Object Recognition Device Occlusion Area Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object recognition devices using image processing and radar measurement units can cause erroneous recognition of objects in occlusion areas, leading to loss of object tracking, incorrect control actions, and overlooking of actual objects, due to differing detection capabilities and overlapping detection zones.
Innovation Solution
An object recognition device with an occlusion area detection processing unit, association processing unit, and update processing unit that determines occlusion areas and associates object detection results from multiple sensors to accurately recognize objects, reducing erroneous recognition by excluding occlusion areas and integrating detection information from non-overlapping zones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If the image processing unit and radar measurement unit are used together for object detection, then the detection coverage is improved, but erroneous recognition occurs in occlusion areas where the image processing unit cannot detect targets
Solution Approach 1:
The detection space is segmented into occlusion areas and non-occlusion areas based on the detected surrounding objects. The association processing is then performed separately in these different regions, allowing the system to leverage radar data in occlusion areas while relying on image processing in non-occlusion areas, thus resolving the contradiction between coverage and accuracy
Solution Approach 2:
The occlusion area determination unit acts as an intermediary that identifies regions where image processing fails and directs the association processing unit to use radar measurement results in those specific areas. This mediator component enables the system to maintain high accuracy while preserving comprehensive detection coverage
2Reliability
If association processing is performed in overlapping detection areas, then object tracking is improved, but false associations occur between different objects
Solution Approach 1:
The system applies different association processing strategies to different spatial regions. In occlusion areas, association is performed using radar data with relaxed matching criteria, while in non-occlusion areas, stricter image-based association is applied. This local differentiation resolves the contradiction by adapting the association method to the specific characteristics of each region
Solution Approach 2:
The association processing dynamically adjusts its behavior based on the determined occlusion status of each detection area. The system transitions between different association modes (radar-dominated in occlusion areas, image-dominated in non-occlusion areas) according to real-time environmental conditions, maintaining both tracking continuity and identification accuracy
3Device complexity
If radar measurement unit is used as the sole detection source in overlapping areas, then detection simplicity is improved, but actual objects are overlooked due to erroneous detection
Solution Approach 1:
The radar measurement unit serves multiple functions: it provides primary detection in occlusion areas where image processing fails, and it provides supplementary verification in non-occlusion areas. The association processing unit intelligently switches between relying on radar alone and combining radar with image data, achieving multi-functionality that resolves the contradiction between simplicity and accuracy
Data Source
AI summary
In this object recognition device, an association between a first object detection result and a second object detection result is taken in a region excluding an occlusion area. When the first object detection result and the second object detection result are determined to be detection results for an identical object, a recognition result of the surrounding object is calculated from the first object detection result and the second object detection result. Thus, occurrences of erroneous recognition of an object can be decreased as compared to a conventional object recognition device of a vehicle.