Neural Network Object Tracking With Kalman ROI Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection and tracking systems face challenges in efficiently and accurately tracking moving objects, particularly when resources are limited, and there is a need for real-time adjustments in field of view to maintain focus on the object.
Innovation Solution
A system utilizing a neural network for object detection and a Kalman filter for predicting future object locations, combined with image cropping and control signals for adjusting the camera's field of view, such as through gimbal and zoom controls, to maintain object tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a neural network is used for object detection and tracking, then detection accuracy is improved, but computational resource requirements increase
Solution Approach 1:
The system segments the image processing task by cropping the image to a region of interest (ROI) before neural network processing. The ROI is determined based on predicted object location from a Kalman filter, reducing the image size and computational burden while maintaining detection accuracy within the focused region.
Solution Approach 2:
The system performs preliminary action by using the Kalman filter to predict the object's future location before neural network detection. This prediction allows the system to pre-crop the image to the relevant region, reducing computational resources needed for full-image processing while maintaining high detection accuracy.
2Measurement precision
If the field of view is adjusted in real-time to track moving objects, then tracking accuracy is improved, but system complexity increases
Solution Approach 1:
The system introduces a Kalman filter as an intermediary component between the neural network detector and the field of view adjustment mechanism. The Kalman filter predicts object location and provides this information to control image cropping and camera positioning, enabling accurate tracking without direct complex coordination between detection and actuation systems.
Solution Approach 2:
The system implements feedback by continuously updating the Kalman filter with detected object locations and using the filter's predictions to adjust the field of view. This closed-loop feedback mechanism maintains tracking accuracy while managing system complexity through a structured information flow between components.
3Productivity
If image cropping is used to reduce processing area, then processing speed is improved, but risk of losing the object increases
Solution Approach 1:
The system performs preliminary action by using the Kalman filter to predict where the object will be located before cropping the image. This prediction ensures that the cropped region is positioned correctly to contain the object, maintaining both processing speed through reduced image size and reliability through accurate object retention.
Solution Approach 2:
The system uses feedback from the neural network's detection results to continuously update the Kalman filter's state estimation. This feedback loop ensures that the predicted location used for cropping remains accurate, preventing object loss while maintaining fast processing through efficient region-of-interest extraction.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
An apparatus, method and computer program is described comprising detecting a first object in a first image of a sequence of images using a neural network (22), wherein the means for detecting the first object provides an object area indicative of a first location of the first object; and tracking the first object (24), wherein the means for tracking the first object further comprises generating a predicted future location of the first object and generating an updated location of the first object using the neural network. The means for generating the predicted future location of the first object may, for example, receive said object area indicative of a first location of the first object and may receive said updated location information of the first object.