Moving Object Imaging with Bounding-Box AX Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computer vision techniques struggle to dynamically control auto exposure, auto focus, and auto white balance for objects moving relative to a camera, especially in dynamic environments like space or military scenarios, leading to suboptimal image quality due to incorrect settings that cannot be corrected post-capture.
Innovation Solution
A method using neural networks to generate candidate bounding boxes, compute feature vectors, and select the closest bounding box to the target object, providing partial image statistics for real-time control of auto exposure, focus, and white balance without requiring training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional computer vision techniques (background subtraction, optical flow, template matching) are used for object tracking, then object location can be determined in image frames, but the ability to dynamically control AX algorithms for optimal image quality is insufficient
Solution Approach 1:
The system implements feedback by continuously tracking the target object across frames and using the determined location to dynamically adjust AX algorithm parameters. The tracking output feeds into the AX control system, which adjusts exposure, focus, and white balance based on the object's current position and characteristics in the image frame.
Solution Approach 2:
The system transitions from static AX settings to dynamic control by continuously updating AX parameters based on real-time object tracking data. As the object moves through different positions and conditions in successive frames, the AX algorithms adapt their parameters dynamically to maintain optimal image quality throughout the sequence.
2Productivity
If AX settings are determined at capture time for moving objects, then real-time image quality optimization is achieved, but incorrect settings cannot be fully compensated post-capture
Solution Approach 1:
The system performs preliminary action by determining appropriate AX settings before the actual image capture of each frame. The object tracking system predicts the object's position and characteristics in the upcoming frame, pre-adjusts the AX parameters accordingly, and then captures the image with optimized settings already in place, ensuring quality is built-in rather than corrected later.
3Quantity of substance
If the entire frame is used for AX algorithm input, then comprehensive image statistics are available, but the ability to optimize for a specific moving object in an uncluttered scene is reduced
Solution Approach 1:
The system applies segmentation by dividing the image frame into regions of interest based on the tracked object's location. Instead of processing statistics from the entire frame, the system focuses computational resources on extracting and analyzing image statistics specifically from the region containing the target object, thereby optimizing AX parameters for the object of interest while ignoring irrelevant background areas.
4Adaptability or versatility
If object tracking is performed to extract information from image frames, then the location of the target object can be determined, but the complexity of the system increases
Solution Approach 1:
The system achieves universality by implementing a multi-functional tracking and control architecture. The same tracking infrastructure serves multiple purposes: determining object location, characterizing object appearance, predicting object motion, and providing input to AX algorithms. This consolidated approach avoids the need for separate specialized systems for each function, thereby reducing overall system complexity while maintaining adaptability.
Data Source
AI summary
The system and methods of the various embodiments enable better capturing of images of objects that are moving relative to a camera. A computing system may be configured to capture a frame containing a target object, generate at least two candidate object bounding boxes for the target object, and determine a candidate object feature vector for the part of the frame contained within each candidate object bounding box. The computing system may compute the distance between each candidate object feature vector and a target object feature vector, selecting the candidate object bounding box associated with the candidate object feature vector that is closest to the target object feature vector, generate partial image statistics on the part of the frame contained within the selected candidate object bounding box, and provide the partial image statistics as input to an AX algorithm that controls the camera.