Method for detecting shipwrecked vessel and drown victims by using aerial hyperspectral image

Patent metadata

Field	Value
Publication number	US-10656018-B2
Application number	US-201816465079-A
Country	US
Kind code	B2
Filing date	May 18, 2018
Priority date	Oct 24, 2017
Publication date	May 19, 2020
Grant date	May 19, 2020

Abstract

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.

First claim

The invention claimed is:

1. A detection method for disabled ships and persons overboard based on airborne hyperspectral images comprising,
a) receiving an observed airborne hyperspectral image and confirming whether the airborne hyperspectral image is suitable for detection analysis, and extracting observation information, location information and reflectivity value of each pixel of a detected disabled ship and overboard persons, via an image receiving unit;
(b) analyzing a degree of similarity of spectral characteristics between spectral reflectivity value of a target material and an observed reflectivity value using pre-constructed spectral library information, and extracting a constituent material and occupation ratio for each pixel of the hyperspectral image to classify detected results via an image analyzing unit; and
(c) displaying the received hyperspectral image, location of the detected disabled ship and persons overboard, and detailed information of the detected results, via an image visualization unit,
wherein said step (b) comprises,
measuring a degree of similarity between reference spectral reflection curves of target materials included in the spectral library and a reflection curve obtained from the hyperspectral image, and
in a case where the obtained reflection curve is measured in a plurality of bands,
the plurality of bands are assumed as vectors, and the degree of similarity is measured through a spectral angle measurement algorithm that defines an angle between each respective vector of the reference spectral reflection curves and a vector of the obtained reflection curve as a spectral angle.

2. The method of claim 1 , wherein said step (a) comprises,
receiving the observed airborne hyperspectral image via an image receiving module;
receiving the received hyperspectral image and confirming whether the observation information and image observed value is suitable for detection analysis by via an image conformance check module; and
extracting the observation information, the location information and the reflectivity value of each pixel required for image analysis and visualization of detected results via an image preprocessing module.

3. The method of claim 2 , wherein the observation information required for image analysis and visualization of detected results is spatial resolution and observation date and time.

4. The method of claim 1 , wherein said step (b) comprises,
fetching the spectral library information pre-constructed through field observations via a spectral library module;
analyzing the degree of similarity of the spectral
characteristics between the spectral reflectivity value of the target material and the observed reflectivity value using the spectral library information to extract the constituent material
and occupation ratio for each pixel of the hyperspectral image via a spectral mixture analysis module; and
classifying the detected results based on the constituent materials and occupation ratio extracted for each pixel and storing the detected results together with detailed information via a detected result classifying module.

5. The method of claim 1 , wherein in said step (b)
the image analyzing unit measures the degree of similarity to calculate the type and content of the constituent material for each pixel.

6. The method of claim 5 , wherein the degree of similarity is a degree of similarity between a subject spectrum and a reference spectrum calculated through the spectral angle measurement algorithm.

7. The method of claim 1 , wherein said step (c) comprises,
displaying and storing the received hyperspectral image in form of an image by selecting a predetermined band via a hyperspectral image visualization module;
displaying and storing a location of a detected body in form of an image by distinguishing the detected disabled ship and persons overboard via a disabled ship and persons overboard location visualization module; and
displaying and storing the detailed information including type, material, latitude and longitude by distinguishing according to classification of the detected results via a detected result detailed information displaying module.

8. The method of claim 1 , wherein the spectral library is constructed by measuring spectral characteristic curves of constituent materials of a ship and a drifting body that are classified by material, paint, and color via a spectrometer.

9. A detection method for disabled ships and persons overboard based on airborne hyperspectral images comprising,
(a) receiving an observed airborne hyperspectral image and confirming whether the airborne hyperspectral image is suitable for detection analysis, and extracting observation information, location information and reflectivity value of each pixel of a detected disabled ship and overboard persons, via an image receiving unit;
(b) analyzing a degree of similarity of spectral characteristics between spectral reflectivity value of a target material and an observed reflectivity value using pre-constructed spectral library information, and extracting a constituent material and occupation ratio for each pixel of the hyperspectral image to classify detected results via an image analyzing unit; and
(c) displaying the received hyperspectral image, location of the detected disabled ship and persons overboard, and detailed information of the detected results, via an image visualization unit;
wherein said step (b) comprises,
measuring a degree of similarity between reference spectral reflection curves of target materials included in the spectral library and a reflection curve obtained from the hyperspectral
image, and
in a case where the obtained reflection curve is measured in a plurality of bands, the plurality of bands are assumed as vectors, and the degree of similarity is measured through a spectral angle measurement algorithm that defines an angle between each respective vector of the reference spectral reflection curves and a vector of the obtained reflection curve as a spectral angle, and if the extracted constituent material within each pixel is multi-mixed, a spectral characteristic curve that is completed based on the hyperspectral image is used to extract the constituent material and occupation ratio of each pixel which enables detection in pixel units or less.

Assignees

• Korea Inst Ocean Sci & Tech

Inventors

• Oh Sangwoo
• Kim Tae-Sung
• Kang Won-Soo
• Lee Moonjin
• Chun Tae-Byung

Classifications

• G06T2207/10036

  Multispectral image; Hyperspectral image · CPC title

• G06T7/11

  Region-based segmentation · CPC title

• G06T17/05 Primary

  Geographic models · CPC title

• G06T2207/10032

  Satellite or aerial image; Remote sensing · CPC title

• G01N21/55

  Specular reflectivity · CPC title