Water surface oil spill detection method based on infrared polarization imaging and polarization characteristic enhancement

Abstract

The embodiment of the invention relates to the technical field of computer vision, in particular to a water surface oil spill detection method based on infrared polarization imaging and polarization characteristic enhancement, which comprises the following steps: shooting a target area within a specific observation angle by using an infrared polarization camera to obtain an infrared polarization mosaic image, the polarization information difference between the spilled oil in the target area and the water surface is maximized; inputting the infrared polarization mosaic image into a polarization feature enhancement network for denoising and demosaicing processing to obtain a multi-dimensional polarization information image so as to enhance the contrast ratio of spilled oil of the target area to the water surface; and inputting the multi-dimensional polarization information image into a detection backbone network, firstly extracting oil spill characteristics through a middle layer of the detection backbone network, then carrying out oil spill detection through a detection head of the detection backbone network, and finally obtaining and outputting an oil spill detection result of the target area. According to the method, the precision and speed of water surface oil spill detection are greatly improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer vision technology, and in particular to a method for detecting oil spills on water surfaces based on infrared polarization imaging and polarization feature enhancement. Background Art

[0002] Polarization information can reflect the physical and chemical characteristics of an object's surface, such as shape, material, surface roughness, and normal direction. Therefore, infrared polarization imaging technology can detect oil spills on water surfaces over a full time period based on the differences in polarization properties between the water surface and crude oil. However, among traditional infrared polarization imaging methods, time-sharing imaging cannot obtain polarization information in real time. However, due to the presence of waves, the water surface is constantly changing, and oil spill detection on water surfaces requires high real-time imaging, making time-sharing imaging unsuitable for oil spill detection. Amplitude-sharing imaging is too bulky and requires high-precision calibration, making it difficult to deploy on various platforms for oil spill detection. Interference between channels in aperture-sharing imaging reduces the effectiveness of the polarization information it obtains, making it difficult to effectively apply to oil spill detection scenarios. In recent years, with the rapid development of micro-nanofabrication technology, the focal plane infrared polarization imaging method has gained widespread attention in detection and identification scenarios due to its advantages of small size, low weight, and strong real-time performance.

[0003] Split-focal-plane infrared polarization imaging obtains a polarization mosaic image by integrating a micro-polarizer in front of the focal plane. The obtained polarization mosaic image needs to be demosaiced to restore images of different polarization directions of the same size as the original image, and then calculate the polarization information. However, the currently commonly used demosaicing methods cannot guarantee the accuracy of the restored demosaiced image while ensuring the real-time performance of the algorithm. If a more real-time method is used, such as quadratic interpolation, bicubic interpolation, etc., the error introduced by the demosaicing process will be amplified in the calculation process of the polarization information, seriously reducing the effectiveness of the polarization information. If a more accurate method is used, such as Newton interpolation polynomials, etc., the collected polarization mosaic image cannot be processed in real time, resulting in the algorithm's real-time performance being unable to meet the actual needs of surface oil spill detection. Summary of the Invention

[0004] In view of this, the embodiments of the present application propose a method for detecting surface oil spills based on infrared polarization imaging and polarization feature enhancement, which successfully overcomes the problem that traditional optical imaging methods cannot quickly and accurately identify surface oil spills at all times, and greatly improves the accuracy and speed of surface oil spill detection.

[0005] In order to achieve the above-mentioned objectives, an embodiment of the present application proposes a surface oil spill detection method based on infrared polarization imaging and polarization feature enhancement, which is suitable for oil spill detection in a target area, including: using an infrared polarization camera to shoot the target area within a specific observation angle to obtain an infrared polarization mosaic image to maximize the polarization information difference between the oil spill and the water surface in the target area; inputting the infrared polarization mosaic image into a polarization feature enhancement network for denoising and demosaicing processing to obtain a multidimensional polarization information image to enhance the contrast between the oil spill and the water surface in the target area; inputting the multidimensional polarization information image into the detection backbone network, first extracting the oil spill features through the middle layer of the detection backbone network, and then performing oil spill detection through the detection head of the detection backbone network, and finally obtaining and outputting the oil spill detection result of the target area.

