ROV underwater adaptive illumination algorithm, storage medium and equipment

Abstract

The invention relates to an ROV underwater adaptive illumination algorithm, a storage medium and equipment, and the algorithm combines a convolutional neural network and a generative adversarial network to improve the image quality, and can intelligently adjust the current of a light source according to illumination data collected in real time, so as to adjust the illumination intensity of the light source, thereby guaranteeing the illumination intensity of the light source under different water depth and water quality conditions. And unnecessary energy consumption is reduced by intelligently adjusting the working mode of the light source, so that the cruising ability of the ROV in a long-time underwater task is improved. According to the ROV underwater adaptive illumination algorithm provided by the invention, the definition and the visual effect of the image can be remarkably improved under a low illumination condition, and the ROV underwater adaptive illumination algorithm provided by the invention has a real-time response capability, so that the illumination condition can be rapidly adjusted in a complex and variable underwater environment.

Description

Technical Field

[0001] The present invention relates to underwater imaging technology, and in particular to an ROV underwater adaptive illumination algorithm, a storage medium and a device. Background Art

[0002] A remotely operated vehicle (ROV) is a device that can move underwater, has a vision and perception system, and uses a manipulator or other tools through remote control to replace or assist humans in completing certain underwater operations.

[0003] The propagation of light in an underwater environment has unique characteristics, mainly including light attenuation, scattering and absorption. As light propagates in water, the intensity of light decays rapidly, especially different wavelengths of light have different attenuation rates in water, blue light and green light have relatively strong penetration, while red light decays the fastest. In addition, suspended particles in the water can cause light to scatter, forming diffuse light, reducing the contrast and clarity of the image. Water also has a strong ability to absorb light, especially long-wavelength light, such as red light and ultraviolet light, which will be absorbed by water molecules and suspended matter, which not only reduces the intensity of visible light, but also affects the lighting effect of the ROV, making underwater photography and detection more difficult.

[0004] Current underwater imaging technology faces certain challenges in dealing with complex lighting environments. Image quality is often affected by the attenuation, scattering, and absorption characteristics of underwater light. Common methods include physical model-based imaging technology and data-driven imaging technology. Physical model technologies such as polarization imaging, structured light scanning, and light field imaging can obtain high-dimensional information, but there are challenges in dealing with dynamic changes and complex nonlinear factors. Most existing underwater lighting systems use fixed light sources and cannot be dynamically adjusted according to environmental changes. Summary of the invention

[0005] In view of this, the present invention provides an ROV underwater adaptive lighting algorithm, a storage medium and a device, so as to solve or at least alleviate one or more of the above-mentioned problems and other problems existing in the prior art.

[0006] In order to achieve the aforementioned object, a first aspect of the present invention provides an ROV underwater adaptive illumination algorithm, wherein the algorithm comprises the following steps: Step S1: input an underwater initial image; Step S2: using a histogram analysis technique to detect the image quality, if the image quality detection result is unqualified, proceed to step S3, if the image quality detection result is qualified, proceed to step S4; Step S3: Determine whether the light source is turned on. If the light source is not turned on, turn on the light source and proceed to step S1. If the light source is turned on, proceed to step S4. Step S4: extracting multiple features of the image based on a convolutional neural network; Step S5: performing three-dimensional reconstruction based on the multiple features extracted from the image, processing the extracted multiple features using a generative adversarial network to generate a three-dimensional image, and the generative adversarial network calculates the quality of the three-dimensional image; Step S6: Determine the quality of the three-dimensional image. If the quality of the three-dimensional image is unqualified, adjust the light intensity of the light source and proceed to step S1. If the quality of the three-dimensional image is qualified, complete the light adjustment and image processing of the current cycle and prepare to enter the next cycle.

[0007] In the algorithm as described above, optionally, the step S1 includes: acquiring an initial underwater image through a sensor and / or a camera of the ROV, and the light source is an LED light.

