Aircraft Infrared Target Recognition Using ASIC and FPGA Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional aircraft-based infrared image recognition systems face challenges with high power consumption and limited real-time performance, particularly in distinguishing foreground and background movements during fast platform movement, which complicates target detection and recognition.
Innovation Solution
The system incorporates a combination of specialized ASIC/SoC chips and FPGA processors for multi-level image processing, including an infrared non-uniformity correction SoC chip, image rotation ASIC chip, multi-level filtering ASIC chip, and connected domain labeling ASIC chip, along with DSP and FPGA processors for efficient feature extraction and recognition, enabling real-time processing and reduced power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a moving-platform-based target detection and tracing algorithm is used to handle fast aircraft movement and background variation, then target detection accuracy is improved, but algorithm complexity increases significantly
Solution Approach 1:
The patent segments the complex moving-platform-based algorithm into multiple functional modules: a static platform simulation module that generates virtual static platform images, a difference calculation module that computes differences between actual and virtual images, and a target detection module. This segmentation reduces overall algorithm complexity while maintaining target detection accuracy by handling background variation through virtual image generation rather than direct complex motion compensation.
Solution Approach 2:
The patent introduces a virtual static platform image as an intermediary element. Instead of directly processing the complex relationship between moving platform and target, the system generates a virtual image that represents what the scene would look like from a static platform perspective. This intermediary simplifies the detection process by converting a moving-platform problem into an equivalent static-platform problem with corrected background variation.
2Measurement precision
If a SIFT algorithm is used for image registration to distinguish foreground from background movement, then registration accuracy is improved, but computation burden increases making real-time processing difficult
Solution Approach 1:
The patent extracts and removes the computationally intensive SIFT algorithm from the real-time processing pipeline. Instead of using SIFT for all image registration tasks, the system uses SIFT only for initial feature extraction and matching, then employs simpler correlation-based methods for ongoing real-time registration. This extraction of the heavy computation allows real-time processing while maintaining registration accuracy through the combination of SIFT feature points with efficient correlation algorithms.
Solution Approach 2:
The patent changes the computational parameters of the registration algorithm by switching from full SIFT descriptor comparison to a hybrid approach using SIFT keypoint locations combined with intensity correlation. This parameter change reduces computation burden significantly while maintaining registration accuracy by using SIFT only for robust feature point selection and simpler methods for precise location matching.
3Adaptability or versatility
If conventional recognition algorithms are used for multi-distance target recognition, then recognition capability is improved, but system workload increases significantly
Solution Approach 1:
The patent implements a universal target recognition algorithm that can handle long-distance, medium-distance, and short-distance targets using the same core processing pipeline. Instead of deploying separate specialized algorithms for each distance range, the system uses a unified approach with adaptive parameters that automatically adjust based on target distance and image characteristics, reducing system workload while maintaining multi-distance recognition capability.
Solution Approach 2:
The patent introduces dynamic adaptation into the recognition algorithm, allowing it to automatically adjust its processing parameters and strategies based on the current imaging conditions and target distance. The system dynamically selects appropriate processing depths, filter parameters, and feature extraction levels according to the detected target characteristics, reducing unnecessary computational workload while maintaining recognition accuracy across all distance ranges.
Data Source
AI summary
An aircraft-based infrared image recognition device for a ground moving target, including an infrared non-uniformity correction module, an image rotation module, an image registration module, a multi-level filtering module, a connected domain labeling module, a target detection and feature recognition module, a process control module, and a FPGA-based interconnection module. The invention uses an ASIC/SoC chip for image processing and target recognition, the DSP processor and the FPGA processor, it is possible to enable a multi-level image processing and target recognition algorithm, to improve system parallel, and to facilitate an aircraft-based infrared image recognition method for a ground moving target. Meanwhile, embodiments of the invention effectively reduce power consumption of the device.