Search

CN-121982411-A - Hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm based on absorption characteristics

CN121982411ACN 121982411 ACN121982411 ACN 121982411ACN-121982411-A

Abstract

The invention relates to the technical field of hyperspectral remote sensing image intelligent processing and hydrocarbon substance identification, in particular to an absorption characteristic-based hyperspectral satellite image hydrocarbon substance intelligent identification algorithm, which is characterized in that hyperspectral satellite image raw data are obtained, an earth surface reflectivity data cube is generated through preprocessing, a wave band near 1100nm is selected as a characteristic identification wave band to avoid spectrum confusion, the wave trough position is dynamically searched in the 1095-1105nm range for each pixel spectrum curve, an absorption baseline is established, the absorption depth and the absorption area are calculated, the absorption characteristics are quantized, the absorption depth, the absorption area and the wave trough screening result are fused to generate a pixel-level absorption index map, clustering analysis is executed based on the absorption index map, and the hydrocarbon substance distribution confidence grading result is output. The algorithm adopts a dynamic band selection and multi-feature fusion mechanism, improves the accuracy and anti-interference performance of weak signal identification, and is suitable for oil-gas exploration and environment monitoring application.

Inventors

  • XU GUANGCHUN
  • WANG YIQUN
  • HE XIAONING

Assignees

  • 西安中科西光航天科技集团有限公司

Dates

Publication Date
20260505
Application Date
20260129

Claims (10)

  1. 1. The hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm based on the absorption characteristics is characterized by comprising the following steps of: step 1, acquiring original data based on hyperspectral satellite images, and preprocessing the original data to obtain a surface reflectivity data cube; step 2, selecting a wave band near 1100nm as a hydrocarbon substance characteristic identification wave band based on the earth surface reflectivity data cube; step 3, searching the trough position in the 1095-1105nm wave band range aiming at the spectral curve of each pixel in the earth surface reflectivity data cube, determining the left shoulder reflectivity value and the right shoulder reflectivity value, and establishing an absorption baseline; Step 4, calculating the absorption depth and the absorption area based on the trough position and the absorption baseline, wherein the absorption depth is the difference value of the baseline reflectivity and the trough reflectivity, and the absorption area is the integral value of the trough curve and the baseline surrounding area; Step 5, fusing the absorption depth, the absorption area and the trough screening result, and generating a pixel-level absorption index map by counting the number of the trough and applying threshold filtering; And 6, performing cluster analysis based on the absorption index map, identifying hydrocarbon substance distribution, and outputting a confidence grading result.
  2. 2. The absorption feature-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm according to claim 1, wherein the preprocessing in step 1 comprises: Performing radiometric scaling conversion based on the sensor quantized values in the raw hyperspectral data, converting the digitized values into radiance values; Correcting by adopting an atmospheric radiation transmission model based on the radiation brightness value, and outputting surface reflectivity data; Quality control and noise filtering are performed based on the surface reflectivity data, and a spectral smoothing algorithm is adopted to eliminate random noise, so that a standardized data cube is formed.
  3. 3. The absorption feature-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm according to claim 2, wherein the selection of the feature recognition band in step 2 includes: Constructing a spectrum database comprising hydrocarbon substances and interference ground objects, and analyzing spectrum characteristic differences within the range of 900-2500 nm; performing spectral derivative analysis based on a spectral database, and determining the stability of absorption characteristics of a band around 1100 nm; Based on the spectral separability analysis result, 1095-1105nm is set as a feature extraction window.
  4. 4. The absorption feature-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm according to claim 3, wherein the dynamically searching for the trough position and establishing the absorption baseline in step 3 comprises: identifying a local reflectivity minimum point of each pixel as a trough position by adopting a self-adaptive sliding window algorithm in a 1095-1105nm spectrum interval; searching in a short wave direction based on the trough position, detecting the rising trend of reflectivity of three continuous wave bands, and determining a left shoulder point when the emissivity value is at a local maximum point; searching in the long wave direction based on the trough position, determining a right shoulder point through iterative comparison, and establishing a linear absorption baseline based on the left shoulder point and the right shoulder point.
  5. 5. The absorption feature-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm according to claim 4, wherein the calculating the absorption depth and the absorption area in step 4 includes: Calculating an absorption depth based on the absorption baseline and the trough position, wherein the absorption depth is the difference value between the baseline reflectivity and the trough reflectivity; and calculating an absorption area based on the absorption base line and the trough position, wherein the absorption area is an integral value of the trough curve and the base line surrounding area, and performing standardization processing on the absorption depth and the absorption area.
  6. 6. The absorption feature-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm according to claim 5, wherein the fusion processing in step 5 comprises: setting a minimum trough number threshold based on trough occurrence frequency statistical results in a characteristic wave band range; marking low quality pixels based on the comparison of the trough number to a threshold; Performing weighted fusion based on the standardized absorption depth and absorption area, and determining a weight coefficient according to analysis of the training sample; and generating a binary quality mask based on the trough screening result, performing dot product operation on the mask and the weighted fusion result, and outputting an absorption index map.
  7. 7. The absorption feature-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm according to claim 6, wherein the clustering analysis and classification output in step 6 comprises: Performing unsupervised classification by adopting a K-means clustering algorithm based on the absorption index map, and determining the clustering quantity according to the contour coefficient; Mapping the classification result to a spatial domain based on the clustering result, and analyzing the spatial distribution and spectral characteristics of the clustering clusters; dividing pixels into a high confidence region, a medium confidence region, and a low confidence region based on absorption index value size and spatial continuity; Based on the medium confidence regions, a spatial context analysis is performed, checking the adjacency of the medium confidence regions to the high confidence regions.
  8. 8. The absorption feature-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm according to claim 7, wherein the collaborative processing of dynamic trough finding, absorption feature calculation and feature fusion comprises: the dynamic trough searches the position of the output trough and the absorption baseline, and is used as the input for calculating the absorption depth and the absorption area; the absorption area calculation is based on absorption baseline supplementary absorption depth calculation, and complete spectrum characteristic description is provided; the trough screening mechanism filters low quality signals based on trough statistics, and suppresses false positive recognition in combination with the feature fusion process.
  9. 9. The absorption feature-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm according to claim 8, further comprising: Performing spatial superposition analysis based on the confidence grading result and the field exploration data, and verifying identification accuracy; performing time sequence analysis based on the recognition results of different periods, and monitoring the change trend of the hydrocarbon abnormal region; A detection report is generated based on the analysis results, the report including a spatial distribution map, a confidence rating, a change detection, and an uncertainty assessment.
  10. 10. The absorption feature-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm according to claim 9, wherein the generating a detection report based on the analysis result, the report including a spatial distribution map, a confidence level, a change detection, and an uncertainty assessment, comprises: Based on the spatial distribution map, the confidence level grading, the change detection and the uncertainty evaluation result, generating a hydrocarbon substance spatial distribution visualization map; Generating a hydrocarbon abnormal region statistical table based on the confidence grading result, wherein the table comprises area occupation ratios and space coordinate information of different confidence grades; generating a time sequence change chart based on the change detection result, and displaying the intensity change trend of the hydrocarbon abnormal region in different periods; generating a quality evaluation index based on the uncertainty evaluation result, wherein the index comprises reliability and error range description of the identification result; and integrating the visual map, the statistical table, the time sequence change chart and the quality evaluation index into a comprehensive detection report document.

