Target Object Tracking with Reduced-Frequency Image Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking solutions based on deep learning models are energy-intensive and require significant computing resources, limiting their deployment in real-world applications.
Innovation Solution
A method for tracking a target object by processing only a portion of images in a stream at a detection frequency lower than the capture frequency, using a tracking phase that skips images between iterations, reducing energy and computing requirements while maintaining performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all images in the stream are processed by object detection and tracking algorithms, then tracking accuracy and reliability are improved, but energy consumption and computational resource requirements increase significantly
Solution Approach 1:
The patent applies partial action by processing only a subset of images from the video stream rather than all images. The system selects specific frames for processing based on criteria such as motion detection or temporal spacing, thereby reducing computational load and energy consumption while maintaining sufficient tracking accuracy through intelligent sampling of the image stream.
Solution Approach 2:
The patent implements periodic action by processing images at reduced frequency intervals rather than continuously. The system processes every Nth frame or uses periodic detection cycles, allowing the tracking algorithm to operate less frequently while still maintaining reliable target tracking through temporal interpolation and motion prediction between processed frames.
2Reliability
If deep learning models are used for object detection and tracking, then tracking performance is improved, but device complexity and computational resource requirements increase
Solution Approach 1:
The patent reduces device complexity by applying deep learning models only to selected images rather than processing every frame. This partial application of complex algorithms significantly reduces computational resource requirements while maintaining tracking performance through strategic selection of key frames for model processing.
Solution Approach 2:
The patent segments the image stream processing into multiple stages: preliminary filtering of images that require processing, selective application of deep learning models to filtered frames, and temporal interpolation for intermediate frames. This segmentation allows complex computational tasks to be distributed and optimized, reducing overall device complexity while maintaining performance.
3Speed
If tracking is performed in real-time on all captured images, then tracking speed and responsiveness are improved, but energy consumption increases
Solution Approach 1:
The patent applies periodic action by performing tracking operations at reduced frequency intervals rather than on every captured image. The system processes images periodically at optimized intervals, reducing energy consumption while maintaining adequate tracking speed through temporal interpolation and motion prediction algorithms that estimate target position between processed frames.
Solution Approach 2:
The patent reduces energy consumption by applying tracking algorithms only to a partial subset of captured images rather than all images. The system identifies and processes only those frames that contain sufficient information for effective tracking updates, thereby reducing overall energy consumption while maintaining tracking responsiveness through intelligent frame selection.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The invention relates to a method (100) for tracking a target object in an image stream captured by a camera at a capture frequency, Fc. The method (100) comprises several iterations of a tracking phase (110), implemented individually for several images, referred to as processed images, of said image stream, and comprising the following steps: - detection (114) of at least one object, and its position, in the processed image, and - identification (116-124) of said target object among the at least one object detected in said processed image. The method is characterized in that said detection phase (110) is performed at a detection frequency, Fs, lower than said capture frequency Fc, such that two images processed during two successive iterations of the tracking phase are separated from at least one unprocessed image to which said tracking phase is not applied. The invention also relates to a computer program, a device, and a system implementing such a method.