Dual-Buffer Frame Aggregation for Slow-Moving Object Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
ADASs using temporal aggregation struggle to effectively track and predict the trajectory of slower moving objects, such as pedestrians and decelerating vehicles, due to their reliance on time-based buffers.
Innovation Solution
Implementing a distance-based buffer in conjunction with a time-based buffer for frame storage, where frames are stored based on the distance traveled by the vehicle, allowing for better recognition and navigation of slower moving objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a time-based buffer is used for frame storage in ADAS, then the system can effectively track and predict the trajectory of fast-moving objects, but it performs poorly with respect to tracking and predicting the trajectory of slower moving objects such as pedestrians and decelerating vehicles
Solution Approach 1:
The patent segments the frame storage mechanism into two distinct buffers: a time-based buffer for capturing frames at fixed time intervals (effective for fast-moving objects) and a distance-based buffer for capturing frames at fixed distance intervals (effective for slow-moving objects). This segmentation allows the system to handle different object speed regimes with specialized buffers, resolving the contradiction between tracking speed and tracking reliability for slow-moving objects.
Solution Approach 2:
The patent implements a dynamic frame selection mechanism that adapts to the speed of detected objects. The system dynamically determines which buffer (time-based or distance-based) to use for tracking based on the detected object's speed characteristics. For fast-moving objects, the time-based buffer is prioritized, while for slow-moving objects, the distance-based buffer is prioritized, making the system's behavior adaptive rather than static.
2Productivity
If frames are stored at fixed time intervals, then the system operates efficiently with consistent processing rates, but it fails to capture sufficient temporal information for slow-moving objects that exhibit minimal movement between frames
Solution Approach 1:
The patent changes the storage parameter from purely time-based to include distance-based sampling. By storing frames at fixed distance intervals in addition to fixed time intervals, the system ensures that sufficient temporal and spatial information is captured for slow-moving objects. This parameter change allows the system to maintain processing efficiency while preventing information loss for slow-moving objects.
3Reliability
If a distance-based buffer is implemented in addition to a time-based buffer, then the system can better recognize and navigate slower moving objects, but the device complexity increases
Solution Approach 1:
The patent designs the dual-buffer system with a unified management architecture that handles both time-based and distance-based buffers through common processing pipelines. The frame selection module universally processes inputs from both buffers using the same object detection and tracking algorithms, reducing the complexity increment. This multi-functional design allows the system to handle both buffer types without requiring completely separate processing systems.
4Duration of action of moving object
If the system stores more frames to improve tracking of slow-moving objects, then the time window for frame capture is extended, but the computational load and processing time increase
Solution Approach 1:
The patent implements a selective frame storage strategy where only frames meeting specific distance-based criteria are stored in the distance-based buffer, rather than storing all frames indiscriminately. This partial action approach extends the effective time window for slow-moving object tracking while avoiding the computational burden of processing excessive frames, thus balancing the trade-off between duration and processing time.
Data Source
AI summary
A system for navigation store, in a time-based buffer, a first set of frames acquired by a sensor; stores, in a distance-based buffer, a second set of frames acquired by the sensor; performs moving object segmentation on the first set of frames and the second set of frames to identify at least one moving object in a scene of the frames; predicts a trajectory of the at least one moving object; and performs a navigation function based on the predicted trajectory.