[0008] In the algorithm as described above, optionally, the step of detecting the image quality using a histogram analysis technique in step S2 includes: Preprocessing the input image, including denoising and normalization; Calculating the grayscale histogram of the image and analyzing the brightness distribution of the image; According to the uniformity and contrast of the grayscale histogram, the quality of the image is evaluated and a target image quality score is set. , and set a predetermined range, the predetermined range defines the target image quality score The acceptable fluctuation interval of the image quality score is calculated. , if the image quality score Within the predetermined range, it indicates that the image quality is qualified. If the image quality score If the image is not within the predetermined range, it means that the image quality is unqualified.

[0009] In the algorithm as described above, optionally, the convolutional neural network includes: an input layer, the input layer receiving the image; A convolutional layer, wherein the convolutional layer extracts features of different scales and levels from the image through a convolution operation, including edges, textures, and high-order features; A pooling layer, wherein the pooling layer reduces the dimension of the feature map and retains the main features through a maximum pooling or an average pooling operation, wherein the main features include low-order features and high-order features, wherein the low-order features include edges and textures, and the high-order features include shapes and postures of objects; An activation function, wherein the activation function includes a linear rectification function ReLU, and the activation function is used to introduce nonlinear characteristics and enhance the expressiveness of the model.

[0010] In the algorithm as described above, optionally, the generative adversarial network includes: A generator, wherein the generator generates a three-dimensional image according to the multiple features extracted by the convolutional neural network; A discriminator is provided to evaluate the quality of the three-dimensional image generated by the generator.

[0011] In the algorithm as described above, optionally, the loss function of the convolutional neural network includes an anti-disturbance loss term, and the anti-disturbance loss term is used to train the robustness of the convolutional neural network under noise conditions. The loss function is: in, is the mean squared error loss, which measures the difference in image pixel values: ; in, is the structural similarity loss, which is used to measure the structural similarity of images: ; in, is the adversarial loss, which is used to enhance the details and contrast of the image and is achieved by generating adversarial networks: ; in, is the anti-disturbance loss, which is used to enhance the robustness of the model under different noise conditions: is the noise parameter; in, is a loss term for special cases, including turbidity and color deviation: in, and is the weight coefficient of the loss term.

[0012] In the algorithm as described above, optionally, the algorithm includes an image quality evaluation equation, and the image quality evaluation equation is: in, is the image quality score; in, It is the contrast evaluation index of the image; in, It is the edge fidelity evaluation index; in, It is an evaluation index of brightness consistency; in, is the water turbidity variable; in, is the light color deviation variable; in, and is the weight coefficient, which is adjusted according to actual conditions.

[0013] In the algorithm as described above, optionally, the step of judging the quality of the three-dimensional image in step S6 includes: Calculate real-time image quality score , If the real-time image quality score Less than the target image quality score , indicating that the current light is insufficient, increasing the current intensity of the light source, thereby increasing the light intensity of the light source; If the real-time image quality score Greater than the target image quality score , indicating that the current light is too strong, and the current intensity of the light source should be appropriately reduced to reduce the light intensity of the light source.

[0014] In order to achieve the aforementioned objectives, the second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions or a computer program, and when the computer-executable instructions or the computer program are executed by a processor, the algorithm as described in any one of the aforementioned first aspects is implemented.

[0015] In order to achieve the aforementioned object, a third aspect of the present invention provides a data processing device, wherein the data processing device comprises: A memory, wherein the memory is used to store computer executable instructions or computer programs; A processor, wherein the processor is used to implement the algorithm as described in any one of the first aspects above when executing the computer executable instructions or computer programs stored in the memory.

[0016] The ROV underwater adaptive lighting algorithm provided by the present invention combines convolutional neural networks and generative adversarial networks to improve image quality, and can intelligently adjust the current of the light source according to the real-time collected lighting data, thereby adjusting the light intensity of the light source to ensure that the best lighting effect is obtained under different water depths and water quality conditions, and reduce unnecessary energy consumption by intelligently adjusting the working mode of the light source to improve the endurance of the ROV in long-term underwater missions. The ROV underwater adaptive lighting algorithm provided by the present invention can significantly improve the clarity and visual effects of the image under low light conditions, and the ROV underwater adaptive lighting algorithm provided by the present invention has real-time response capabilities, ensuring that the lighting conditions can be quickly adjusted in complex and changeable underwater environments.

