Frequency Imaging for High-Speed Video Information Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video camera technologies, both in visible and infrared wavebands, are limited in extracting new information due to their reliance on pattern recognition and contrast methods, which have reached the limits of utility in processing high frame rate and long video recording times.
Innovation Solution
Frequency Imaging, a technique that processes and visualizes images in the frequency domain to extract time-varying information from scenes, using high-speed cameras to reveal behaviors and objects not detectable by conventional methods, by transforming time series data from pixels into frequency information and displaying it as spectral power across pixels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional pattern recognition or contrast methods are used to process video images, then objects or actions can be identified based on silhouette, shape, color, or thermal signature, but these methods have reached the limit of utility in extracting new information from high frame rate camera technology
Solution Approach 1:
The patent transforms video data from the time domain to the frequency domain by analyzing temporal variations in pixel intensity values. This parameter transformation allows extraction of frequency-based characteristics that are invisible to conventional pattern recognition and contrast methods, thereby overcoming the utility limits of traditional approaches and enabling new information extraction from high frame rate camera data
Solution Approach 2:
The invention adds a frequency dimension to traditional video analysis by computing frequency spectra from time-series pixel data. This dimensional transformation from spatial-temporal analysis to frequency-domain analysis creates a new information space where previously undetectable objects and behaviors become visible, resolving the information extraction bottleneck
2Speed
If high-speed cameras operating at 1000 frames per second or higher are used to capture rapid behaviors, then very high frequency behavior can be revealed, but scene resolution decreases
Solution Approach 1:
The patent applies frequency analysis to transform the evaluation criteria from spatial resolution metrics to temporal frequency metrics. By analyzing the frequency content of pixel intensity variations over time, the system can detect and characterize high-frequency behaviors even in lower-resolution images, effectively decoupling the trade-off between frame rate and resolution
Solution Approach 2:
The invention introduces frequency spectrum analysis as an intermediary processing stage between image capture and interpretation. This intermediary transformation extracts meaningful behavioral information from temporal variations, allowing high-speed capture to reveal dynamics that compensate for reduced spatial resolution
Data Source
AI summary
Frequency imaging of different areas or object in an image is created by a visible light, infrared or other cameras taking multiple sequential images is disclosed. The images are recorded and stacked. Pixels that vary in the images yield time varying data on a pixel by pixel basis. The time varying data is processed to extract pixel by pixel signal spectrum or another similar signal metric. Frequency at each pixel is displayed and distinguished, such as by recoloring the pixels based on spectral power rather than intensity contrast.