Rear Vision Object Tracking for Stable Reverse Collision Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle imaging systems struggle to accurately detect objects behind the vehicle during reversing maneuvers, particularly in low-speed conditions and when objects are small or moving, leading to potential collisions due to false positives and intermittent detection.
Innovation Solution
A vision system equipped with a rear-facing camera and image processing algorithms that segment images into Points-of-Interest (POIs), track motion vectors, and utilize filters to stabilize object detection, including a prediction filter, confirmation filter, and Kalman filter to minimize false positives and ensure consistent object tracking across frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image processing algorithms are used to detect objects behind the vehicle, then object detection capability is improved, but false positives and intermittent detection occur leading to reduced reliability
Solution Approach 1:
The system performs preliminary actions by segmenting the image into Points-of-Interest (POIs) and tracking their motion vectors before making detection decisions. This preliminary processing allows the system to establish a baseline of expected motion patterns, which improves the reliability of subsequent detection by comparing actual observations against these pre-established expectations.
Solution Approach 2:
The system implements feedback mechanisms through confirmation filters that verify detection results against multiple criteria including motion vector consistency, POI tracking stability, and detection persistence across frames. This feedback loop eliminates false positives by requiring that detected objects satisfy multiple confirming conditions before being reported, thereby improving detection reliability.
2Measurement precision
If motion vector tracking is used to detect moving objects, then detection of small or moving objects is improved, but false positives increase due to noise and intermittent detection
Solution Approach 1:
The confirmation filter implements feedback by continuously monitoring motion vector patterns and comparing them against expected patterns for legitimate objects. When motion vectors exhibit characteristics of noise or intermittent detection rather than consistent object motion, the feedback mechanism rejects the detection, thereby reducing false positives while maintaining sensitivity to actual moving objects.
Solution Approach 2:
The system applies dynamics by adapting its detection thresholds and criteria based on the observed motion patterns and environmental conditions. The confirmation filter dynamically adjusts its stringency based on the consistency of motion vectors across frames, allowing flexible detection that maintains high precision while adapting to varying scene conditions.
3Reliability
If multiple filters are applied to stabilize object detection, then detection stability is improved, but system complexity increases
Solution Approach 1:
The system segments the detection process into distinct functional modules: POI segmentation, motion vector tracking, prediction filtering, and confirmation filtering. Each module handles a specific aspect of the detection task, which improves reliability through specialized processing while managing complexity by organizing functions into separable, maintainable units with clear interfaces.
Solution Approach 2:
The confirmation filter serves multiple functions simultaneously: it verifies detection results, filters false positives, tracks object consistency, and validates motion patterns. This multi-functionality improves detection stability without proportionally increasing system complexity, as a single unified filter structure performs multiple detection-related tasks.
Data Source
AI summary
A vehicular vision system includes a vehicular camera disposed at a vehicle and operable to capture frames of image data. The vehicular vision system is operable to detect, via processing by an image processor of multiple frames of captured image data, when multiple objects are viewed by the vehicular camera. During a driving maneuver, and when multiple objects are detected, the vehicular vision system determines which detected object of the multiple detected objects is closest to the vehicle, and determines movement of at least the determined closest object relative to the vehicle. During the driving maneuver, and based at least in part on the determined relative movement of the determined closest object, the vehicular vision system determines whether the determined relative movement is indicative of the determined closest object being in or moving toward a position where collision may occur between the determined closest object and the vehicle.