[0017] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The disclosure of the present invention will become more apparent with reference to the accompanying drawings. It should be understood that these drawings are only for illustrative purposes and are not intended to limit the scope of protection of the present invention. In the drawings: Figure 1 A schematic diagram of a flow chart of an embodiment of an ROV underwater adaptive illumination algorithm of the present invention; Figure 2 An image that has not been processed by the ROV underwater adaptive illumination algorithm of the present invention; Figure 3 Processed by the ROV underwater adaptive illumination algorithm of the present invention Figure 2 The result diagram is shown in Figure 2. DETAILED DESCRIPTION

[0019] Embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not intended to limit the scope of protection of the present invention.

[0020] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment". The relevant definitions of other terms will be given in the following description.

[0021] Figure 1 The figure is a flow chart of an embodiment of the ROV underwater adaptive illumination algorithm of the present invention.

[0022] like Figure 1As shown, the ROV underwater adaptive lighting algorithm may include the following steps S1 to S6.

[0023] In step S1, an underwater initial image is input.

[0024] In some embodiments, the underwater initial image is an original image captured by the sensor and / or camera of the ROV in the underwater environment. It is understandable that the particularity of the underwater environment, such as the absorption and scattering of water, will cause the image quality to deteriorate, so the ROV needs to be equipped with lighting equipment to optimize the imaging effect. In an optional embodiment, the type of lighting equipment is not limited, and it can be an LED light source. LED lights can provide high-brightness lighting, and because LED lights do not have fragile filaments, they can better withstand the vibration and impact of underwater operations.

[0025] In step S2, the image quality is detected using a histogram analysis technique. If the image quality detection result is unqualified, the process proceeds to step S3; if the image quality detection result is qualified, the process proceeds to step S4.

[0026] It can be understood that histogram analysis technology is an image processing and analysis method that evaluates image quality by counting the frequency of occurrence of each pixel value in the image. Pixel is the basic unit of the image, and each pixel has one or more pixel values ​​corresponding to it. For example, in a color image, a pixel can have three pixel values ​​RGB (red, green, and blue). In a grayscale image, a pixel can have a pixel value representing the brightness of the pixel.

[0027] In an optional embodiment, the step of detecting image quality using a histogram analysis technique includes: Preprocess the input image, including denoising and normalization; Calculate the grayscale histogram of the image and analyze the brightness distribution of the image; Evaluate the image quality based on the uniformity and contrast of the grayscale histogram and set a target image quality score , and set a predetermined range, which defines the target image quality score The acceptable fluctuation range of the image quality score is calculated. If the image quality score If the image quality score is within the preset range, it means that the image quality is qualified. If it is not within the predetermined range, it means that the image quality is unqualified.

[0028] Denoising is the process of reducing or eliminating noise in an image. Noise in an image refers to random changes that are generated during image acquisition, transmission or processing and are unrelated to the original image. It can appear as random spots, stripes or other irregular patterns in the image. Image resolution and quality can be improved by denoising the image. In an optional embodiment, the denoising method is not limited, and denoising methods such as filter-based methods and model-based methods are all considered to fall within the scope of protection of the present invention.

[0029] Specifically, normalization is to limit the image data to a specific range, and normalization can make the image processing algorithm more stable and efficient. In an optional embodiment, the normalization method is not limited, and normalization methods such as minimum-maximum normalization and nonlinear normalization methods are all considered to fall within the protection scope of the present invention.

[0030] Specifically, the grayscale histogram is a graph used in image processing to represent the number of pixels at each grayscale level in an image. Grayscale refers to the number of different brightness levels that each pixel in the image can represent. Each grayscale level corresponds to a specific brightness value. The smaller the brightness value, the lower the brightness, and the closer the color is to black. The larger the brightness value, the higher the brightness, and the closer the color is to white. The grayscale histogram can intuitively show how much each grayscale level occupies in the image. By calculating the grayscale histogram of an image, we can have an intuitive understanding of the image's contrast, brightness, and grayscale distribution.