[0006] In order to achieve the above-mentioned purpose, the embodiment of the present application also proposes a surface oil spill detection system based on infrared polarization imaging and polarization feature enhancement, which is suitable for oil spill detection in the target area, including: a shooting and acquisition module, which is used to use an infrared polarization camera to shoot the target area within a specific observation angle to obtain an infrared polarization mosaic image to maximize the polarization information difference between the oil spill and the water surface in the target area; a detection execution module, which is used to input the infrared polarization mosaic image into the trained polarization feature enhancement network for denoising and demosaicing processing to obtain a multi-dimensional polarization information image to enhance the contrast between the oil spill and the water surface in the target area, and input the multi-dimensional polarization information image into the trained detection backbone network, first extract the oil spill features through the middle layer of the trained detection backbone network, and then perform oil spill detection through the detection head of the trained detection backbone network, and finally obtain and output the oil spill detection result of the target area.

[0007] In order to achieve the above-mentioned objectives, an embodiment of the present application also proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a surface oil spill detection method based on infrared polarization imaging and polarization feature enhancement as described above.

[0008] In order to achieve the above-mentioned objectives, an embodiment of the present application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement a surface oil spill detection method based on infrared polarization imaging and polarization feature enhancement as described above.

[0009] The present application proposes a method for detecting oil spills on water surfaces based on infrared polarization imaging and polarization feature enhancement. First, an infrared polarization camera is used to shoot the target area within a specific observation angle to obtain an infrared polarization mosaic image to maximize the polarization information difference between the oil spill in the target area and the water surface. The infrared polarization mosaic image is then input into a polarization feature enhancement network for denoising and demosaicing to obtain a multi-dimensional polarization information image to enhance the contrast between the oil spill and the water surface in the target area. Finally, the multi-dimensional polarization information image is input into the detection backbone network. The oil spill features are first extracted through the middle layer of the detection backbone network, and then the oil spill detection is performed through the detection head of the detection backbone network. Finally, the oil spill detection result of the target area is obtained and output. This method successfully overcomes the problem that traditional optical imaging methods cannot quickly and accurately identify oil spills on water surfaces at all times, and greatly improves the accuracy and speed of surface oil spill detection.

[0010] Optionally, the specific observation angle ranges from 20 degrees to 50 degrees.

[0011] Optionally, the polarization feature enhancement network is specifically composed of a first-layer denoising network, a second-layer demosaicing network and a third-layer enhancement network; the infrared polarization mosaic image is input into the first-layer denoising network, the first-layer denoising network denoises the infrared polarization mosaic image to obtain a denoised mosaic image, and at the same time, the polarization feature enhancement network uses a Sobel edge extraction operator to extract edge information from the infrared polarization mosaic image; the denoised mosaic image and edge information are input into the second-layer demosaicing network, the second-layer demosaicing network demosaics the denoised mosaic image based on the edge information to obtain preliminary demosaiced images of several channels, and at the same time, the polarization feature enhancement network again uses the Sobel edge extraction operator to extract edge information of different channels from the preliminary demosaiced images of several channels; the preliminary demosaiced image and edge information of different channels are input into the third-layer enhancement network, the third-layer enhancement network enhances the preliminary demosaiced image based on the edge information of different channels to obtain final demosaiced images of several channels.

[0012] Optionally, the main structures of the first denoising network and the second demosaicing network are composed of convolutional layers and dense residual structures. The dense residual structure consists of The main structure of the third layer enhanced network consists of The residual channel attention structure is composed of a series of residual channel attention structures, each of which consists of a convolutional layer, a residual layer, an average pooling layer, an activation function and a cross-layer connection; and are all integers greater than 2.

[0013] Optionally, the first-layer denoising network performs denoising on the infrared polarization mosaic image to obtain a denoised mosaic image, which is achieved by the following formula: ; in, represents an infrared polarization mosaic image, represents the convolutional layer, represents a dense residual structure, represents the mosaic image after denoising; The polarization feature enhancement network uses the Sobel edge extraction operator to extract edge information from the infrared polarization mosaic image, which is achieved by the following formula: ; in, represents the convolution operation, represents the horizontal gradient operator, represents the vertical gradient operator, Represents the extracted edge information; The second-layer demosaicing network performs demosaicing on the denoised mosaic image based on edge information to obtain preliminary demosaiced images of several channels, which is achieved through the following formula: ; in, , respectively represent the preliminary demosaiced images of 0 degree, 45 degree, 90 degree, and 135 degree channels; The polarization feature enhancement network again uses the Sobel edge extraction operator to extract edge information of different channels from the preliminary demosaiced images of several channels, which is achieved by the following formula: ; in, , respectively represent the edge information of 0 degree, 45 degree, 90 degree, and 135 degree channels; The third layer enhancement network enhances the preliminary demosaiced image based on the edge information of different channels to obtain the final demosaiced images of several channels, which is achieved by the following formula: ; in, , respectively represent the final demosaiced images of 0 degree, 45 degree, 90 degree, and 135 degree channels, express A concatenated residual channel attention structure.

