CN114049347A - Crude oil leakage detection method based on feature enhancement - Google Patents

Crude oil leakage detection method based on feature enhancement Download PDF

Info

Publication number
CN114049347A
CN114049347A CN202111423141.0A CN202111423141A CN114049347A CN 114049347 A CN114049347 A CN 114049347A CN 202111423141 A CN202111423141 A CN 202111423141A CN 114049347 A CN114049347 A CN 114049347A
Authority
CN
China
Prior art keywords
feature
target
crude oil
shadow
oil leakage
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111423141.0A
Other languages
Chinese (zh)
Inventor
宫法明
高亚婷
李云静
嵇晓峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China University of Petroleum East China
Original Assignee
China University of Petroleum East China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China University of Petroleum East China filed Critical China University of Petroleum East China
Priority to CN202111423141.0A priority Critical patent/CN114049347A/en
Publication of CN114049347A publication Critical patent/CN114049347A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/60—Rotation of whole images or parts thereof
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Business, Economics & Management (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Marketing (AREA)
  • Agronomy & Crop Science (AREA)
  • Mining & Mineral Resources (AREA)
  • Probability & Statistics with Applications (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Animal Husbandry (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a crude oil leakage detection method based on feature enhancement, which comprises the following steps: solving the problem of small samples by using a data enhancement method; denoising an input image; in the shadow removing stage, a Mask-shadow GAN-based shadow detection and removal network is used, an unpaired image data training model is adopted, and shadow Mask is used for guiding generation of shadow images, so that shadow removal in a complex scene is realized; an attention mechanism is introduced into a feature pyramid, and the convolution span between a shallow feature map and a high feature map is reduced by reconstructing the pyramid from top to bottom so as to enrich fine-grained features contained in the high feature map; a PANet structure is added behind the characteristic pyramid network, so that shallow layer characteristic information is better reserved; and providing powerful judgment basis for the detector by utilizing the target peripheral context information, taking the peripheral characteristics of the predicted frame as the context information, and providing supplementary information for the detection network to adjust the result of the detection network so as to realize the detection of the crude oil leakage. The method can effectively solve the problem of crude oil leakage detection in the complex scene of the petroleum production area.