[0031] In step S3, it is determined whether the light source is turned on. If the light source is not turned on, the light source is turned on and step S1 is performed. If the light source is turned on, step S4 is performed. In an optional embodiment, the light source is used to compensate for the lack of natural light in the underwater environment, so that the ROV can obtain clearer images. In different embodiments, the type of light source is not limited. In this embodiment, an LED light source can be used, which has the advantages of high efficiency, durability, low energy consumption and adjustability, and can provide suitable lighting in different underwater environments.

[0032] In step S4, multiple features of the image are extracted based on a convolutional neural network; In an optional embodiment, the convolutional neural network includes: Input layer, the input layer receives the image; Convolutional layer: The convolutional layer extracts features of different scales and levels from the image through convolution operations, including edges, textures, and high-order features; Pooling layer: The pooling layer reduces the dimension of the feature map and retains the main features through maximum pooling or average pooling operations. The main features include low-order features and high-order features. The low-order features include edges and textures, and the high-order features include the shape and posture of the object. Activation function, the activation function includes the linear rectification function ReLU, the activation function is used to introduce nonlinear characteristics and enhance the expressive power of the model.

[0033] It is understood that a convolutional neural network is a deep learning model. The convolution layer of a convolutional neural network extracts image features through convolution operations. The convolution operation uses a learnable convolution kernel (or filter) to slide over the image and calculate the dot product with the local image area at each position to obtain a new feature map. The convolution kernel is a small matrix, and the size of the convolution kernel determines the size of the image area covered by each convolution operation.

[0034] Specifically, the low-order features of an image refer to the original or basic features extracted directly from the image data, including texture, edges, etc. The low-order features of an image reflect the basic attributes and local structure of the image; the high-order features of an image are obtained through calculation and analysis based on the low-order features, which can represent more complex semantics and concepts in the image, including the shape and posture of objects.

[0035] In an optional embodiment, the pooling layer uses a pooling operation to replace the local area of ​​the convolutional layer input feature map. The pooling operations include maximum pooling and average pooling. Maximum pooling is to retain the maximum value in the area. Maximum pooling can capture the most significant features in the local area of ​​the feature map. Average pooling is to calculate the average of all values ​​in the local area. Average pooling can smooth the feature map and capture the average features in the local area. The presence of the pooling layer in the convolutional neural network helps to reduce computational complexity and prevent overfitting. It can be understood that the convolutional neural network in this embodiment can extract features from low-quality underwater images and help identify and retain important image details.

[0036] In different embodiments, the type and quantity of activation functions are not limited, such as linear rectification function ReLU, hyperbolic tangent activation function or exponential linear unit activation function, which can introduce nonlinear characteristics to convolutional neural networks, all fall within the protection scope of the present invention. In convolutional neural networks, activation functions enable the model to perform nonlinear transformations, improve the network's expressive power and the ability to model complex functions.

[0037] In step S5, three-dimensional reconstruction is performed based on the multiple features extracted from the image, the extracted multiple features are processed using a generative adversarial network to generate a three-dimensional image, and the generative adversarial network calculates the quality of the three-dimensional image; It is important to understand that a 3D image is an image that represents an object or scene in a 3D space. A 3D image can show the object’s three-dimensional shape, size, and spatial position. Generative adversarial networks can create 3D images based on multi-dimensional features in 2D image data.

[0038] In an optional embodiment, the generative adversarial network includes: Generator, which generates a three-dimensional image based on multiple features extracted by a convolutional neural network; Discriminator,The discriminator evaluates the quality of the 3D images produced by the generator.

[0039] It can be understood that in the generative adversarial network, the generator learns to generate high-quality images through continuous improvement, while the discriminator helps the generator improve its generation ability by distinguishing between real images and generated images, and ensures that the generated images are visually realistic and high-quality by evaluating the quality of the generated images. In this embodiment, the generative adversarial network can be used to denoise and enhance the contrast and brightness of the image, thereby significantly improving the clarity and visual effect of the image under low light conditions.

