Vehicle Target Feature Tracking Using Dynamics-Based Bounding Boxes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated vehicle systems require significant processing time and capability for image processing to track target features, affecting driver assist system efficiencies.
Innovation Solution
A driver assist system that utilizes vehicle dynamics and a bounding box to predict the position of target features in subsequent images, reducing the need for full-image segmentation and enhancing processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full-image segmentation is used to track target features, then tracking accuracy is maintained, but processing time and computational capability requirements increase significantly
Solution Approach 1:
The patent divides the image processing task into two stages: initial full-image segmentation to locate the target feature, followed by bounded-region tracking using a bounding box. This segmentation of the processing task allows the system to maintain accuracy while reducing computational load in subsequent frames by only processing the region of interest rather than the entire image.
Solution Approach 2:
The patent applies different processing quality to different regions of the image. The initial full-image segmentation ensures high accuracy for target detection, while subsequent tracking uses a localized bounding box approach that processes only the relevant region. This local quality approach maintains tracking accuracy for the target feature while significantly reducing the processing burden compared to continuous full-image analysis.
2Reliability
If full-image segmentation is performed for every frame, then target feature detection remains robust, but system efficiency and real-time performance deteriorate
Solution Approach 1:
The patent performs full-image segmentation as a preliminary action only in the initial frame to establish the target feature location and create the bounding box. This preliminary processing ensures robust detection while avoiding the computational cost of repeated full-image segmentation. The bounding box created in advance is then used for efficient tracking in subsequent frames, maintaining reliability while improving productivity.
Solution Approach 2:
The patent applies partial action by performing complete image segmentation only when necessary (initially) and using partial region processing (bounding box) for subsequent tracking. This selective application of processing intensity maintains detection robustness for target acquisition while significantly improving system efficiency during the tracking phase by processing only the necessary portion of each frame.
3Measurement precision
If comprehensive image processing is used to ensure accurate target tracking, then tracking precision is maintained, but computational resource consumption increases
Solution Approach 1:
The patent segments the computational workload by performing comprehensive image processing only initially to establish accurate target location and bounding box parameters. Once the bounding box is established, the system transitions to processing only the bounded region in subsequent frames. This segmentation of computational effort maintains tracking precision while dramatically reducing ongoing computational resource consumption and energy usage.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A vehicle driver assist system provides for the location of a target feature in an initial image and updates the location of that target feature based on information indictive of vehicle dynamic operation. The system predicts a relative position of the target feature within subsequent images such that only a portion of subsequent images need by analyzed and searched.