Neural Network Freespace Detection for Automotive Driver Assistance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current driver assistance systems face challenges in accurately detecting accessible freespace, particularly in complex scenarios like parking, due to limitations in object detection methods that often miss objects and result in false positives or negatives, which can impact safety and accuracy in semi-automated or fully automated driving.
Innovation Solution
The method employs a capturing device with multiple cameras and a computing device using dense depth maps, neural networks, and classical algorithms to detect and classify objects at the border of freespace, categorizing it into semantic regions based on object types, thereby providing a robust and accurate semantic freespace detection that adapts driving behavior to the environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If object detection methods are used to detect obstacles in the environment, then obstacles can be identified, but missed objects and false positives occur reducing detection accuracy
Solution Approach 1:
The patent combines multiple detection approaches: classical freespace detection methods (based on depth information and geometric constraints) are merged with neural network-based object detection. This integration allows the system to leverage the strengths of both methods - the geometric accuracy of classical approaches and the pattern recognition capabilities of neural networks - to reduce missed objects and false positives in freespace detection
Solution Approach 2:
The patent introduces an intermediary processing step where detected objects are used to refine and correct freespace detection results. The neural network detects objects that may be missed by classical methods, and these detections serve as intermediary information to adjust the freespace map, thereby improving overall detection precision and reliability
2Productivity
If classical freespace detection methods are used, then depth information can be processed, but false positives occur and static obstacles may be incorrectly erased
Solution Approach 1:
The patent implements a feedback mechanism where neural network object detection results are fed back to correct and refine the freespace detection output. When the neural network detects objects that classical methods missed or misclassified, this information feedback allows the system to adjust the freespace boundaries, reducing false positives while maintaining detection speed through efficient processing pipelines
3Measurement precision
If dense depth maps with millions of flow vectors are used, then freespace detection accuracy is improved, but computational complexity increases significantly
Solution Approach 1:
The patent applies segmentation by dividing the computational domain into relevant regions. Instead of processing all pixels uniformly, the system focuses computational resources on border regions and areas with high uncertainty, while using coarser representations in stable regions. This segmentation strategy maintains high precision where needed while reducing overall computational complexity
Solution Approach 2:
The patent uses partial action by applying full dense depth map processing only to critical regions (such as areas near detected objects or boundaries), while using sparse or simplified processing in other regions. This selective approach achieves sufficient precision for safety-critical areas without the excessive computational cost of uniform dense processing throughout the entire scene
Data Source
AI summary
The invention relates to a method for operating a driver assistance system (2) of a motor vehicle (1), including a) Capturing an environment (4) of the motor vehicle (1) by a capturing device (3) of the driver assistance system (2); b) Detecting an accessible freespace (6) in the captured environment (4) by a computing device (5) of the driver assistance system (2); c) Detecting and Classifying at least one object (7a-7e) in the captured environment (4) that is located at a border (8) of the freespace (6) by a neural network (9) of the driver assistance system (2); d) Assigning a part (10a-10e) of the border (8) of the freespace (6) to the detected and classified object (7a-7e); and e) Categorizing a part (11a-11e) of the freespace (6) adjacent to the part (10a-10e) of the border (8) that is assigned to the detected and classified object (7a-7e) in dependence upon the class of that classified object (7a-7e), so as to enable improved safety in driving.