CN-121999357-A - Method for identifying sea surface spilled oil

CN121999357ACN 121999357 ACN121999357 ACN 121999357ACN-121999357-A

Abstract

The invention discloses a method for identifying sea surface spilled oil, which relates to the technical field of image identification, and comprises the following steps of S1, collecting and obtaining an original spilled oil image with a sea surface oil film and an SAR spilled oil image, carrying out denoising and enhancing pretreatment on the original spilled oil image to obtain a pretreated spilled oil image, S2, extracting and obtaining color features from the pretreated spilled oil image, extracting and obtaining texture features from the SAR spilled oil image, carrying out multi-source information fusion on the color features and the texture features to obtain a spilled oil image after feature fusion, and S4, adopting a sea surface spilled oil identification network to identify an image to be detected to obtain an identification result, wherein the sea surface spilled oil identification network comprises a C3 module, an SPPF module, a Upsanple module, a Concat module, a C2f module, a SimAM attention mechanism and the like, and sea surface spilled oil identification is more accurate through the feature fused spilled oil image, the sea surface spilled oil identification network and the like.

Inventors

WU XIAODONG
ZHU KEHAI
LI QIAN
FAN LU
WANG YANYAN
JIANG CHUNJUAN
ZHANG JING
MENG ZHAOKAI
ZHANG DAPENG

Assignees

中国石油化工股份有限公司
中国石油化工股份有限公司胜利油田分公司

Dates

Publication Date: 20260508
Application Date: 20241101

Claims (10)

1. A method for identifying sea surface spilled oil is characterized by comprising the following steps: Step S1, acquiring an original oil spill image with a sea surface oil film and an SAR oil spill image, and carrying out denoising and enhancing pretreatment on the original oil spill image to obtain a pretreated oil spill image; S2, extracting color features from the preprocessed oil spill image, extracting texture features from the SAR oil spill image, and carrying out multi-source information fusion on the color features and the texture features to obtain a feature fused oil spill image; and S4, identifying the image to be detected by adopting a sea surface oil spill identification network to obtain an identification result.
2. A method of identifying a sea surface spill as claimed in claim 1, further comprising a step S3 following step S2, S3, segmenting the oil spill image after feature fusion and removing an invalid target without oil spill to obtain an oil spill data set, wherein the oil spill data set comprises a plurality of segmented oil spill images; In the step S4, the image to be detected is all the images in the spilled oil data set obtained in the step S3, the sea surface spilled oil identification network is trained, a trained sea surface spilled oil identification network is obtained, and the trained sea surface spilled oil identification network is used for identifying sea surface spilled oil in the images.
3. The method for identifying sea surface spilled oil according to claim 1, wherein in the step S4, the sea surface spilled oil identification network is a sea surface spilled oil identification network trained in advance, the image to be detected is a spilled oil image obtained in the step S2 and obtained after feature fusion, and the step S4' is formed.
4. The method of claim 1, wherein in the step S4, the sea surface spilled oil identification network comprises a first convolution layer, a second convolution layer, a first C3 module, a third convolution layer, a second C3 module, a fourth convolution layer, a third C3 module, a fifth convolution layer, a fourth C3 module, an SPPF module, a first Upsanple module, a first Concat module, a first C2f module, a second Upsanple module, a second Concat module, a second C2f module, a first SimAM attention mechanism, a sixth convolution layer, a third Concat module, a third C2f module, a second SimAM attention mechanism, a seventh convolution layer, a fourth Concat module, a fourth C2f module, and a third SimAM attention mechanism, and a first, a second, and a third detection module, the first and the second detection module being connected to the first SimAM attention mechanism, the second detection module being connected to the third Upsanple attention mechanism, and the third detection module being connected to the third detection module 28.
5.A method for identifying a sea surface spill according to claim 1, wherein in step S1, the original spill image and the SAR spill image of the sea surface area are obtained by a satellite, an unmanned aerial vehicle or an onboard sensor.
6. A method of identifying sea surface spilled oil as defined in claim 1 wherein, in step S1, the step of denoising includes removing random noise from the image using non-local mean filtering and the step of enhancing includes processing the image by gamma correction.
7. The method for identifying sea surface spillover according to claim 1, wherein in the step S2, the step of extracting and obtaining color features includes extracting color features in the preprocessed spillover image by using color histograms, converting the image from RGB color space to HSV color space, calculating a histogram of each channel of the HSV color space, normalizing the histogram to obtain a normalized color histogram, the step of extracting and obtaining texture features includes extracting texture features from the SAR spillover image by using local binary patterns, calculating and obtaining local binary patterns of each pixel to obtain LBP histograms for describing texture features of pixel neighborhood, and the step of multi-source information fusion includes performing weighted average fusion on the normalized color histograms and the LBP histograms.
8. A method for identifying sea surface spilled oil according to claim 2, wherein in step S3 the step of segmenting comprises random rotation, flipping, cropping and color adjustment.
9. A method for identifying spilled oil on the sea according to claim 2, wherein the step S3 is performed by manual operation to remove the ineffective target of no spilled oil.
10. The method of claim 2, wherein in step S3, the segmented spilled oil image is a 160X 160 pixel image.