Description

Crude oil leakage detection method based on feature enhancement
Technical Field
The invention belongs to the field of computer vision, and relates to a crude oil leakage detection method based on feature enhancement.
Background
In recent years, the development of computer vision technology is rapidly advanced, and a deep learning algorithm is rapidly developed in the field of computer vision and is gradually and widely applied to the field of video monitoring. Different from the traditional method for manually making the features, the deep learning algorithm can automatically learn the features of interest in the data, so that various tasks such as classification, identification, detection and the like are realized.
There is little current research on the detection of crude oil leaks in oilfield production areas. The problem of crude oil leakage not only causes certain environmental pollution, but also brings certain potential safety hazard to the working area of the oil field. In the detection and management process of the oil field production area, the labor investment is very large, and the cost is high; the safety of manual monitoring is not high, and economic loss and even casualties are easily caused. The remote video monitoring system is introduced to the crude oil leakage detection of the oil field production area, so that the intelligent management of the oil field is facilitated, the production and operation efficiency of the oil field is further improved, and the safety problem of staff is also facilitated to be guaranteed. The crude oil leakage detection in the oil field production area faces a plurality of problems, such as difficulty in obtaining crude oil leakage samples in the oil field production area, strong interference of illumination shadow and the like in the oil field production area, complex scene of video monitoring images, fuzzy target characteristics and the like. The interference causes the crude oil leakage detection effect to be reduced sharply, so the crude oil leakage detection in the field of oil field production is a difficult problem to be solved urgently.
Disclosure of Invention
In order to overcome the defects, the invention provides a crude oil leakage detection method based on feature enhancement, which comprises the following specific steps:
s1, amplifying the data set by using a data enhancement method, and performing operations such as cutting, turning, zooming, color gamut enhancement and the like on the image according to the characteristics of the crude oil leakage target to be detected;
s2, carrying out image preprocessing on the data set after data enhancement to remove noise in the image;
s3, detecting and removing shadow areas in the image by using a Mask-ShadowGAN-based shadow detection network;
s4, extracting the deep level features of the detection target by using a backbone with CSPDarknet53 as a core;
s5, connecting adjacent feature layers in a feature fusion mode based on an attention mechanism, and constructing an enhanced feature pyramid with rich target information and remarkable features from bottom to top;
s6, introducing a PANet structure on the effective feature layers for feature extraction;
s7, using the context information around the target to provide a powerful criterion for the detector
And S8, outputting the detection result of the crude oil leakage target.
The technical scheme of the invention is characterized by comprising the following steps:
for step S1, the present invention mainly uses random scaling, flipping, clipping, rotating, etc. to amplify the data, and according to the characteristics of the crude oil leakage target, data enhancement is performed on the detection target from the aspect of geometric, gray scale, texture, etc. features, and a Mosaic data enhancement method is used to enrich the background of the image, and further amplify the data amount of the small target.
For step S3, the invention uses a shadow detection network based on Mask-ShadowGAN to detect and remove the shadow area in the image, because a large number of paired images with and without shadow based on the oil field scene cannot be collected, the Mask-ShadowGAN network uses unpaired image data training model, makes a difference through the shadow image (input) and the image without shadow (output result of the network), uses Otsu automatic threshold to obtain a binary image to obtain a shadow Mask, and guides the generation of the shadow image through the shadow Mask to realize the shadow removal in the complex scene.
For step S5, the present invention proposes to introduce an attention mechanism into the feature pyramid, and to obtain richer crude oil leakage features, the convolution span between the shallow feature map and the high feature map is reduced by reconstructing the pyramid from top to bottom, so as to enrich fine-grained features contained in the high feature map. And (3) connecting adjacent feature layers by using a feature fusion mode based on a spatial attention mechanism so as to achieve the purpose of generating the salient features and realize the top-down connection of the feature pyramid. The spatial attention module used by the invention consists of a maximum pooling layer and an average pooling layer along the channel direction, a 3 multiplied by 3 convolution layer and a Sigmoid function, an effective information area is highlighted through pooling operation, and a spatial attention diagram is output to extract the characteristics of the region of interest. The feature fusion process can be expressed as:
Figure BDA0003377244950000021
Figure BDA0003377244950000022
in the formula: att (-) is the attention mechanism; σ (-) is a Sigmoid function; f. of3×3Is a 3 × 3 convolution; f. of1×1Is a 1 × 1 convolution; avgpool, MaxPool, g2x-upAverage pooling, maximum pooling and 2-fold down-sampling, respectively;
Figure BDA0003377244950000023
is a concatenate feature superposition mode.
For step S6, the method introduces a PANET structure into a feature pyramid, the shallow features are mostly features such as edge shapes of images, the features are key features to be extracted from the crude oil leakage image, a Bottom-up Path Augmentation is added behind the feature pyramid network, the shallow features are transmitted to the top layer along the PANET structure through FPN, and the shallow feature information can be well retained.
For step S7, in order to improve the accuracy of the determination, the present invention provides a strong determination basis for the detector by using the target peripheral context information, and uses the predicted frame peripheral features as context information to provide supplementary information for the detection network. According to the frame coordinate (b)x,by,bw,bh) Extracting context feature of the periphery of the target on the corresponding prediction feature map, and extracting context coordinate region (b)x,by,βbw,βbh) And beta is the magnification factor. The context characteristics are firstly integrated by two convolution layers of 3 multiplied by 3, then reduced in dimension by an average pooling layer, and finally output a target fraction adjusting parameter lambda by a Sigmoid functionobjAnd a class confidence adjustment parameter λkAnd adjusting PobjAnd Pk：
Figure BDA0003377244950000031
In the formula: σ (-) is a Sigmoid function, Pr (object) and Pr (class)k) The target score and the category confidence value which are directly output by the convolutional layer and are not normalized by a Sigmoid function; pobjRepresenting the probability of the frame containing the target as the target score; pkThe category confidence represents the probability that the target in the frame belongs to the kth category. The invention aims at the detection of a crude oil leakage target, and only one target classification label 'oil spill' is set for privacy. The classifier only needs to calculate the probability P that the target belongs to the background0And probability P of being a crude oil leak1。
