Dynamic Region of Interest for Radar Ghost Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection systems in autonomous vehicles face challenges with ghost detections due to multi-path propagation in radar signals, which are difficult to distinguish from valid reflections using existing filters, leading to noise, clutter, and interference.
Innovation Solution
The system defines a dynamic region of interest (ROI) based on various factors such as the field-of-view of sensors, road geometry, speed limits, and potential encounters, where objects outside this ROI are disregarded to reduce ghost detections, and those inside are identified as relevant for scene building.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If radar sensors are used for object detection in autonomous vehicles, then the system becomes more robust against adverse weather conditions, but the system becomes more susceptible to ghost detections caused by multi-path propagation
Solution Approach 1:
The patent divides the detection space into multiple regions of interest (ROIs) based on road geometry, vehicle maneuver, and speed limit. By segmenting the environment into relevant and irrelevant zones, the system can selectively process radar returns only from meaningful areas, thereby reducing ghost detections while preserving weather robustness
Solution Approach 2:
The patent applies different processing qualities to different spatial regions. High-priority processing is applied to objects within the dynamic ROI, while low-priority or discarded processing is applied to objects outside the ROI. This local differentiation reduces computational overhead and eliminates ghost detections in irrelevant areas
2Measurement precision
If existing filters such as CFAR are used to discern between noise and valid reflections, then the system attempts to reduce noise, but ghost detections with similar frequency to valid reflections still surpass these filters
Solution Approach 1:
The patent introduces a geometric intermediary (the dynamic ROI defined by road geometry and vehicle context) between the radar signal processing and object classification stages. This intermediary spatial filter acts as an additional layer of discrimination that complements traditional amplitude-based filters like CFAR, enabling the system to reject ghost detections that have similar frequency characteristics to valid reflections
3Reliability
If the system processes all detected objects for scene building, then comprehensive object detection is achieved, but computational overhead and processing time increase
Solution Approach 1:
The patent extracts and processes only the essential subset of detected objects that fall within the dynamic ROI, discarding objects outside this region. This extraction approach maintains comprehensive detection within relevant areas while significantly reducing computational overhead by eliminating processing of irrelevant objects
Solution Approach 2:
The patent applies partial action by selectively processing only those objects that are material to navigation based on the dynamic ROI criteria. Rather than processing all detected objects equally, the system applies processing intensity proportional to the object's relevance to current vehicle operations
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
This approach effectively reduces the likelihood of ghost detections, simplifies data processing, and enhances the accuracy of autonomous driving systems by focusing on relevant objects within the dynamic ROI, thereby improving navigation and reducing communication overhead.
Implementation Method 1
Radar sensors are more robust against adverse weather conditions such as fog, snow, and rain when compared to cameras or LiDAR
Implementation Method 2
sometimes the reflected signal follows an indirect reflection path between the radar sensor and the target object, which may be referred to as multi-path propagation
Data Source
AI summary
An objection detection system for a vehicle includes one or more range sensors and one or more controllers in communication with the one or more range sensors. The one or more controllers execute instructions to instruct the one or more range sensors to emit a signal. The one or more range sensors receive a reflected signal from the environment surrounding the vehicle. The one or more controllers determine a position of the detected object based on the reflected signal and compare the position of the detected object with a dynamic region of interest (ROI). The one or more controllers determine the position of the detected object is outside of the dynamic ROI. In response to determining the position of the detected object is outside the dynamic ROI, the one or more controllers disregard the detected object for purposes of scene building.