Hyperspectral Detection of Shipwrecks and Drown Victims
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image analysis methods using multi-spectral images struggle to accurately and quickly detect disabled ships and persons overboard in marine accidents due to similarities in reflectivity and morphological features with the surrounding environment, leading to high missed detection rates and potential fatal casualties.
Innovation Solution
A method utilizing airborne hyperspectral images that analyzes spectral characteristics through a pre-constructed spectral library to extract constituent materials and occupation ratios, enabling accurate classification and visualization of detected targets, thereby improving detection accuracy and reducing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional multi-spectral image analysis methods are used for marine search, then the detection system is simple and easy to operate, but the detection accuracy is low due to similar reflectivity and morphological features between targets and background
Solution Approach 1:
The patent transitions from multi-spectral imaging (3-10 bands) to hyperspectral imaging (tens to hundreds of continuous bands), adding spectral dimensionality. This enables differentiation of targets from background by analyzing continuous spectral reflectance characteristics across many narrow bands, resolving the contradiction between detection accuracy and system complexity.
Solution Approach 2:
The patent changes the spectral resolution parameter from coarse (100-200 nm bandwidth per band) to fine (10 nm bandwidth per band) by using hyperspectral sensors. This parameter change allows detection of subtle spectral differences between targets and seawater, significantly improving detection accuracy while accepting increased system complexity.
2Area of stationary object
If aerial wide area observation is performed to cover large search areas, then the coverage area increases, but the spatial resolution and spectral sensitivity per pixel decrease leading to missed detections
Solution Approach 1:
The patent compensates for reduced per-pixel spectral sensitivity in wide-area aerial observation by utilizing the high spectral dimensionality (tens to hundreds of bands) of hyperspectral imaging. The increased spectral information per pixel allows detection algorithms to identify targets even when individual spectral bands have lower signal-to-noise ratios.
Solution Approach 2:
The patent combines information across multiple spectral bands through spectral mixture analysis and target detection algorithms. By merging spectral information from tens to hundreds of bands, the system achieves high detection sensitivity in wide-area observations where individual pixels have limited spectral signal.
3Reliability
If conventional multi-spectral imaging is used, then the equipment cost and data processing load are lower, but the missed detection rate is very high due to inability to distinguish target spectral characteristics
Solution Approach 1:
The patent uses hyperspectral imaging to capture tens to hundreds of continuous spectral bands, preserving complete spectral characteristic curves for each pixel. This high-dimensional spectral information prevents information loss and enables reliable detection by providing sufficient spectral features for distinguishing targets from background.
Solution Approach 2:
The patent performs spectral library construction and target spectral characteristic analysis in advance before actual search operations. This preliminary action prepares detection templates and spectral signatures, enabling rapid and reliable target identification during real-time search while minimizing information loss.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method significantly reduces detection time and increases the probability of detecting disabled ships and persons overboard, achieving detection rates greater than 90% with low false alarms, supporting rapid search operations and enhancing marine search operations.
Implementation Method 1
the reflectivity or morphological feature differences between a target material (ship body, overboard persons) and a background material (seawater) is used for target detection
Data Source
AI summary
A method for detecting a shipwrecked vessel and drown victims by using an aerial hyperspectral image, according to the present invention, comprises the steps of: (a) allowing an image reception unit to receive an observed aerial hyperspectral image, check whether the received aerial hyperspectral image is suitable for detection analysis, and extract observation information, and location information and a reflectance value for each pixel with respect to the detected shipwrecked vessel and drown victims; (b) allowing an image analysis unit to analyze spectral characteristic similarity between a spectral reflection value of a target object and an observed reflection value by using pre-constructed spectral library information and extract constituent materials and an occupation ratio for each pixel of the hyperspectral image, thereby classifying a detection result; and (c) allowing an image visualization unit to display the received hyperspectral image, locations of the detected shipwrecked vessel and drown victims, and detailed information of the detection result.