The crude oil leakage detection method based on feature enhancement solves the problem of crude oil leakage detection in the complex scene of an oil field production area, and has the following advantages:
(1) the method adopts a mosaic data enhancement mode, and greatly enriches the background of the oil leakage target.
(2) The method adopts an unpaired data training model, can solve the problem of interference of illumination shadow and the like on oil leakage target detection, and realizes crude oil leakage detection in a complex scene;
(3) the method introduces a deep level feature extraction network, starts with feature enhancement from two aspects of enriching feature map information and improving significance features, so as to adapt to low-resolution and small-target crude oil leakage detection, thereby improving the accuracy of the detection network.
Drawings
FIG. 1 is a flow chart of a feature enhancement based crude oil leak detection method of the present invention.
FIG. 2 is a flowchart of a Mask-ShadowGAN-based shadow detection and removal method according to the present invention.
FIG. 3 is a flow chart of crude oil leakage detection based on deep level feature extraction network in the present invention.
Detailed Description
The invention is described in further detail below with reference to the following figures and detailed description:
a method for detecting crude oil leakage based on feature enhancement, as shown in fig. 1, is a flow chart of the method for detecting crude oil leakage based on feature enhancement of the present invention, and the method comprises:
and S1, data enhancement, wherein according to the characteristics and environment of the crude oil leakage site of the oil field, aiming at the problem that the crude oil leakage image data set is difficult to obtain, the crude oil leakage site is simulated in the oil field scene to acquire and enhance samples required by the experiment. According to the characteristics, on the basis of traditional data enhancement modes such as random scaling, random cutting, shifting, horizontal/vertical overturning and the like, data expansion is carried out by using a Mosaic mode and the like, four training images are combined into one image according to a certain proportion, so that the data volume of small objects can be increased, and the positions of detected objects in the images are changed. So as to increase the change of the oil leakage target dimension, position and background and improve the generalization of the model.
S2, preprocessing the image, wherein the image in reality can be influenced by a plurality of factors so as to contain certain noise, and Gaussian bilateral filtering is adopted to remove the image noise while keeping the edge information of the detected target, thereby laying a foundation for further target detection.
S3, detecting and removing shadow areas in the image by using a Mask-ShadowGAN-based shadow detection network, and because a large number of paired shadow-based and shadow-free images based on the oil field scene cannot be collected, the Mask-ShadowGAN network adopts an unpaired image data training model, performs subtraction through a shadow-based image (input) and a shadow-free image (output result of the network), obtains a binary image by using an Otsu automatic threshold value to obtain a shadow Mask, and guides generation of the shadow image through the shadow Mask to realize shadow removal in a complex scene.
And S4, extracting the deep-level features of the detection target by using a backbone with CSPDarknet53 as a core, inputting the image subjected to denoising and shadow removal into a backbone feature network with CSPDarknet53 as a core, adjusting the size of the input image to be 416 x 416, inputting the adjusted input image into a CSPDarknet53 network, and performing feature extraction by using a Mish activation function.
And S5, connecting adjacent feature layers in a feature fusion mode based on an attention mechanism, and constructing an enhanced feature pyramid with rich target information and remarkable features from bottom to top. The attention module consists of a maximum pooling layer and an average pooling layer along the channel direction, a 3 multiplied by 3 convolution layer and a Sigmoid function, and the effective information area is highlighted through a pooling operation, and the attention module outputs a spatial attention diagram to extract the characteristics of the region of interest. The feature fusion process can be expressed as:
Figure BDA0003377244950000041
Figure BDA0003377244950000042
in the formula: att (-) is the attention mechanism; σ (-) is a Sigmoid function; f. of3×3Is a 3 × 3 convolution; f. of1×1Is a 1 × 1 convolution; avgpool, MaxPool, g2x-upAverage pooling, maximum pooling and 2-fold down-sampling, respectively;
Figure BDA0003377244950000043
is a concatenate feature superposition mode.
S6, introducing a PANET structure to a plurality of effective feature layers for feature extraction, adding a Bottom-up Path evaluation behind the feature pyramid network, transmitting the shallow feature to the top layer along the PANET structure after passing through the FPN, and well retaining the shallow feature information.
S7, inputting the extracted multi-layer feature information into the inspection network, providing powerful judgment basis for the detector by using the target surrounding context information, and according to the frame coordinate (b)x,by,bw,bh) Extracting context feature of the periphery of the target on the corresponding prediction feature map, and extracting context coordinate region (b)x,by,βbw,βbh) And beta is the magnification factor. The context characteristics are firstly integrated by two convolution layers of 3 multiplied by 3, then reduced in dimension by an average pooling layer, and finally output a target fraction adjusting parameter lambda by a Sigmoid functionobjAnd a class confidence adjustment parameter λkAnd adjusting PobjAnd Pk：
Figure BDA0003377244950000051
In the formula: σ (-) is a Sigmoid function, Pr (object) and Pr (class)k) The target score and the category confidence value which are directly output by the convolutional layer and are not normalized by a Sigmoid function; pobjRepresenting the probability of the frame containing the target as the target score; pkThe category confidence represents the probability that the target in the frame belongs to the kth category. The invention aims at the detection of a crude oil leakage target, and only one target score is set for privacyThe class label "oil spill". The classifier only needs to calculate the probability P that the target belongs to the background0And probability P of being a crude oil leak1。
S8, outputting the detection result of the crude oil leakage target, including the coordinates (b) of the prediction framex,by,bw,bh) And P1。
In conclusion, the crude oil leakage detection method based on feature enhancement solves the problem of crude oil leakage detection in a complex scene of an oil field production area under a big data environment, and the proposed deep level feature extraction network starts with feature enhancement from two aspects of enriching feature map information and improving significance features, and can be suitable for detection of low-resolution and small target objects.
While the present invention has been described in detail with reference to the preferred embodiments, it should be understood that the above description should not be taken as limiting the invention. Various modifications and alterations to this invention will become apparent to those skilled in the art upon reading the foregoing description. Accordingly, the scope of the invention should be determined from the following claims.

