A method for 3D reconstruction of underwater fish bodies in smart fisheries based on neural radiation fields

Abstract

A method for 3D reconstruction of underwater fish in a smart fish farm based on neural radiation field includes the following steps: S1, obtaining visible light image data of individual underwater fish in the fish farm through a camera; S2, calculating the internal and external parameters of the camera corresponding to the visible light image; S3, constructing a neural radiation field model of underwater fish in a smart fish farm; S4, using the obtained visible light image data and the corresponding internal and external parameters of the camera as training data to train the neural radiation field model of underwater fish; during the training, the predicted image under the corresponding viewing angle is calculated by a neural radiation field volume rendering method based on the underwater light propagation attenuation coefficient; S5, using the trained neural radiation field model of underwater fish to perform 3D reconstruction of the fish. By introducing the underwater light propagation attenuation factor in the neural radiation field volume rendering method, the attenuation characteristics of light propagating in water can be effectively simulated. This method can better adapt to the underwater environment and significantly improve the accuracy and realism of 3D reconstruction of fish in the underwater environment.

Description

Technical Field

[0001] The present invention relates to computer vision and three-dimensional reconstruction technology, and in particular to a three-dimensional reconstruction method for underwater fish bodies in a smart fishery based on a neural radiation field. Background Art

[0002] In the digital twin task of smart fish farms, the 3D reconstruction of digital twins of individual fish is the basis for fish behavior analysis, fish growth and health monitoring. Existing 3D reconstruction methods for fish mainly include manual reconstruction, active 3D reconstruction, and passive 3D reconstruction. Among them, the manual reconstruction method is usually done by professional reconstructionists referring to the morphological characteristics and anatomical structure of fish, and building models in computer-aided design software; the active 3D reconstruction method often completes the 3D reconstruction by performing 3D scanning of static fish bodies through laser radar, structured light, etc.; the passive 3D reconstruction method completes the 3D reconstruction by collecting static fish photos and feature extraction and matching algorithms.

[0003] The process of reconstructing fish bodies by manual reconstruction usually includes several main steps: data collection, model design, and model correction. ① First, collect the anatomical data and morphological characteristics of the fish, such as body proportions, scale arrangement, fin shape, etc. ② Then, use 3D reconstruction software such as Maya, 3ds Max, etc. to manually build a 3D model of the fish based on the collected data, including the process of building basic geometry and adding details. ③ Then, adjust and correct the material, texture, etc. of the model to ensure that the model is as consistent as possible with the actual fish morphology. ④ Finally, correct and optimize the model as needed to obtain more realistic and accurate results.

[0004] The process of reconstructing fish bodies using active 3D reconstruction usually includes several main steps: equipment configuration, scanning and acquisition, data processing, and model generation. ① First, configure an appropriate 3D scanner or other imaging device and adjust its parameters to suit the shape and size of the fish. ② Then, scan and collect the fish to obtain the geometric information of its surface through the device. ③ Next, process and analyze the collected data, including noise removal, alignment, point cloud reconstruction, and other steps to obtain accurate morphological information. ④ Finally, use the 3D reconstruction algorithm to convert the processed data into a 3D model of the fish, and optimize and correct it to obtain a more realistic and accurate result.

[0005] Traditional passive 3D reconstruction methods usually include several main steps: data acquisition, feature extraction, motion estimation, 3D reconstruction, and model optimization. ① First, use a camera or other equipment to shoot or record the fish to obtain a sequence of images in different postures or motion states. ② Then, computer vision technology is used to extract the morphological information of the fish, and its motion trajectory and posture changes are inferred. ③ Next, the 3D reconstruction algorithm is used to convert the extracted morphological information into a point cloud or mesh model in 3D space. ④ Finally, the generated 3D model is optimized and corrected to obtain more accurate and realistic results.

[0006] The above traditional methods face many challenges when reconstructing fish bodies in complex underwater environments. Manual reconstruction consumes a lot of manpower time and expertise, and it is difficult to meet efficiency requirements; active 3D reconstruction methods are expensive and require the object to be scanned to remain still, which cannot be applied to fish swimming in water; in particular, the uneven attenuation of light propagation in water causes image color distortion and reduced contrast, which affects the effect of traditional passive 3D reconstruction methods and leads to low reconstruction accuracy.

[0007] It should be noted that the information disclosed in the above background technology section is only used for understanding the background of the present application, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the invention

[0008] The main purpose of the present invention is to overcome the defects of the above-mentioned background technology and provide a method for three-dimensional reconstruction of underwater fish bodies in a smart fishery based on neural radiation fields, so as to solve the problem that the existing technology is difficult to efficiently and accurately reconstruct the three-dimensional model of fish bodies in complex underwater environments.

[0009] To achieve the above object, the present invention adopts the following technical solutions:

[0010] A method for three-dimensional reconstruction of underwater fish bodies in a smart fishery based on neural radiation fields comprises the following steps:

[0011] S1, obtaining visible light image data of individual fish in the fishing ground through a camera;

[0012] S2, calculating the internal and external parameters of the camera corresponding to the visible light image;

[0013] S3. Construct a neural radiation field model of underwater fish in smart fish farms;

[0014] S4, using the acquired visible light image data and the corresponding internal and external parameters of the camera as training data to train the underwater fish body neural radiation field model; wherein, in the training iteration process of the underwater fish body neural radiation field model, according to the predicted output of the neural network, the predicted image under the corresponding viewing angle is calculated by underwater attenuation volume rendering based on the underwater light propagation attenuation coefficient, and the loss is calculated, and the parameters of the neural network are iteratively updated;

[0015] S5. Use the trained underwater fish neural radiation field model to reconstruct the fish body in three dimensions based on the new visible light image data.

[0016] Further:

[0017] In step S2, feature point extraction, matching and three-dimensional point cloud construction are performed on the collected images, and the camera extrinsic parameters and intrinsic parameters corresponding to each image are calculated. The extrinsic parameters include the position and posture of the camera, and the intrinsic parameters include the focal length and distortion coefficient.

[0018] In step S3, a neural network containing multiple fully connected layers is constructed as an underwater fish neural radiation field model. The input is the spatial position of any sampling point in three-dimensional space and the observation direction from the camera to the point, and the output is the color and density of the sampling point.

[0019] In step S3, a neural radiation field model neural network with 10 fully connected layers is constructed, where the first 9 layers are 256-dimensional, the 10th layer is 128-dimensional, and the nonlinear activation function ReLU is used between layers.

[0020] In step S4, a gradient descent optimization algorithm is used, and the square error photometric loss function is used as the loss function to calculate the loss.

[0021] In step S4, according to the prediction output of the neural network, the prediction image under the corresponding viewing angle is calculated by underwater attenuation volume rendering based on the underwater light propagation attenuation coefficient, which specifically includes the following steps:

[0022] S41. Ray parameterization: determining the starting point of the ray and direction ;

[0023] S42, calculate sampling points: in the camera light Upper uniform sampling Sampling points ;

[0024] S43, obtaining color and density: obtaining each sampling point predicted and output by the underwater fish body nerve radiation field model Color and density ;

[0025] S44. Calculate the residual energy coefficient: For each sampling point , calculate its residual energy coefficient ;

[0026] S45. Calculate cumulative transparency: Calculate each sampling point Cumulative transparency ;

[0027] S46, volume rendering: sampling points predicted by the underwater fish body nerve radiation field model Color and density , and the calculated residual energy coefficient and cumulative transparency , calculate the final color of the light , and use the final color Generate a predicted image.

[0028] In step S46, the volume rendering calculation formula is:

[0029]

[0030] in, The light reaches The accumulated transparency before the sampling points, is the distance between adjacent sampling points.

[0031] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method for three-dimensional reconstruction of underwater fish bodies in a smart fishery based on neural radiation fields.

[0032] A computer program product includes a computer program, which, when executed by a processor, implements the method for three-dimensional reconstruction of underwater fish bodies in a smart fishery based on neural radiation fields.

[0033] The present invention has the following beneficial effects:

[0034] The present invention provides a method for three-dimensional reconstruction of underwater fish bodies in a smart fish farm based on neural radiation fields. By considering the attenuation effect of light in the process of underwater propagation, the underwater light propagation attenuation coefficient is introduced in the neural radiation field (NRRFs) volume rendering process to solve the influence of the underwater environment, thereby realizing efficient and high-precision three-dimensional reconstruction of swimming fish bodies in complex underwater environments. Specifically, the present invention constructs a neural radiation field neural network, and integrates an underwater attenuation volume rendering method based on the underwater light propagation attenuation coefficient into the training process of the neural radiation field model. During the neural network training process, the neural radiation field volume rendering method based on the underwater light propagation attenuation coefficient is used to convert the color and density predicted by the neural network into a complete predicted image, and the loss is calculated to guide the neural network to learn how to predict color and density, so as to train and optimize the neural network. As a result, the neural network learns how to predict the color and density of each sampling point according to the input spatial position and observation direction, so that the trained model can accurately predict the three-dimensional morphology and appearance of the swimming fish body underwater under new inputs, and can maintain high accuracy even under the condition of light attenuation.

[0035] Other beneficial effects of the embodiments of the present invention will be further described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of a method for three-dimensional reconstruction of underwater fish bodies in a smart fishery based on neural radiation fields according to an embodiment of the present invention.

[0037] Figure 2 This is a flow chart of drawing an underwater attenuation body based on the underwater light propagation attenuation coefficient according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following is a detailed description of the embodiments of the present invention. It should be emphasized that the following description is only exemplary and is not intended to limit the scope and application of the present invention.

[0039] See also Figure 1 The embodiment of the present invention provides a method for 3D reconstruction of underwater fish bodies in a smart fish farm based on neural radiation fields, comprising the following steps:

[0040] S1, obtaining visible light image data of individual fish in the fishing ground through a camera;

[0041] S2, calculating the internal and external parameters of the camera corresponding to the visible light image;

[0042] S3. Construct a neural radiation field model of underwater fish in smart fish farms;

[0043] S4, using the acquired visible light image data and the corresponding internal and external parameters of the camera as training data to train the underwater fish body neural radiation field model; wherein, in the training iteration process of the underwater fish body neural radiation field model, according to the predicted output of the neural network, the predicted image under the corresponding viewing angle is calculated by underwater attenuation volume rendering based on the underwater light propagation attenuation coefficient, and the loss is calculated, and the parameters of the neural network are iteratively updated by minimizing the error between the predicted image and the real image;

[0044] S5. Use the trained underwater fish neural radiation field model to reconstruct the fish body in three dimensions based on the new visible light image data.

[0045] See also Figure 2 In a preferred embodiment, in step S4, according to the prediction output of the neural network, the prediction image under the corresponding viewing angle is calculated by underwater attenuation volume rendering based on the underwater light propagation attenuation coefficient, which specifically includes the following steps:

[0046] S41. Ray parameterization: determining the starting point of the ray and direction ;

[0047] S42. Calculate sampling points: uniform sampling on the light Sampling points ;

[0048] S43, obtaining color and density: obtaining each sampling point predicted and output by the underwater fish body nerve radiation field model Color and density ;

[0049] S44. Calculate the residual energy coefficient: For each sampling point , calculate its residual energy coefficient ;

[0050] S45. Calculate cumulative transparency: Calculate each sampling point Cumulative transparency ;

[0051] S46, volume rendering: sampling points predicted by the underwater fish body nerve radiation field model Color and density , and the calculated residual energy coefficient and cumulative transparency , where the latter two are coefficients that take into account the attenuation effect of underwater light propagation. The final color of the light is calculated by underwater attenuation volume rendering. , and use the final color Generate a predicted image.

[0052] The embodiment of the present invention proposes a three-dimensional reconstruction method of underwater fish bodies in a smart fish farm based on neural radiation fields, constructs a neural radiation field model of underwater fish bodies in a smart fish farm, and in the process of neural network training, uses an underwater attenuation volume rendering method based on the underwater light propagation attenuation coefficient to convert the prediction of the neural network into a complete prediction image, and calculates the loss, so as to train and optimize the neural network. As a result, the neural network learns how to predict the color and density of each sampling point based on the input spatial position and observation direction, so that the trained model can accurately predict the three-dimensional shape and appearance of the swimming fish body underwater under new inputs, and maintain high accuracy even under the condition of light attenuation.

[0053] The specific embodiments of the present invention are further described below.

[0054] like Figure 1 As shown, a three-dimensional reconstruction method of underwater fish bodies in a smart fishery based on neural radiation fields mainly includes the following steps:

[0055] Step 1: Obtain visible light image data of individual underwater fish in the smart fishery.

[0056] Use a high-resolution underwater camera to collect multiple visible light images of the target fish body with overlapping areas from different angles and positions.

[0057] Step 2: Calculate the camera internal and external parameters corresponding to the visible light image.

[0058] Through algorithms such as SfM, the captured images are subjected to feature point extraction, matching, and 3D reconstruction, and the camera external parameters (position and posture) and internal parameters (focal length, distortion coefficient, etc.) corresponding to each image are calculated.

[0059] Step 3: Construct an underwater fish neural radiation field model for smart fisheries.

[0060] Construct a neural network with 10 fully connected layers, where the first 9 layers are 256-dimensional and the 10th layer is 128-dimensional. The nonlinear activation function ReLU (Rectified Linear Unit) is used between layers. The input of the neural network is the spatial position and viewing direction , the output of the neural network is the color of the sampling point and density .

[0061] Step 4: Use the underwater attenuation volume rendering method based on the underwater light propagation attenuation coefficient to train the underwater fish neural radiation field model of the smart fishery.

[0062] Construct a training data set, combining the image obtained in step 1 with the camera pose and parameter data (including external parameters: camera position and posture, and internal parameters: focal length, distortion coefficient, etc.) obtained in step 2 as the input data of the neural network.

[0063] During the iterative training process of the neural radiation field model of underwater fish, the predicted output of the neural radiation field model is used to calculate the image under the corresponding perspective through the underwater attenuation volume rendering method based on the underwater light propagation attenuation coefficient, so as to obtain the predicted image. Then, the gradient descent optimization algorithm is used, and the square error photometric loss function (Photometric LossFunction) is preferably used as the loss function to calculate the loss. The parameters of the neural network are iteratively updated by minimizing the error between the predicted image and the real image.

[0064] like Figure 2 As shown, in a preferred embodiment, the calculation process of the underwater attenuation volume rendering method based on the underwater light propagation attenuation coefficient specifically includes the following steps:

[0065] (1) Ray parameterization: determining the starting point of the ray and direction ;

[0066] (2) Calculate sampling points: in the camera light Upper uniform sampling Sampling points ;

[0067] (3) Get color and density: Get each sample point predicted and output by the neural radiation field model Color and density ;

[0068] (4) Calculate the residual energy coefficient: For each sampling point , calculate its residual energy coefficient ;

[0069] (5) Calculate the cumulative transparency: Calculate the cumulative transparency of each sampling point. Cumulative transparency ;

[0070] (6) Volume rendering: The final color of light is obtained through underwater attenuation volume rendering , and use the final color Generate a predicted image.

[0071] Camera Light The color of the dot on (where the camera position is , the observation direction is ) is calculated as:

[0072]

[0073] in Indicates the light is at position The residual energy coefficient at is the attenuation intensity coefficient of light in water; is the cumulative transmittance; is the density of points on the ray, The color of the points on the ray.

[0074] In order to implement it in a computer, the integral is discretized and the sampling points, each sampling point The color and density at and , the volume rendering calculation formula is:

[0075]

[0076] in The light reaches The accumulated transparency before the sampling points, is the distance between adjacent sampling points.

[0077] Step 5: Terminate the training process of the underwater fish neural radiation field model of the smart fishery, and use the trained fish neural radiation field model to complete the three-dimensional reconstruction of the fish body.

[0078] Set stop conditions (such as training time, number of iterations, etc.) during the training process. After training, save the final model parameters. Use the model to complete the 3D reconstruction of the fish body and generate a high-precision 3D model.

[0079] In summary, the present invention proposes a method for three-dimensional reconstruction of underwater fish bodies in a smart fish farm based on neural radiation fields, and uses an underwater attenuation volume rendering method to perform three-dimensional reconstruction of underwater fish bodies during neural radiation field training. Considering the attenuation effect of light during underwater propagation, the present invention introduces an underwater light propagation attenuation coefficient in the neural radiation field volume rendering process, which solves the problem of the influence of the underwater environment, thereby realizing efficient and high-precision three-dimensional reconstruction of swimming fish bodies in complex underwater environments. Specifically, the present invention constructs a neural radiation field neural network, and integrates an underwater attenuation volume rendering method based on the underwater light propagation attenuation coefficient into the training process of the neural radiation field model. During the neural network training process, the underwater attenuation volume rendering method based on the underwater light propagation attenuation coefficient is used to convert the color and density predicted by the neural network into a complete predicted image, and the loss is calculated to guide the neural network to learn how to predict color and density, so as to train and optimize the neural network. As a result, the neural network learns how to predict the color and density of each sampling point according to the input spatial position and observation direction, so that the trained model can accurately predict the three-dimensional morphology and appearance of the swimming fish body underwater under new inputs, and can maintain high accuracy even under the condition of light attenuation.

[0080] Compared with traditional methods, the method of the present invention adopts neural radiation field technology to perform three-dimensional reconstruction of underwater fish bodies, which can not only capture the details of the fish body surface and texture, but also the reconstructed three-dimensional model has higher resolution and detail restoration. In particular, by introducing the underwater light propagation attenuation coefficient in the neural radiation field volume rendering method, the present invention can effectively simulate the attenuation characteristics of light when propagating in water. This method can better adapt to the underwater environment and significantly improve the accuracy and realism of three-dimensional reconstruction of fish bodies in underwater environments.

[0081] An embodiment of the present invention further provides a storage medium for storing a computer program, which at least performs the above method when executed.

[0082] An embodiment of the present invention further provides a control device, comprising a processor and a storage medium for storing a computer program; wherein the processor is configured to execute at least the method described above when executing the computer program.

[0083] An embodiment of the present invention further provides a processor, wherein the processor executes a computer program and at least executes the method described above.

[0084] The storage medium can be implemented by any type of non-volatile storage device, or a combination thereof. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a ferromagnetic random access memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk memory or a tape memory. The storage medium described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0085] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0086] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0087] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0088] A person skilled in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, etc. Various media that can store program codes.

[0089] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0090] The methods disclosed in the several method embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0091] The features disclosed in several product embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0092] The features disclosed in several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0093] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art of the present invention, several equivalent substitutions or obvious variations can be made without departing from the concept of the present invention, and the performance or use is the same, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for 3D reconstruction of underwater fish bodies in a smart fishery based on neural radiation field, characterized in that: The steps include: S1, obtaining visible light image data of individual fish in the fishing ground through a camera; S2, calculating the internal and external parameters of the camera corresponding to the visible light image; S3. Construct a neural radiation field model of underwater fish in smart fisheries; S4, using the acquired visible light image data and the corresponding internal and external parameters of the camera as training data to train the underwater fish body neural radiation field model; wherein, in the training iteration process of the underwater fish body neural radiation field model, according to the predicted output of the neural network, the predicted image under the corresponding viewing angle is calculated by underwater attenuation volume rendering based on the underwater light propagation attenuation coefficient, and the loss is calculated, and the parameters of the neural network are iteratively updated; S5, using the trained underwater fish body neural radiation field model to perform three-dimensional reconstruction of the fish body based on the new visible light image data; In step S4, according to the prediction output of the neural network, the prediction image under the corresponding viewing angle is calculated by underwater attenuation volume rendering based on the underwater light propagation attenuation coefficient, which specifically includes the following steps: S41. Ray parameterization: Determine the starting point of the ray and direction ; S42, calculate sampling points: in the camera light Upper uniform sampling Sampling points ; S43, obtaining color and density: obtaining each sampling point predicted and output by the underwater fish body nerve radiation field model Color and density ; S44. Calculate the residual energy coefficient: For each sampling point , calculate its residual energy coefficient ; S45. Calculate cumulative transparency: Calculate each sampling point Cumulative transparency ; S46, volume rendering: sampling points predicted by the underwater fish body nerve radiation field model Color and density , and the calculated residual energy coefficient and cumulative transparency , calculate the final color of the light , and use the final color Generate a prediction image; In step S46, the volume rendering calculation formula is: ； in, The light reaches The accumulated transparency before the sampling points, is the distance between adjacent sampling points, Indicates that the light is in the camera light The residual energy coefficient at the corresponding position, is the attenuation intensity coefficient of light in water, is the camera position.

2. The method for 3D reconstruction of underwater fish bodies in a smart fishery based on neural radiation field according to claim 1, characterized in that: In step S2, feature point extraction, matching and three-dimensional point cloud construction are performed on the collected images, and the camera extrinsic parameters and intrinsic parameters corresponding to each image are calculated. The extrinsic parameters include the position and posture of the camera, and the intrinsic parameters include the focal length and distortion coefficient.

3. The method for 3D reconstruction of underwater fish bodies in a smart fishery based on neural radiation field according to claim 1 or 2, characterized in that: In step S3, a neural network containing multiple fully connected layers is constructed as an underwater fish neural radiation field model. The input is the spatial position of any sampling point in three-dimensional space and the observation direction from the camera to the point, and the output is the color and density of the sampling point.

4. The method for 3D reconstruction of underwater fish bodies in a smart fishery based on neural radiation field according to claim 3, characterized in that: In step S3, a neural radiation field model neural network with 10 fully connected layers is constructed, where the first 9 layers are 256-dimensional, the 10th layer is 128-dimensional, and the nonlinear activation function ReLU is used between layers.

5. The method for 3D reconstruction of underwater fish bodies in a smart fishery based on neural radiation field according to any one of claims 1 to 2, characterized in that: In step S4, a gradient descent optimization algorithm is used, and the square error photometric loss function is used as the loss function to calculate the loss.

6. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for three-dimensional reconstruction of underwater fish bodies in a smart fishery based on neural radiation fields is implemented as described in any one of claims 1 to 5.

7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for three-dimensional reconstruction of underwater fish bodies in a smart fishery based on neural radiation fields is implemented as described in any one of claims 1 to 5.

Images