[0014] Optionally, before training the polarization feature enhancement network and the detection backbone network, the polarization feature enhancement network is pre-trained, and then the pre-trained polarization feature enhancement network and the detection backbone network are trained synchronously. This training method can ensure the effectiveness of the training results.

[0015] Optionally, when pre-training the polarization feature enhancement network, a common infrared camera with an external polarizer and an infrared polarization camera are used to synchronously capture scene data to obtain a demosaiced true value image and an infrared polarization mosaic image, respectively. The demosaiced true value image and the result of the third-layer enhancement network reconstruction and enhancement are used to calculate the true value multidimensional polarization information image and the enhanced multidimensional polarization information image, respectively. The multidimensional polarization information image is used to calculate the loss function to improve the accuracy of the training results. The multi-dimensional polarization information image is expressed by the formula: ; ; ; ; ; ; ; ; ; ; ; in, 、 、 、 Represents the demosaiced true value images of 0 degree, 45 degree, 90 degree, and 135 degree channels respectively, Indicates that the mosaic image is constructed pixel by pixel based on the structure of the demosaiced true value image and polarization mosaic image of different channels. represents the denoised mosaic ground truth image constructed pixel by pixel, and They represent the true polarization degree image and the true polarization angle image calculated based on the demosaiced true value image, respectively. 、 and They represent the Stokes vectors S0, S1 and S2 images calculated from the demosaiced true value images of different channels, and They represent the polarization degree reconstructed image and polarization angle reconstructed image calculated based on the results of the third layer enhancement network reconstruction enhancement, respectively. 、 and They represent the Stokes vectors S0, S1 and S2 images calculated from the reconstruction enhancement results of different channels respectively; When pre-training the polarization feature enhancement network, the total loss function used is expressed as follows: ; ; ; ; ; in, represents the total loss function, 、 、 、 、 、 are weight coefficients, represents the L1 norm. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the related technologies, the following is a brief introduction to the drawings required for use in the embodiments of the present application or the description of the related technologies. Obviously, the following drawings are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. The drawings described here are only used to explain the present application and are not used to limit the present application.

[0017] Figure 1 This is a flow chart of a method for detecting oil spills on a water surface based on infrared polarization imaging and polarization feature enhancement, provided in one embodiment of the present application; Figure 2 This is a visual schematic diagram of the process of a method for detecting oil spills on a water surface based on infrared polarization imaging and polarization feature enhancement, provided in one embodiment of the present application; Figure 3 An embodiment of the present application provides a polarization pseudo-color image of oil spill on a water surface collected at different observation angles using an infrared polarization camera; Figure 4 This is a schematic structural diagram of a polarization characteristic enhancement network provided in one embodiment of the present application; Figure 5 This is a diagram of the results of oil spill detection on a water surface provided in one embodiment of the present application; Figure 61 is a schematic structural diagram of a surface oil spill detection system based on infrared polarization imaging and polarization feature enhancement, provided in another embodiment of the present application; Figure 7 It is a structural diagram of an electronic device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below with reference to the accompanying drawings. Those skilled in the art will appreciate that in the embodiments of the present application, many technical details are proposed to enable the reader to better understand. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in the present application can be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation on the specific implementation of the present application. The following embodiments can be combined with each other and referenced to each other under the premise of no contradiction.

[0019] One embodiment of the present application proposes a method for detecting oil spills on a water surface based on infrared polarization imaging and polarization feature enhancement. The method is suitable for detecting oil spills on a target water surface and is applied to an electronic device, where the electronic device can be a terminal or a server. This embodiment and the following embodiments are described using a server as an example. The following describes the implementation details of the method for detecting oil spills on a water surface based on infrared polarization imaging and polarization feature enhancement proposed in this embodiment. The following content is provided for ease of understanding only and is not required for implementing this solution.

[0020] The specific process of the surface oil spill detection method based on infrared polarization imaging and polarization feature enhancement proposed in this embodiment can be as follows: Figure 1 As shown, the visual details are as follows Figure 2 As shown, the method specifically includes: Step 101 : Use an infrared polarization camera to photograph the target area within a specific observation angle to obtain an infrared polarization mosaic image to maximize the difference in polarization information between the oil spill in the target area and the water surface.

[0021] In the specific implementation, the first step in water surface oil spill detection is to collect raw data, that is, to use an infrared polarization camera based on focal plane technology to shoot the target water surface within a specific observation angle to obtain (collect) a 14-bit infrared polarization mosaic image. And upload it to the server synchronously to maximize the difference in polarization information between the oil spill and the water surface in the target area.

[0022] In one example, Figure 2As shown in the figure, the polarization pseudo-color images generated by calculating the oil spill image on the water surface acquired by an infrared polarization camera at different observation angles are different. When the observation angle is less than 20 degrees, the polarization information difference between the oil droplets and the water surface is significant, making it difficult to distinguish the oil film from the water surface. When the observation angle is greater than 50 degrees, the polarization information difference between the oil film and the water surface is absent, and when it is greater than 60 degrees, the oil film and the water surface cannot be distinguished at all. Therefore, when acquiring infrared polarization mosaic images, the observation angle should be between 20 and 50 degrees to improve the recognition of the oil film.

[0023] In step 102 , the infrared polarization mosaic image is input into a polarization feature enhancement network for denoising and demosaicing to obtain a multi-dimensional polarization information image to enhance the contrast between the oil spill and the target water surface.

[0024] In the specific implementation, after receiving the uploaded infrared polarization mosaic image, the server needs to input the infrared polarization mosaic image into the polarization feature enhancement network for denoising and demosaicing processing to obtain a multi-dimensional polarization information image to enhance the contrast between the oil spill and the water surface in the target area.

[0025] In one example, Figure 4 As shown in the figure, the polarization feature enhancement network is specifically composed of a first-layer denoising network, a second-layer demosaicing network and a third-layer enhancement network.

[0026] The main structures of the first-layer denoising network and the second-layer demosaicing network are composed of convolutional layers and dense residual structures. The dense residual structure is composed of The residual modules of the dense residual structure are composed of convolutional layers, residual layers, normalization layers and cross-layer connections. The convolutional layers, residual layers and normalization layers are connected in series in sequence, and the cross-layer connections connect the input of the convolutional layer and the output of the normalization layer. The main structure of the third layer enhancement network is composed of The residual channel attention structure is composed of a series of residual channel attention structures, each of which consists of a convolutional layer, a residual layer, an average pooling layer, an activation function and a cross-layer connection. and are all integers greater than 2.

[0027] In the residual channel attention structure, a residual layer is sandwiched between two convolutional layers to form a three-layer structure. The first three-layer structure, the average pooling layer, and the second three-layer structure are connected sequentially. The second three-layer structure is followed by an activation function. The first cross-layer connection is responsible for connecting the output of the first three-layer structure and the output of the activation function for element-by-element multiplication, and the second cross-layer connection is responsible for connecting the input of the first three-layer structure and the output of the first cross-layer connection for element-by-element addition.

[0028] The infrared polarization mosaic image is input to the first layer denoising network, which uses convolutional layers and contains A dense residual structure of residual modules is used to denoise the infrared polarization mosaic image to obtain a denoised mosaic image. At the same time, the polarization feature enhancement network uses the Sobel edge extraction operator to extract edge information from the infrared polarization mosaic image.

[0029] It should be noted that the Sobel edge extraction operator includes a horizontal edge extraction operator and a vertical edge extraction operator, which work together.

[0030] The first-layer denoising network denoises the infrared polarization mosaic image to obtain the denoised mosaic image, which can be achieved by the following formula: ; in, represents an infrared polarization mosaic image, represents the convolutional layer, represents a dense residual structure, Represents the mosaic image after denoising.

[0031] The polarization feature enhancement network uses the Sobel edge extraction operator to extract edge information from the infrared polarization mosaic image, which can be achieved by the following formula: ; in, represents the convolution operation, represents the horizontal gradient operator, represents the vertical gradient operator, Represents the extracted edge information.

[0032] The denoised mosaic image and edge information are input to the second demosaicing network, which uses convolutional layers and A dense residual structure of residual modules is constructed. Based on the edge information, the denoised mosaic image is demosaiced to obtain preliminary demosaiced images of several channels. At the same time, the polarization feature enhancement network again uses the Sobel edge extraction operator to extract edge information of different channels from the preliminary demosaiced images of several channels.

[0033] The second-layer demosaicing network performs demosaicing on the denoised mosaic image based on edge information to obtain preliminary demosaiced images of several channels, which can be achieved by the following formula: ; in, , which represent the preliminary demosaiced images of 0 degree, 45 degree, 90 degree, and 135 degree channels, respectively.

[0034] The polarization feature enhancement network again uses the Sobel edge extraction operator to extract edge information of different channels from the preliminary demosaiced images of several channels, which can be achieved by the following formula: ; in, , representing the edge information of 0 degree, 45 degree, 90 degree, and 135 degree channels respectively.

[0035] The preliminary demosaiced image and the edge information of different channels are input into the third-layer enhancement network. The third-layer enhancement network uses the residual channel attention structure to enhance the preliminary demosaiced image based on the edge information of different channels to obtain the final demosaiced images of several channels.

[0036] The third layer enhancement network enhances the preliminary demosaiced image based on the edge information of different channels to obtain the final demosaiced images of several channels, which can be achieved by the following formula: ; in, , respectively represent the final demosaiced images of 0 degree, 45 degree, 90 degree, and 135 degree channels, express A concatenated residual channel attention structure.

[0037] It should be noted that the final demosaiced image is a part of the multi-dimensional polarization information image, which also includes a polarization degree image and a polarization angle image calculated based on the final demosaiced image.

[0038] In step 103, the multi-dimensional polarization information image is input into the detection backbone network. The oil spill features are first extracted through the middle layer of the detection backbone network, and then the oil spill detection is performed through the detection head of the detection backbone network. Finally, the oil spill detection result of the target area is obtained and output.

[0039] In the specific implementation, after obtaining the multi-dimensional polarization information image, the server can input the multi-dimensional polarization information image into the detection backbone network, first extract the oil spill features from the multi-dimensional polarization information image through the middle layer of the detection backbone network, and then use the detection head of the detection backbone network to perform oil spill detection using the extracted oil spill features, and finally obtain and output the oil spill detection results of the target area.

[0040] In one example, the results of oil spill detection on the water surface are as follows: Figure 5 As shown. Figure 5 It can be seen that this method can use polarization information to effectively improve the contrast between oil spill and water surface, and use the difference in polarization information to accurately detect oil spill on the water surface at different angles.

[0041] In one example, the polarization feature enhancement network is pre-trained before training the polarization feature enhancement network and the detection backbone network. The pre-trained polarization feature enhancement network and the detection backbone network are then trained synchronously. This training method ensures the effectiveness of the training results.

[0042] In one example, when pre-training the polarization feature enhancement network, an ordinary infrared camera with an external polarizer and an infrared polarization camera are used to synchronously collect scene data to obtain a demosaiced true value image and an infrared polarization mosaic image, respectively. The demosaiced true value image and the result of the third-layer enhancement network reconstruction enhancement are used to calculate the true value multidimensional polarization information image and the enhanced multidimensional polarization information image, respectively. The multidimensional polarization information image is used to calculate the loss function to improve the accuracy of the training results.

[0043] The multi-dimensional polarization information image is expressed by the formula: ; ; ; ; ; ; ; ; ; ; ; in, 、 、 、 Represents the demosaiced true value images of 0 degree, 45 degree, 90 degree, and 135 degree channels respectively, Indicates that the mosaic image is constructed pixel by pixel based on the structure of the demosaiced true value image and polarization mosaic image of different channels. represents the denoised mosaic ground truth image constructed pixel by pixel, and They represent the true polarization degree image and the true polarization angle image calculated based on the demosaiced true value image, respectively. 、 and They represent the Stokes vectors S0, S1 and S2 images calculated from the demosaiced true value images of different channels, and They represent the polarization degree reconstructed image and polarization angle reconstructed image calculated based on the results of the third layer enhancement network reconstruction enhancement, respectively. 、 and They represent the Stokes vector S0, S1 and S2 images calculated from the reconstruction enhancement results of different channels.

[0044] When pre-training the polarization feature enhancement network, the total loss function used is expressed as follows: ; ; ; ; ; in, represents the total loss function, 、 、 、 、 、 are weight coefficients, represents the L1 norm.

[0045] This allows the polarization feature enhancement network to be pre-trained before detection to ensure the validity of the polarization information it recovers. Afterward, the pre-trained polarization feature enhancement network and the detection backbone network are trained simultaneously. Once training is complete, the trained polarization feature enhancement network and the detection backbone network can be used to achieve rapid and accurate detection of oil spills on the surface.

[0046] In this embodiment, an infrared polarization camera is first used to photograph the target area within a specific observation angle to obtain an infrared polarization mosaic image to maximize the difference in polarization information between the oil spill in the target area and the water surface. The infrared polarization camera then uploads the infrared polarization mosaic image to a server deployed with a polarization feature enhancement network and a detection backbone network. The server inputs the infrared polarization mosaic image into the polarization feature enhancement network for denoising and demosaicing to obtain a multi-dimensional polarization information image to enhance the contrast between the oil spill and the water surface in the target area. Finally, the multi-dimensional polarization information image is input into the detection backbone network, and the oil spill features are first extracted through the middle layer of the detection backbone network. The oil spill detection is then performed through the detection head of the detection backbone network, and finally the oil spill detection result of the target area is obtained and output. This method successfully overcomes the problem that traditional optical imaging methods cannot quickly and accurately identify oil spills on the water surface at all times, and greatly improves the accuracy and speed of oil spill detection on the water surface.

[0047] The steps of the various methods above are divided for clarity of description only. During implementation, they can be combined into a single step, or some steps can be broken down into multiple steps. As long as they contain the same logical relationships, they are all within the scope of protection of this application. Adding minor modifications or introducing minor design changes to the algorithm or process, but not changing the core design of the algorithm or process, are also within the scope of protection of this application.

[0048] Another embodiment of the present application proposes a surface oil spill detection system based on infrared polarization imaging and polarization feature enhancement, which is suitable for oil spill detection on the target water surface. The following is a detailed description of the details of the surface oil spill detection system based on infrared polarization imaging and polarization feature enhancement proposed in this embodiment. The following content is only the implementation details provided for the convenience of understanding and is not necessary for the implementation of this embodiment. The specific structure of the surface oil spill detection system based on infrared polarization imaging and polarization feature enhancement proposed in this embodiment can be as follows: Figure 6 As shown, it includes: a shooting and collecting module 201 and a detection execution module 202.

[0049] The shooting and acquisition module 201 is used to shoot the target area using an infrared polarization camera within a specific observation angle to obtain an infrared polarization mosaic image to maximize the difference in polarization information between the oil spill in the target area and the water surface.

[0050] The detection execution module 202 is used to input the infrared polarization mosaic image into the trained polarization feature enhancement network for denoising and demosaicing processing to obtain a multi-dimensional polarization information image to enhance the contrast between the oil spill and the water surface in the target area. The multi-dimensional polarization information image is input into the trained detection backbone network, the oil spill features are first extracted through the middle layer of the trained detection backbone network, and then the oil spill detection is performed through the detection head of the trained detection backbone network, and finally the oil spill detection result of the target area is obtained and output.

[0051] It is not difficult to find that this embodiment is a system embodiment corresponding to the above-mentioned method embodiment, and this embodiment can be implemented in conjunction with the above-mentioned method embodiment. The relevant technical details and technical effects mentioned in the above-mentioned method embodiment are still valid in this embodiment, and to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above-mentioned method embodiment.

[0052] It is worth mentioning that all modules and modules involved in this embodiment are logical modules. In actual applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, to highlight the innovation of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by this application. However, this does not mean that other units do not exist in this embodiment.

[0053] Another embodiment of the present application provides an electronic device, such as Figure 7 As shown, it includes: at least one processor 301; and a memory 302 communicatively connected to the at least one processor 301; wherein the memory 302 stores instructions that can be executed by the at least one processor 301, and the instructions are executed by the at least one processor 301 to enable the at least one processor 301 to execute a surface oil spill detection method based on infrared polarization imaging and polarization feature enhancement as described in the above method embodiment.