Claims (6)

1. A crude oil leakage detection method based on feature enhancement is characterized by comprising the following specific steps:
s1, amplifying the data set by using a data enhancement method, and performing operations such as cutting, turning, zooming, color gamut enhancement and the like on the image according to the characteristics of the crude oil leakage target to be detected;
s2, carrying out image preprocessing on the data set after data enhancement to remove noise in the image;
s3, detecting and removing shadow areas in the image by using a Mask-ShadowGAN-based shadow detection network;
s4, extracting the deep level features of the detection target by using a backbone with CSPDarknet53 as a core;
s5, connecting adjacent feature layers in a feature fusion mode based on an attention mechanism, and constructing an enhanced feature pyramid with rich target information and remarkable features from bottom to top;
s6, introducing a PANet structure on the effective feature layers for feature extraction;
s7, using the context information around the target to provide a powerful criterion for the detector
And S8, outputting the detection result of the crude oil leakage target.
2. The method for detecting crude oil leakage based on feature enhancement as claimed in claim 1, wherein for step S1, the method mainly uses random scaling, flipping, clipping, rotating, etc. to amplify the data, and according to the features of the crude oil leakage target, data enhancement is performed on the detected target from the aspect of geometric, gray, texture, etc. features, and a Mosaic data enhancement method is used to enrich the background of the image and further amplify the data size of the small target.
3. The method for detecting crude oil leakage based on feature enhancement as claimed in claim 1, wherein for step S3, the invention employs a Mask-ShadowGAN-based shadow detection network to detect and remove shadow regions in the image, since a large number of pairs of shadow-based and non-shadow images in the oil field scene cannot be collected, the Mask-ShadowGAN network employs an unpaired image data training model, performs subtraction through the shadow-based image (input) and the non-shadow image (output result of the network), obtains a binary image by using Otsu automatic threshold to obtain a shadow Mask, and guides generation of the shadow image by the shadow Mask to realize shadow removal in a complex scene.
4. The feature-enhancement-based crude oil leakage detection method as claimed in claim 1, wherein for step S5, the invention proposes to introduce an attention mechanism into the feature pyramid, and to obtain richer crude oil leakage features, the convolution span between the shallow feature map and the high feature map is reduced by reconstructing the pyramid from top to bottom, so as to enrich fine-grained features contained in the high feature map. And (3) connecting adjacent feature layers by using a feature fusion mode based on a spatial attention mechanism so as to achieve the purpose of generating the salient features and realize the top-down connection of the feature pyramid. The spatial attention module used by the invention consists of a maximum pooling layer and an average pooling layer along the channel direction, a 3 multiplied by 3 convolution layer and a Sigmoid function, an effective information area is highlighted through pooling operation, and a spatial attention diagram is output to extract the characteristics of the region of interest. The feature fusion process can be expressed as:
Figure FDA0003377244940000011
Figure FDA0003377244940000021
Figure FDA0003377244940000022
in the formula: att (-) is the attention mechanism; σ (-) is a Sigmoid function; f. of3×3Is a 3 × 3 convolution; f. of1×1Is a 1 × 1 convolution; avgpool, MaxPool, g2x-upAverage pooling, maximum pooling and 2-fold down-sampling, respectively;
Figure FDA0003377244940000023
is a concatenate feature superposition mode.
5. The feature enhancement-based crude oil leakage detection method according to claim 1, characterized in that, for step S6, the present invention introduces a PANet structure into a feature pyramid, and the shallow features are mostly features such as edge shapes of images, which are key features to be extracted from the crude oil leakage image, and a Bottom-up Path Augmentation is added behind the feature pyramid network, and the shallow features are transmitted to the top layer along the PANet structure through the FPN, so that the shallow feature information can be better retained.
6. The method for detecting crude oil leakage based on feature enhancement as claimed in claim 1, wherein for step S7, in order to improve the accuracy of the determination, the present invention provides powerful criterion for the detector by using the context information around the target, and provides supplementary information for the detection network by using the feature around the predicted frame as the context information. According to the frame coordinate (b)x，by，bw，bh) Extracting context feature of the periphery of the target on the corresponding prediction feature map, and extracting context coordinate region (b)x，by，βbw，βbh) And beta is the magnification factor. The context characteristics are firstly integrated by two convolution layers of 3 multiplied by 3, then reduced in dimension by an average pooling layer, and finally output a target fraction adjusting parameter lambda by a Sigmoid functionobjAnd a class confidence adjustment parameter λkAnd adjusting PobjAnd Pk：
Figure FDA0003377244940000024
In the formula: σ (-) is a Sigmoid function, Pr (object) and Pr (class)k) The target score and the category confidence value which are directly output by the convolutional layer and are not normalized by a Sigmoid function; pobjRepresenting the probability of the frame containing the target as the target score; pkThe category confidence represents the probability that the target in the frame belongs to the kth category. The invention aims at the detection of a crude oil leakage target, and only one target classification label 'oil spill' is set for privacy. The classifier only needs to calculate the probability P that the target belongs to the background0And probability P of being a crude oil leak1。
CN202111423141.0A 2021-11-26 2021-11-26 Crude oil leakage detection method based on feature enhancement Pending CN114049347A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111423141.0A CN114049347A (en) 2021-11-26 2021-11-26 Crude oil leakage detection method based on feature enhancement