Description

Method for identifying sea surface spilled oil Technical Field The invention relates to the technical field of image recognition, in particular to a method for recognizing sea surface spilled oil. Background The marine oil spill event causes serious damage to the marine environment and ecosystem. Therefore, the rapid and accurate identification and monitoring of sea surface spilled oil is an important task for marine environmental protection. Currently, data from a single sensor cannot provide enough information to accurately identify oil spills. Aiming at the technical problems that the complex background recognition rate is low, the robustness is insufficient due to uneven distribution of the thin and thick oil spilled on the sea surface, the recognition effect of the oil spilled on the sea surface is poor, and the like. Disclosure of Invention The invention provides a method for identifying sea surface spilled oil, which solves the technical problem of poor sea surface spilled oil identification effect. In order to solve the technical problems, the technical scheme adopted by the invention is as follows: A method of identifying sea surface spilled oil comprising the steps of: Step S1, acquiring an original oil spill image with a sea surface oil film and an SAR oil spill image, and carrying out denoising and enhancing pretreatment on the original oil spill image to obtain a pretreated oil spill image; S2, extracting color features from the preprocessed oil spill image, extracting texture features from the SAR oil spill image, and carrying out multi-source information fusion on the color features and the texture features to obtain a feature fused oil spill image; and S4, identifying the image to be detected by adopting a sea surface oil spill identification network to obtain an identification result. The further technical proposal is that the method also comprises a step S3 positioned after the step S2, S3, segmenting the oil spill image after feature fusion and removing an invalid target without oil spill to obtain an oil spill data set, wherein the oil spill data set comprises a plurality of segmented oil spill images; In the step S4, the image to be detected is all the images in the spilled oil data set obtained in the step S3, the sea surface spilled oil identification network is trained, a trained sea surface spilled oil identification network is obtained, and the trained sea surface spilled oil identification network is used for identifying sea surface spilled oil in the images. In the step S4, the sea surface spilled oil identification network is a sea surface spilled oil identification network trained in advance, the image to be detected is a spilled oil image obtained in the step S2 and subjected to feature fusion, and a step S4' is formed. In a further technical scheme, in the step S4, the sea surface spilled oil identification network includes a first convolution layer, a second convolution layer, a first C3 module, a third convolution layer, a second C3 module, a fourth convolution layer, a third C3 module, a fifth convolution layer, a fourth C3 module, an SPPF module, a first Upsanple module, a first Concat module, a first C2f module, a second Upsanple module, a second Concat module, a second C2f module, a first SimAM attention mechanism, a sixth convolution layer, a third Concat module, a third C2f module, a second SimAM attention mechanism, a seventh convolution layer, a fourth Concat module, a fourth C2f module, and a third SimAM attention mechanism, and a first Detect module, a second Detect module, and a third Detect module that are sequentially connected to the first SimAM attention mechanism, and the second Detect module to the second SimAM attention mechanism, and the third Detect module to the third SimAM attention mechanism. In the step S1, an original oil spill image and an SAR oil spill image of the sea surface area are obtained through a satellite, an unmanned aerial vehicle or a ship-borne sensor. The further technical scheme is that in the step S1, the denoising step comprises the step of eliminating random noise in the image by using non-local mean filtering, and the step of enhancing comprises the step of processing the image by gamma correction. The technical scheme is that in the step S2, the step of extracting and obtaining color features comprises the steps of extracting color features in a preprocessed oil spill image by using a color histogram, converting the image from an RGB color space to an HSV color space, calculating a histogram of each channel of the HSV color space, normalizing the histogram to obtain a normalized color histogram, the step of extracting and obtaining texture features comprises the steps of extracting texture features from an SAR oil spill image by using a local binary pattern, calculating and obtaining a local binary pattern of each pixel to obtain an LBP histogram for describing texture features of a pixel neighborhood, and the step of multi-sou