Real-Time Aerial Object Detection Using Multi-GPU Parallel Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current visual detection algorithms are inadequate for processing large-scale aerial images from Wide Area Motion Imagery systems, as they are computationally inefficient and struggle with detecting multiple moving objects in high-resolution images, making it difficult to utilize such data for urban surveillance.
Innovation Solution
A method and system utilizing a CPU host and multiple GPUs for parallel computation, performing image segmentation, fast block-wise registration, background and foreground generation, and support vector machine classification, with the Hyper-Q computation framework to enhance processing efficiency and reduce synchronization time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional visual detection algorithms are used for WAMI images, then the algorithms can detect objects in small scenes, but they cannot efficiently process large-scale high-resolution images with multiple moving objects
Solution Approach 1:
The patent divides large-scale WAMI images into multiple smaller image partitions or blocks. Each partition is processed independently by separate detection algorithms running in parallel, which reduces the computational burden on each individual processing unit while maintaining the ability to detect multiple objects across the entire large-scale image.
Solution Approach 2:
The patent transitions from single-CPU sequential processing to multi-GPU parallel processing architecture. By utilizing multiple graphics processing units working simultaneously on different image partitions, the system achieves exponential speedup in processing large-scale images while distributing computational complexity across multiple devices.
2Productivity
If human operators manually monitor WAMI data, then detailed analysis is possible, but it is not feasible to monitor the overwhelming amount of data in real-time
Solution Approach 1:
The patent implements an automated detection system that processes WAMI images without requiring continuous human intervention. The multi-GPU parallel processing system automatically detects, tracks, and classifies moving objects in real-time, enabling the system to handle overwhelming data volumes independently while maintaining rapid response times for surveillance applications.
3Measurement precision
If high-resolution large-scale images are processed, then detection accuracy improves, but processing time increases significantly
Solution Approach 1:
By segmenting high-resolution images into smaller partitions, the system maintains detection accuracy within each partition while enabling parallel processing that dramatically reduces overall processing time. Each partition retains sufficient resolution for accurate object detection, and the aggregation of results from multiple partitions preserves global detection precision.
Solution Approach 2:
The patent performs preliminary processing steps such as image partitioning and feature extraction in advance, preparing the data structure for efficient parallel processing. This preliminary organization of high-resolution image data enables subsequent rapid processing while maintaining the quality and detail necessary for accurate object detection.
Data Source
AI summary
In accordance with various embodiments of the disclosed subject matter, a method and a system for detecting multiple objects from real-time images are provided. The method comprises: performing, using a CPU host, an image segmentation process to divide real-time input images into a plurality of image partitions; performing, by multiple GPUs, a fast block-wise registration process, a mark setting process, a background generation process, a foreground generation process based on a Hyper-Q computation infrastructure, and a support vector machine classification process; and generating, by the CPU host, visualization classification images.