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111423141.0A CN114049347A (en) 2021-11-26 2021-11-26 Crude oil leakage detection method based on feature enhancement

Publications (1)

Publication Number Publication Date
CN114049347A true CN114049347A (en) 2022-02-15

Family

ID=80211290

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111423141.0A Pending CN114049347A (en) 2021-11-26 2021-11-26 Crude oil leakage detection method based on feature enhancement

Country Status (1)

Country Link
CN (1) CN114049347A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116485682A (en) * 2023-05-04 2023-07-25 北京联合大学 Image shadow removing system and method based on potential diffusion model
CN116704316A (en) * 2023-08-03 2023-09-05 四川金信石信息技术有限公司 Substation oil leakage detection method, system and medium based on shadow image reconstruction

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116485682A (en) * 2023-05-04 2023-07-25 北京联合大学 Image shadow removing system and method based on potential diffusion model
CN116485682B (en) * 2023-05-04 2024-03-15 北京联合大学 Image shadow removing system and method based on potential diffusion model
CN116704316A (en) * 2023-08-03 2023-09-05 四川金信石信息技术有限公司 Substation oil leakage detection method, system and medium based on shadow image reconstruction

Similar Documents

Publication Publication Date Title
CN111553929B (en) Mobile phone screen defect segmentation method, device and equipment based on converged network
CN110544251B (en) Dam crack detection method based on multi-migration learning model fusion
CN110232380B (en) Fire night scene restoration method based on Mask R-CNN neural network
CN110852316B (en) Image tampering detection and positioning method adopting convolution network with dense structure
Wang et al. RENet: Rectangular convolution pyramid and edge enhancement network for salient object detection of pavement cracks
CN105069779B (en) A kind of architectural pottery surface detail pattern quality detection method
CN108009591A (en) A kind of contact network key component identification method based on deep learning
CN114049347A (en) Crude oil leakage detection method based on feature enhancement
WO2023039959A1 (en) Remote sensing image marine and non-marine area segmentation method based on pyramid mechanism
CN109840483B (en) Landslide crack detection and identification method and device
CN111862143B (en) Automatic monitoring method for river dike collapse
CN112907598B (en) Method for detecting falsification of document and certificate images based on attention CNN
CN110472628B (en) Improved Faster R-CNN network floater detection method based on video characteristics
CN110619623B (en) Automatic identification method for heating of joint of power transformation equipment
CN111611861B (en) Image change detection method based on multi-scale feature association
CN108304786A (en) A kind of pedestrian detection method based on binaryzation convolutional neural networks
CN111524117A (en) Tunnel surface defect detection method based on characteristic pyramid network
CN116612106A (en) Method for detecting surface defects of optical element based on YOLOX algorithm
CN114241310B (en) Improved YOLO model-based intelligent identification method for piping dangerous case of dike
CN112700476A (en) Infrared ship video tracking method based on convolutional neural network
CN116206112A (en) Remote sensing image semantic segmentation method based on multi-scale feature fusion and SAM
CN114820541A (en) Defect detection method based on reconstructed network
CN114743103A (en) Island reef remote sensing image geological classification method based on Deeplabv3+ network model
CN114386504A (en) Engineering drawing character recognition method
CN110660048A (en) Leather surface defect detection algorithm based on shape characteristics

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication