Roadway Object Classification Using Virtual Surface Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies for classifying objects on a roadway, such as lost cargo, struggle to reliably distinguish raised objects from flat ones, leading to potential false emergency brake activations and liability issues in highly automated driving systems.
Innovation Solution
A method and device that generate a virtual roadway model from markings to identify and classify raised objects using optical flow and parallax deviation, allowing for accurate distance measurement and classification by comparing distance values against a continuity criterion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optical flow-based algorithms or stereo cameras are used for distance estimation, then depth information can be obtained, but reliable distinction between raised and flat objects cannot be achieved
Solution Approach 1:
The patent transitions from two-dimensional image analysis to three-dimensional spatial reasoning by generating a virtual roadway model with depth information. This model incorporates elevation data to represent the roadway surface topology, enabling the system to distinguish raised objects from flat objects based on their vertical dimension. The continuity criterion evaluates distance deviations in 3D space rather than merely in the image plane, providing reliable classification.
Solution Approach 2:
The virtual roadway model serves as an intermediary between the raw image data and the object classification decision. This model acts as a reference framework that mediates the comparison between detected objects and the expected roadway surface, allowing the system to identify deviations that indicate raised objects. The model translates complex image data into a structured spatial representation that facilitates reliable distinction.
2Reliability
If raised objects are detected using current algorithms, then distance information can be obtained, but false classification of flat objects as raised occurs
Solution Approach 1:
The system employs a feedback mechanism where the virtual roadway model provides continuous reference information about the expected roadway surface. The continuity criterion compares actual detected distances against this reference model, and when deviations exceed a threshold, the object is classified as raised. This feedback loop ensures that only objects with significant spatial deviations trigger emergency brake activation, preventing false positives from flat objects like manhole covers.
3Measurement precision
If a virtual roadway model is generated from roadway markings, then a reference for distance comparison is obtained, but the model generation complexity increases
Solution Approach 1:
The system performs preliminary action by pre-generating the virtual roadway model from detected roadway markings before conducting object classification. This model is created in advance as a reference framework, allowing subsequent object detections to be quickly compared against it. The model generation is performed once per scene or roadway segment, and then reused for multiple object classifications, reducing the overall computational burden despite the initial complexity.
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
Enables reliable detection and classification of raised objects, preventing false emergency brake applications and ensuring safe traversability, thereby reducing liability risks and improving the accuracy of object classification in highly automated driving scenarios.
Implementation Method 1
reading in image data from an interface to at least one vehicle camera of the vehicle, the image data representing an area of the surroundings including at least the roadway
Data Source
AI summary
A method for classifying objects on a roadway in surroundings of a vehicle. The method includes: reading in image data from a vehicle camera of the vehicle. The image data represent an area of the surroundings which includes the roadway; evaluating the image data including generating a model of a surface of the roadway using identified roadway markings, and an object on the roadway being identified; ascertaining first distance values between the vehicle camera and object image points of the object represented by the image data, and second distance values between the vehicle camera and roadway image points, defined by the model, of the surface of the roadway in surroundings of the object; and comparing the distance values to at least one continuity criterion for distinguishing raised objects from flat objects to classify the object as a raised or flat object as a function of a result of the comparison.