CNN Automatic Target Recognition on Resource-Constrained Satellites
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural networks for image identification and automatic target recognition are computationally intensive and require significant memory, making them challenging to implement in resource-limited environments such as satellites.
Innovation Solution
The implementation of a convolutional neural network (CNN) based automatic target recognition system in a hardware-constrained environment, such as a satellite, using a field programmable gate array (FPGA). This involves breaking up image data into smaller pixel image chips, processing them in parallel, and using quantized weight values to reduce memory and computational requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a neural network is implemented for image identification on a satellite, then automatic target recognition capability is improved, but memory usage and computational complexity increase significantly
Solution Approach 1:
The image data is divided into smaller pixel image chips that can be processed independently and in parallel. This segmentation allows the neural network to process manageable portions of image data, reducing the memory required to hold the entire image in memory at once while maintaining the automatic target recognition capability.
Solution Approach 2:
The patent applies quantization to the neural network weight values, changing them from high-precision floating-point numbers to lower-precision integer representations. This parameter change significantly reduces the memory required to store weights while preserving sufficient accuracy for satellite-based automatic target recognition applications.
2Extent of automation
If a neural network is implemented for image identification on a satellite, then automatic target recognition capability is improved, but processing power requirements increase significantly
Solution Approach 1:
By segmenting the image into pixel chips and processing them in parallel, the computational workload is distributed across multiple processing units. This parallel processing approach reduces the computational complexity burden on any single processor while achieving the desired automatic target recognition performance.
Solution Approach 2:
Quantizing the weight values to lower precision integers simplifies the mathematical operations required during neural network inference. This parameter change reduces computational complexity by enabling the use of simpler arithmetic operations that require less processing power while maintaining adequate recognition accuracy for satellite applications.
Data Source
AI summary
Methods and structures are presented for implementing an automatic target recognition system as a convolutional neural network (CNN) in a satellite or other environment with constrained resources, such as limited memory capacity and limited processing capability. For example, this allows for the automatic target recognition to be implemented on a field programmable gate array (FPGA). Image data is split into subsets of contiguous pixels, with the subsets processed in parallel in a CNN of a corresponding processing node using quantized weight values that are determined in a training process that accounts for the constraints of the automatic target recognition system. The results of the automatic target recognition process is based on the combined output of the processing nodes.