[0054] The memory and processor are connected using a bus, which includes any number of interconnected buses and bridges. The bus connects various circuits of one or more processors and memories. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits. These are all well known in the art and therefore will not be described further in this article. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. Data processed by the processor is transmitted on a wireless medium via an antenna. Furthermore, the antenna also receives data and transmits it to the processor.

[0055] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0056] Another embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it can implement a surface oil spill detection method based on infrared polarization imaging and polarization feature enhancement as described in the above method embodiment.

[0057] That is, those skilled in the art will understand that all or part of the steps in the above-described method embodiments can be implemented by instructing the relevant hardware through a program, which is stored in a storage medium and includes a number of instructions for causing a device (such as a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps in the method embodiments of the present application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory, a random access memory, a magnetic disk, or an optical disk.

[0058] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application. In actual applications, various modifications may be made to the embodiments in form and detail without departing from the spirit and scope of the present application. Those skilled in the art will appreciate that improvements and modifications may be made without departing from the principles of the present application, and such improvements and modifications are also considered to be within the scope of protection of the present application.

Claims

1. A method for detecting oil spills on a water surface based on infrared polarization imaging and polarization feature enhancement, suitable for detecting oil spills in a target area, characterized in that: The method comprises: Use an infrared polarization camera to photograph the target area within a specific observation angle to obtain an infrared polarization mosaic image to maximize the difference in polarization information between the oil spill and the water surface in the target area; The infrared polarization mosaic image is input into the polarization feature enhancement network for denoising and demosaicing to obtain a multi-dimensional polarization information image to enhance the contrast between the oil spill and the water surface in the target area. The multi-dimensional polarization information image is input into the detection backbone network. The oil spill features are first extracted through the middle layer of the detection backbone network, and then the oil spill detection is performed through the detection head of the detection backbone network. Finally, the oil spill detection results of the target area are obtained and output.

2. The method for detecting oil spills on water surfaces based on infrared polarization imaging and polarization feature enhancement according to claim 1, characterized in that: The specific observation angle ranges from 20 degrees to 50 degrees.

3. The method for detecting oil spills on water surfaces based on infrared polarization imaging and polarization feature enhancement according to claim 1, characterized in that: The polarization feature enhancement network consists of a first-layer denoising network, a second-layer demosaicing network, and a third-layer enhancement network. The infrared polarization mosaic image is input into the first-layer denoising network, which denoises the infrared polarization mosaic image to obtain a denoised mosaic image. At the same time, the polarization feature enhancement network uses the Sobel edge extraction operator to extract edge information from the infrared polarization mosaic image. The denoised mosaic image and edge information are input to the second-layer demosaicing network. The second-layer demosaicing network demosaic the denoised mosaic image based on the edge information to obtain preliminary demosaiced images of several channels. At the same time, the polarization feature enhancement network again uses the Sobel edge extraction operator to extract edge information of different channels from the preliminary demosaiced images of several channels. The preliminary demosaiced image and edge information of different channels are input into the third-layer enhancement network. The third-layer enhancement network enhances the preliminary demosaiced image based on the edge information of different channels to obtain the final demosaiced images of several channels.

4. The method for detecting oil spills on water surfaces based on infrared polarization imaging and polarization feature enhancement according to claim 3, characterized in that: The main structures of the first-layer denoising network and the second-layer demosaicing network are composed of convolutional layers and dense residual structures. The dense residual structure is composed of The network consists of residual modules with residual connections, each of which consists of a convolutional layer, a residual layer, a normalization layer, and a cross-layer connection. The main structure of the third layer enhanced network is composed of It consists of a series of residual channel attention structures, each of which consists of a convolutional layer, a residual layer, an average pooling layer, an activation function and a cross-layer connection; in, and are all integers greater than 2.

