monocular cameras require predefined camera locations, limiting cost-effectiveness and flexibility.
Innovation Solution
A monocular camera system that can be retrofitted and adjusted within a vehicle, using single view geometry and Inverse Perspective Mapping (IPM) to estimate camera height and distance to objects based on lane boundary calculations and known lane widths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple cameras or sensors (Radar, Lidar) are used for object detection, then detection reliability is improved, but system cost increases
Solution Approach 1:
The patent replaces expensive, long-lasting sensors (Radar, Lidar) with a cheap monocular camera that can be easily replaced or repositioned. The camera serves as a disposable or adjustable component that provides sufficient functionality without the high cost of advanced sensors.
Solution Approach 2:
The patent creates a virtual 3D representation of the road scene by processing 2D images from the monocular camera through geometric transformations (IPM, homography). This copying approach reconstructs depth information computationally rather than capturing it directly with expensive 3D sensors.
2Device complexity
If a monocular camera is used for distance estimation, then system cost is reduced, but the camera location must be predefined which reduces flexibility
Solution Approach 1:
The system performs self-calibration by automatically detecting lane markings in the captured images and computing the camera's height and position relative to the road plane. This self-service approach eliminates the need for manual preconfiguration of camera parameters, allowing the camera to be installed at any location and automatically adapt to its position.
Solution Approach 2:
The patent dynamically determines camera parameters (height, position, orientation) based on the detected scene geometry rather than using fixed predefined values. The system adjusts its internal model of the camera position to match the actual installation configuration, enabling flexibility in mounting locations.
3Measurement precision
If deep learning algorithms are used for continuous processing, then measurement precision is improved, but computational resource consumption increases
Solution Approach 1:
The patent replaces continuous deep learning processing with a geometric computation approach based on single-view geometry and projective transformations. This substitution uses mathematical models (IPM, homography matrices) to achieve accurate distance estimation without the high computational cost of continuous neural network inference.
Solution Approach 2:
The system performs preliminary calibration once during initialization to establish the camera's geometric relationship with the road plane. This preliminary action creates a transformation model that can be reused for subsequent distance measurements, eliminating the need for continuous heavy processing while maintaining precision.
Data Source
AI summary
Disclosed are methods, devices, and computer-readable media for detecting lanes and objects in image frames of a monocular camera. In one embodiment, a method is disclosed comprising receiving a sample set of image frames; detecting a plurality of markers in the sample set of image frames using a convolutional neural network (CNN); fitting lines based on the plurality of markers; detecting a plurality of vanishing points based on the lines; identifying a best fitting horizon for the sample set of image frames via a RANSAC algorithm; computing an inverse perspective mapping (IPM) based on the best fitting horizon; and computing a lane width estimate based on the sample set of image frames using the IPM in a rectified view and the parallel line fitting.