Description

Hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm based on absorption characteristics Technical Field The invention relates to the technical field of hyperspectral remote sensing image intelligent processing and hydrocarbon substance identification, in particular to an absorption characteristic-based hyperspectral satellite image hydrocarbon substance intelligent identification algorithm. Background The hyperspectral satellite image captures electromagnetic wave information reflected by an earth surface object through a continuous narrow-band sensor, and a data cube comprising hundreds of spectral bands is constructed, wherein hydrocarbon substances generate unique absorption characteristics at specific wavelengths based on molecular chemical bond vibration and electronic transition of the hydrocarbon substances, and the characteristic is represented as a concave spectral curve. Based on the principle, the intelligent recognition method utilizes a characteristic spectrum pattern diagram of a specific object to automatically extract and classify hydrocarbon abnormal signals in the image, so as to recognize the surface hydrocarbon micro-leakage and the oil-bearing sandstone. The technology can be used for oil and gas exploration and pollution monitoring, and the underground resource distribution can be deduced by combining stratum and construction information. The existing hyperspectral remote sensing oil gas detection technology based on 1730nm wave band has the technical pain point, and specifically, (1) the spectrum specificity is insufficient, the existing method relies on 1730nm wave band as hydrocarbon diagnosis characteristics, but the wave band is highly overlapped with absorption characteristics of artificial ground objects such as asphalt ways and parking lots, so that spectrum signal confusion and characteristic extraction uncertainty are increased. It is difficult to characterize the unique absorption pattern of hydrocarbon in complex scenes with 1730nm alone. (2) Weak signal detection capability is limited, namely weak oil gas leakage signal strength is low, and the weak oil gas leakage signal is easy to be absorbed and covered by a man-made structure, so that effective information is covered by strong absorption characteristics, algorithm distinction is insufficient, and omission ratio is increased. (3) The false alarm risk is remarkable, for example, in an environment monitoring case similar to a Rake-Mei Gangdi g crude oil leakage event, the 1730nm wave band extraction result shows that abnormal points are distributed in a factory area instead of a leakage core area, and the single wave band is easy to generate false positive, so that the reliability of pollution evolution tracking is weakened. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an absorption characteristic-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm, which solves the technical problem of low recognition accuracy of weak oil gas leakage information because the existing 1730nm wave band is easily interfered by artificial ground objects in hyperspectral remote sensing oil gas detection. In order to solve the technical problems, the invention comprises the following specific contents: the invention provides an absorption characteristic-based hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm, which comprises the following steps: step 1, acquiring original data based on hyperspectral satellite images, and preprocessing the original data to obtain a surface reflectivity data cube; step 2, selecting a wave band near 1100nm as a hydrocarbon substance characteristic identification wave band based on the earth surface reflectivity data cube; step 3, searching the trough position in the 1095-1105nm wave band range aiming at the spectral curve of each pixel in the earth surface reflectivity data cube, determining the left shoulder reflectivity value and the right shoulder reflectivity value, and establishing an absorption baseline; Step 4, calculating the absorption depth and the absorption area based on the trough position and the absorption baseline, wherein the absorption depth is the difference value of the baseline reflectivity and the trough reflectivity, and the absorption area is the integral value of the trough curve and the baseline surrounding area; Step 5, fusing the absorption depth, the absorption area and the trough screening result, and generating a pixel-level absorption index map by counting the number of the trough and applying threshold filtering; And 6, performing cluster analysis based on the absorption index map, identifying hydrocarbon substance distribution, and outputting a confidence grading result. Further, according to the hyperspectral satellite image hydrocarbon substance intelligent recognition algorithm based on the absorption characteristics, the prepro