5. The method for detecting oil spills on water surfaces based on infrared polarization imaging and polarization feature enhancement according to claim 4, characterized in that: The first-layer denoising network denoises the infrared polarization mosaic image to obtain the denoised mosaic image, which is achieved by the following formula: ； in, represents an infrared polarization mosaic image, represents the convolutional layer, represents a dense residual structure, represents the mosaic image after denoising; The polarization feature enhancement network uses the Sobel edge extraction operator to extract edge information from the infrared polarization mosaic image, which is achieved by the following formula: ； in, represents the convolution operation, represents the horizontal gradient operator, represents the vertical gradient operator, Represents the extracted edge information; The second-layer demosaicing network performs demosaicing on the denoised mosaic image based on edge information to obtain preliminary demosaiced images of several channels, which is achieved through the following formula: ； in, , respectively represent the preliminary demosaiced images of 0 degree, 45 degree, 90 degree, and 135 degree channels; The polarization feature enhancement network again uses the Sobel edge extraction operator to extract edge information of different channels from the preliminary demosaiced images of several channels, which is achieved by the following formula: ； in, , respectively represent the edge information of 0 degree, 45 degree, 90 degree, and 135 degree channels; The third layer enhancement network enhances the preliminary demosaiced image based on the edge information of different channels to obtain the final demosaiced images of several channels, which is achieved by the following formula: ； in, , respectively represent the final demosaiced images of 0 degree, 45 degree, 90 degree, and 135 degree channels, express A concatenated residual channel attention structure.

6. A method for detecting oil spills on water surfaces based on infrared polarization imaging and polarization feature enhancement according to any one of claims 3 to 5, characterized in that: Before training the polarization feature enhancement network and the detection backbone network, the polarization feature enhancement network is pre-trained first, and then the pre-trained polarization feature enhancement network and the detection backbone network are synchronously trained.

7. The method for detecting oil spills on water surfaces based on infrared polarization imaging and polarization feature enhancement according to claim 6, characterized in that: When pre-training the polarization feature enhancement network, a common infrared camera with an external polarizer and an infrared polarization camera are used to synchronously capture scene data to obtain a demosaiced true value image and an infrared polarization mosaic image, respectively. The demosaiced true value image and the results of the third-layer enhancement network reconstruction and enhancement are used to calculate the true value multi-dimensional polarization information image and the enhanced multi-dimensional polarization information image, respectively. The multi-dimensional polarization information image is used to calculate the loss function to improve the accuracy of the training results. The multi-dimensional polarization information image is expressed by the formula: ； ； ； ； ； ； ； ； ； ； ； in, 、 、 、 Represents the demosaiced true value images of 0 degree, 45 degree, 90 degree, and 135 degree channels respectively, Indicates that the mosaic image is constructed pixel by pixel based on the structure of the demosaiced true value image and polarization mosaic image of different channels. represents the denoised mosaic ground truth image constructed pixel by pixel, and They represent the true polarization degree image and the true polarization angle image calculated based on the demosaiced true value image, respectively. 、 and They represent the Stokes vectors S0, S1 and S2 images calculated from the demosaiced true value images of different channels, and They represent the polarization degree reconstructed image and polarization angle reconstructed image calculated based on the results of the third layer enhancement network reconstruction enhancement, respectively. 、 and They represent the Stokes vectors S0, S1 and S2 images calculated from the reconstruction enhancement results of different channels respectively; When pre-training the polarization feature enhancement network, the total loss function used is expressed as follows: ； ； ； ； ； in, represents the total loss function, 、 、 、 、 、 are weight coefficients, represents the L1 norm.

8. A surface oil spill detection system based on infrared polarization imaging and polarization feature enhancement, suitable for oil spill detection in a target area, characterized by: The system comprises: The shooting and acquisition module is used to use an infrared polarization camera to shoot the target area within a specific observation angle to obtain an infrared polarization mosaic image to maximize the difference in polarization information between the oil spill and the water surface in the target area; The detection execution module is used to input the infrared polarization mosaic image into the trained polarization feature enhancement network for denoising and demosaicing processing to obtain a multi-dimensional polarization information image to enhance the contrast between the oil spill and the water surface in the target area. The multi-dimensional polarization information image is input into the trained detection backbone network, and the oil spill features are first extracted through the middle layer of the trained detection backbone network. Then, oil spill detection is performed through the detection head of the trained detection backbone network, and finally the oil spill detection results of the target area are obtained and output.

9. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a surface oil spill detection method based on infrared polarization imaging and polarization feature enhancement as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it can implement a surface oil spill detection method based on infrared polarization imaging and polarization feature enhancement as described in any one of claims 1 to 7.

Images