UAV Onboard Image Recognition for Moving Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional UAVs face challenges in recognizing and tracking moving objects due to limited computing power and poor communication quality, leading to difficulties in accurately identifying and monitoring objects over a wide range.
Innovation Solution
A UAV system equipped with an image capturing device, storage medium, and processor that includes modules for object recognition, moving object detection, and tracking, allowing for on-board image recognition and transmission of tracking results, reducing computational load and enabling wider surveillance ranges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If the UAV compresses the captured image and transmits it to a server for computation, then the computing power requirement on the UAV is reduced, but the communication quality deteriorates and the server cannot fully restore the compressed image for accurate recognition
Solution Approach 1:
The patent extracts the image recognition function from the server and implements it on the UAV's onboard processor. By taking out the recognition capability from the remote server and placing it directly on the UAV, the system eliminates the compression and transmission bottleneck while maintaining high recognition accuracy without requiring powerful external computing resources.
Solution Approach 2:
The UAV performs image recognition autonomously using its onboard processor and storage medium. The system serves itself by processing images locally rather than relying on external server assistance. This self-service capability allows the UAV to recognize objects in real-time without depending on communication quality or server processing power.
2Device complexity
If the server observes whether a scene in an image changes to recognize moving objects, then the recognition method is simple for stationary cameras, but it becomes ineffective for moving UAVs where scene changes are caused by the UAV's own movement
Solution Approach 1:
The patent implements a feedback mechanism where the UAV continuously captures images, the onboard processor analyzes scene changes, and the system compares current frame changes with historical data to distinguish between UAV-induced scene changes and actual object movements. This feedback loop enables reliable moving object detection despite the UAV's own movement by continuously adapting to changing scenes.
Solution Approach 2:
The system performs preliminary actions by capturing and storing historical image data before detecting potential moving objects. By having pre-stored reference images and analyzing changes relative to these historical data, the UAV can distinguish between scene changes caused by its own movement and actual object movements, improving detection accuracy before false alarms occur.
3Area of stationary object
If the UAV moves constantly on patrol in the air, then the surveillance coverage is expanded, but the communication quality deteriorates making it difficult for the server to restore compressed images
Solution Approach 1:
The patent extracts the image processing and recognition functions from the server and implements them on the UAV's onboard processor. By taking out these computational tasks from the remote server, the UAV can process high-resolution images locally without relying on communication quality, thus maintaining full image information despite constant movement and poor communication conditions.
Solution Approach 2:
The UAV performs self-service by processing images autonomously using its onboard computational resources. This self-service capability allows the system to maintain high image quality and recognition accuracy regardless of communication conditions or surveillance area size, as the UAV independently analyzes images without needing server assistance.
Data Source
AI summary
The image recognition method includes: obtaining an image data stream, wherein the image data frame includes a current frame; performing image recognition on an object in the current frame to generate a first box corresponding to the current frame; detecting movement of the object to generate a second box corresponding to the current frame; and determining the object as a tracking target according to the first box and the second box.