[0040] In step S6, the quality of the three-dimensional image is judged. If the quality of the three-dimensional image is unqualified, the light intensity of the light source is adjusted and step S1 is performed. If the quality of the three-dimensional image is qualified, the light adjustment and image processing of the current cycle are completed and preparations are made to enter the next cycle.

[0041] In an optional embodiment, the loss function of the convolutional neural network includes an anti-disturbance loss term, which is used to train the robustness of the convolutional neural network under noise conditions. The loss function is: in, is the mean squared error loss, which measures the difference in image pixel values: ; in, is the structural similarity loss, which is used to measure the structural similarity of images: ; in, is the adversarial loss, which is used to enhance the details and contrast of the image and is achieved by generating adversarial networks: ; in, is the anti-disturbance loss, which is used to enhance the robustness of the model under different noise conditions: is the noise parameter; in, is a loss term for special cases, including turbidity and color deviation: in, and is the weight coefficient of the loss term.

[0042] Specifically, the anti-disturbance term mainly refers to the technology used to prevent the model from overfitting. Overfitting means that the model performs well on the training data, but has poor generalization ability to new and unseen data. The anti-disturbance term can help the model reduce its sensitivity to noise and outliers in the training data. By adding the anti-disturbance term to the loss function, it helps to train the robustness of the convolutional neural network under noisy conditions and ensure that the model works stably in the face of various environmental changes. This is especially important when the underwater lighting and environmental conditions may change rapidly.

[0043] In an optional embodiment, the ROV underwater adaptive illumination algorithm of the present invention includes an image quality evaluation equation, which is used to calculate the quality of the image. The image quality evaluation equation is: in, is the image quality score; in, It is the contrast evaluation index of the image; in, It is the edge fidelity evaluation index; in, It is an evaluation index of brightness consistency; in, is the water turbidity variable; in, is the light color deviation variable; in, and is the weight coefficient, which is adjusted according to actual conditions.

[0044] In an alternative embodiment, is the contrast weight coefficient, which controls the proportion of contrast in the score. The recommended range is 0.5 ≤ ≤1.5; is the edge fidelity weight coefficient, which controls the weight of edge fidelity in the score. The recommended range of β is 0.5≤ ≤1.5; is the brightness consistency weight coefficient, which controls the proportion of brightness consistency in the score. The recommended range is 0.5 ≤ ≤2.0; is the water quality turbidity variable weight coefficient, The recommended range is 0.5 ≤ ≤1.5; is the light color deviation variable, The recommended range is 0.5 ≤ ≤1.5.

[0045] In some embodiments, the features extracted by the convolutional neural network are combined with the quality evaluation index of the features to calculate the overall quality score of the image. The quality evaluation indexes include contrast, edge fidelity and brightness consistency. Among them, contrast refers to the degree of difference between light and dark areas in the image, which is related to the visual distinction of the image; edge fidelity refers to the clarity and accuracy of edge information in the image, which is related to the recognition rate and analysis accuracy of the image; brightness consistency refers to the uniformity of brightness in different areas of the image, which affects the visual effect and readability of the image.

[0046] Through the above loss function and image quality evaluation equation, the quality of underwater images can be effectively improved to ensure that high-quality images are obtained under different lighting and special conditions, thereby improving the work efficiency and data acquisition capabilities of ROV.

[0047] In an optional embodiment, the step of judging the quality of the three-dimensional image in step S6 of the present invention includes: Calculate real-time image quality score , If the real-time image quality rating Less than target image quality score , indicating that the current light is insufficient, increasing the current intensity of the light source, thereby increasing the light intensity of the light source; If the real-time image quality rating Greater than the target image quality score , indicating that the current light is too strong, and the current intensity of the light source should be appropriately reduced to reduce the light intensity of the light source.

[0048] The algorithm provided by the present invention can intelligently adjust the current of the LED light source according to the real-time collected illumination data, thereby adjusting the intensity of the light source to ensure the best lighting effect under different water depths and water quality conditions. In addition, this intelligent light source adjustment working mode can reduce unnecessary energy consumption.

