FPGA SOSE Pipeline for Parallel Spacecraft Terrain Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing descriptor-based feature matching approaches for spacecraft navigation are not amenable to parallelization or FPGA implementation due to random data access in scale-space representation, leading to inefficient processing and poorer performance in scale-invariance.
Innovation Solution
A novel FPGA implementation of a scale and orientation simultaneous estimation (SOSE) methodology, comprising a scale and orientation (SO) module and an extract descriptor (ED) module, which includes a three-stage computation pipeline for each module, optimizing memory access and parallel processing to estimate feature scale and orientation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If descriptor-based feature matching approaches are used for spacecraft navigation, then feature matching capability is improved, but processing efficiency deteriorates due to random data access in scale-space representation
Solution Approach 1:
The patent segments the scale-space representation into discrete scale levels and organizes feature detection and description into separate modular stages. This segmentation allows independent optimization of each stage and enables parallel processing across different scale levels, resolving the contradiction between maintaining feature matching accuracy and improving processing efficiency.
Solution Approach 2:
The patent implements dynamic pipeline processing where feature detection, scale estimation, and descriptor extraction operate in coordinated stages. The system dynamically adjusts processing based on detected features, enabling efficient resource utilization while maintaining accurate feature matching through adaptive pipeline execution.
2Adaptability or versatility
If scale-space representation with random data access is implemented, then scale-invariant feature detection is achieved, but parallelization and FPGA implementation become difficult
Solution Approach 1:
The patent segments the continuous scale-space representation into discrete, manageable scale levels that can be independently processed. This segmentation transforms the random access pattern into a structured sequence of operations suitable for FPGA implementation, maintaining scale-invariance while reducing device complexity.
Solution Approach 2:
The patent performs preliminary feature detection and scale estimation before descriptor extraction, organizing data access patterns in advance. This preliminary action prepares data in a structured format that eliminates random access requirements during the computationally intensive descriptor extraction phase, enabling efficient FPGA implementation.
3Measurement precision
If traditional terrain relative navigation instruments are used, then navigation accuracy is maintained, but system weight and power consumption increase
Solution Approach 1:
The patent replaces traditional mechanical and electronic navigation instruments with a vision-based system implemented on an FPGA. This substitution uses image processing and feature matching algorithms to achieve navigation accuracy without the heavy hardware, significantly reducing power consumption while maintaining measurement precision.
Solution Approach 2:
The patent creates a virtual model of the terrain by matching features between reference maps and captured images. This copying approach allows the system to infer navigation information from visual data rather than relying on physical sensors, reducing power consumption while maintaining navigation accuracy.
Data Source
AI summary
A system provides descriptor-based feature matching during terrain relative navigation (TRN). A scale and orientation (SO) module acquires a source image, image and slope pixel windows, and ring mask. The SO module combines corresponding pixels from the image pixel window and the slope pixel window to generate intermediate values, accumulates the intermediate values into ring accumulators, sums the accumulated values to generate a final ring value, and determines an orientation stability measure, and final scale and orientation values. An extract descriptors (ED) module acquires the source image, the image and slope pixel windows, final scale and orientation values, sector values, and a rink mask value. The ED module identifies pixels of interest, reorients the sector values. combines corresponding pixels from the image pixel window and the slope pixel window, accumulates and normalizes the intermediate values, and generates an image feature descriptor per coordinate.