Vehicle Vision Quality Evaluation for ROI-Based Function Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing evaluation methods for deep neural networks in vehicle perception systems do not account for the difficulty of input data, which is crucial for reliable and fast object detection in various driving scenarios, especially in highly automated driving systems.
Innovation Solution
A method for determining a local quality parameter that assesses the difficulty of a specific region of interest in sensor data, using metrics such as intensity and size contrast with the background, to adjust vehicle functions accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing evaluation methods are used for deep neural networks in vehicle perception systems, then the evaluation process is simple, but the reliability of object detection is insufficient because input data difficulty is not accounted for
Solution Approach 1:
The evaluation process is segmented into multiple components: determining a region of interest in sensor data, determining a background region, calculating intensity contrast between the region of interest and background, calculating size contrast, and combining these to determine a local quality parameter. This segmentation allows the system to account for input data difficulty while maintaining a structured evaluation approach.
Solution Approach 2:
The patent introduces a new dimension of evaluation by adding the local quality parameter that assesses input data difficulty. This goes beyond traditional evaluation metrics by incorporating intensity contrast and size contrast dimensions, creating a multi-dimensional evaluation framework that captures the complexity of real-world driving scenarios.
2Measurement precision
If the local quality parameter is determined using multiple metrics (intensity contrast, size contrast), then the evaluation accuracy improves, but the computational complexity increases
Solution Approach 1:
The evaluation focuses on local quality by determining a region of interest and its corresponding background region, then calculating contrasts specifically for these regions. This localized approach allows precise measurement of sensor data quality in critical areas without unnecessarily processing the entire sensor data set, balancing precision with computational efficiency.
Solution Approach 2:
The system performs preliminary actions by determining the region of interest and background region before calculating the final quality parameter. This preliminary segmentation allows the subsequent intensity and size contrast calculations to be performed on focused, relevant data portions, reducing overall computational complexity while maintaining measurement precision.
3Productivity
If the vehicle function is adjusted based on the local quality parameter, then the safety and efficiency of vehicle operations improve, but the control system complexity increases
Solution Approach 1:
The system implements feedback by using the determined local quality parameter to adjust the execution of the vehicle function. The control device receives the local quality parameter and modifies vehicle function execution accordingly, creating a closed-loop system that continuously adapts to input data quality, thereby improving safety and efficiency with a relatively simple feedback mechanism.
Data Source
AI summary
Disclosed is a method for automatically executing a vehicle function of a vehicle based on spatially resolved raw sensor data for environment perception generated by at least one sensor of the vehicle. The method comprisesreceiving spatially resolved raw sensor data generated by the at least one sensor andprocessing sensor data which are characteristic of the spatially resolved raw sensor data using a processor. The processor may determine at least one region of interest of the sensor data and at least one class for classifying the region of interest. The method further comprisesprocessing the sensor data based on the determined region of interest and hereby determining at least one local quality parameter which is characteristic for the quality of the sensor data with respect to at least a section of the region of interest andexecuting the vehicle function in dependence on the local quality parameter.