[0049] Figure 2 This is an image that has not been processed by the ROV underwater adaptive illumination algorithm of the present invention.

[0050] Figure 3 After the ROV underwater adaptive algorithm processing of the present invention Figure 2 The result diagram is shown in Figure 2.

[0051] exist Figure 3 In the process of ROV underwater adaptive algorithm processing of the present invention shown in the figure, the image evaluation quality equation is: in, is the image quality score; in, It is the contrast evaluation index of the image; in, is the edge fidelity evaluation index; in, It is an evaluation index of brightness consistency; in, is the water turbidity variable; in, is the light color deviation variable; in, and is the weight coefficient.

[0052] exist Figure 3 In the embodiment shown, the weight coefficient of the image evaluation quality equation is set, where: =1.0, =1.0, =1.5, =0.5, =0.5. For example Figure 3 , the contrast of the image is 0.8, the edge fidelity of the image is 0.7, the brightness consistency of the image The water turbidity is 0.9. and light color deviation The default value is 0. Calculated according to the image quality evaluation equation Figure 3 Quality rating: Q=1.0×0.8+1.0×0.7+1.5×0.9+0.5×0+0.5×0; Furthermore, Q=0.8+ 0.7+1.35+0+0=2.85.

[0053] By adjusting the weight coefficient of the image quality evaluation equation, the influence of each parameter on the image quality score Q is controlled. The advantage of this implementation is that it helps to maintain the consistency of image quality under different environments.

[0054] like Figure 2 and Figure 3 As shown, compared with the original Figure 2 After being processed by the algorithm provided by the present invention, the image is clearer and the details of the original image are also retained, so that the image maintains a higher quality.

[0055] Furthermore, the present invention also provides a computer-readable storage medium, which stores computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the ROV underwater adaptive lighting algorithm in any of the aforementioned embodiments is implemented.

[0056] It should be noted that the computer-readable medium of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer readable signal media may also be any computer readable medium other than computer readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0057] Computer program code for performing operations of the present invention may be written in one or more programming languages, or a combination thereof, including but not limited to object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages.

[0058] Furthermore, the present invention also provides a data processing device, which may include: a memory and a processor, wherein the memory can be used to store computer-executable instructions or computer programs, and the processor is used to execute the computer-executable instructions or computer programs stored in the memory to implement the ROV underwater adaptive lighting algorithm in any of the aforementioned embodiments.

[0059] The flow chart in the accompanying drawings illustrates the possible architecture, function and operation of the system, method and computer program product according to the embodiment of the present invention. In this regard, each square frame in the flow chart can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square frame can also occur in a sequence different from that marked in the accompanying drawings. For example, two square frames represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square frame in the flow chart, and the combination of the square frames in the flow chart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0060] The above description is only a preferred embodiment of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present invention is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present invention (but not limited to) to form a technical solution.

[0061] In addition, although each operation is described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the present invention. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0062] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. An ROV underwater adaptive illumination algorithm, characterized in that: The algorithm includes: Step S1: input an underwater initial image; Step S2: Use the histogram analysis technology to detect the image quality. If the image quality detection result is If the image quality detection result is unqualified, proceed to step S3; if the image quality detection result is qualified, proceed to step S4; Step S3: Determine whether the light source is turned on. If the light source is not turned on, turn on the light source and proceed to step S1. If the light source is turned on, proceed to step S4. Step S4: extracting multiple features of the image based on a convolutional neural network; Step S5: performing three-dimensional reconstruction based on the multiple features extracted from the image, processing the extracted multiple features using a generative adversarial network to generate a three-dimensional image, and the generative adversarial network calculates the quality of the three-dimensional image; Step S6: Determine the quality of the three-dimensional image. If the quality of the three-dimensional image is unqualified, adjust the light intensity of the light source and proceed to step S1. If the quality of the three-dimensional image is qualified, complete the light adjustment and image processing of the current cycle and prepare to enter the next cycle.

2. The ROV underwater adaptive illumination algorithm as claimed in claim 1, characterized in that: The step S1 comprises: acquiring an initial underwater image through a sensor and / or a camera of the ROV, and the light source is an LED lamp.

3. The ROV underwater adaptive illumination algorithm as claimed in claim 1, characterized in that: The step of using the histogram analysis technology to detect the image quality in step S2 includes: Preprocessing the input image, including denoising and normalization; Calculating the grayscale histogram of the image and analyzing the brightness distribution of the image; According to the uniformity and contrast of the grayscale histogram, the quality of the image is evaluated and a target image quality score is set. , and set a predetermined range, the predetermined range defines the target image quality score The acceptable fluctuation interval of the image quality score is calculated. , if the image quality score Within the predetermined range, it indicates that the image quality is qualified. If the image quality score If the image is not within the predetermined range, it means that the image quality is unqualified.

4. The ROV underwater adaptive illumination algorithm according to claim 1, characterized in that: The convolutional neural network comprises: an input layer, the input layer receiving the image; A convolutional layer, wherein the convolutional layer extracts features of different scales and levels from the image through a convolution operation, including edges, textures, and high-order features; A pooling layer, wherein the pooling layer reduces the dimension of the feature map and retains the main features through a maximum pooling or an average pooling operation, wherein the main features include low-order features and high-order features, wherein the low-order features include edges and textures, and the high-order features include shapes and postures of objects; An activation function, wherein the activation function includes a linear rectification function ReLU, and the activation function is used to introduce nonlinear characteristics and enhance the expressiveness of the model.

5. The ROV underwater adaptive illumination algorithm as claimed in claim 1, characterized in that: The generative adversarial network comprises: A generator, wherein the generator generates a three-dimensional image according to the multiple features extracted by the convolutional neural network; A discriminator is provided to evaluate the quality of the three-dimensional image generated by the generator.

6. The ROV underwater adaptive illumination algorithm according to claim 4 or 5, characterized in that: The loss function of the convolutional neural network includes an anti-disturbance loss term, which is used to train the robustness of the convolutional neural network under noise conditions. The loss function is: in, is the mean squared error loss, which measures the difference in image pixel values: ； in, is the structural similarity loss, which is used to measure the structural similarity of images: ； in, is the adversarial loss, which is used to enhance the details and contrast of the image and is achieved by generating adversarial networks: ； in, is the anti-disturbance loss, which is used to enhance the robustness of the model under different noise conditions: is the noise parameter; in, is a loss term for special cases, including turbidity and color deviation: in, and is the weight coefficient of the loss term.

7. The ROV underwater adaptive lighting algorithm as claimed in claim 1, characterized in that: The algorithm includes an image quality evaluation equation, which is used to calculate the image quality score. The image quality evaluation equation is: in, is the image quality score; in, It is the contrast evaluation index of the image; in, It is the edge fidelity evaluation index; in, It is an evaluation index of brightness consistency; in, is the water turbidity variable; in, is the light color deviation variable; in, and is the weight coefficient, which is adjusted according to actual conditions.

8. The ROV underwater adaptive illumination algorithm according to claim 1, characterized in that: The step of judging the quality of the three-dimensional image in step S6 includes: Calculate real-time image quality score , If the real-time image quality score Less than the target image quality score , indicating that the current light is insufficient, increasing the current intensity of the light source, thereby increasing the light intensity of the light source; If the real-time image quality score Greater than the target image quality score , indicating that the current light is too strong, and the current intensity of the light source should be appropriately reduced to reduce the light intensity of the light source.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions or a computer program, and when the computer-executable instructions or the computer program are executed by a processor, the algorithm as described in any one of claims 1 to 8 is implemented.

10. A data processing device, characterized in that: The data processing device comprises: A memory, wherein the memory is used to store computer executable instructions or computer programs; A processor, wherein the processor is configured to implement the algorithm as claimed in any one of claims 1 to 8 when executing the computer executable instructions or computer program stored in the memory.

Images