Customization method of deep-sea fish trapping strategy based on partitioned visual data

Abstract

The present invention relates to a method for customizing a deep-sea fish trapping strategy based on partitioned visual data, comprising: utilizing an intelligent deep-sea fish prediction model corresponding to a target deep-sea water area to intelligently predict the number of fish appearing in each sub-sea area of ​​the target deep-sea water area in a future time segment after the current moment based on multiple copies of layer-by-layer sea area visual data corresponding to multiple past time segments before the current moment; and determining, based on the intelligent prediction results, a key trapping sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment. Through the present invention, it is possible to utilize an intelligent deep-sea fish prediction model designed for a customized structure of the current deep-sea water area to intelligently predict the number of fish appearing in each sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment, and then determine a matching deep-sea fish trapping strategy for the target deep-sea water area, thereby improving the effectiveness and efficiency of deep-sea fish trapping in different time segments.

Description

Technical Field

[0001] The present invention relates to the field of deep-sea fish trapping, and in particular to a method for customizing deep-sea fish trapping strategies based on partitioned visual data. Background Art

[0002] Deep-sea fish are highly prized for their white, delicate texture and authentic, tender, and refreshing taste. They are also a representative example of pollution-free, low-fat, high-protein, and green food. Rich in protein and low in fat, deep-sea fish are beneficial to human health. Their nutritional profile is more comprehensive and richer than that of regular fish. Currently, popular deep-sea fish such as horse mackerel, hairtail, yellow croaker, cod, and pomfret are all caught from deep-sea fishing. These fish mostly inhabit the deep sea, between 600 and 2,700 meters below sea level, far from land-based pollution and enjoying a uniquely favorable habitat. The deep-sea fishing process is influenced by numerous factors, including season, water temperature, and ocean currents, and can sometimes take months or even years. Deep-sea fishing is a complex process that requires overcoming numerous natural constraints and utilizing advanced technology and equipment. Therefore, deep-sea fish trapping is the standard method for catching deep-sea fish.

[0003] Among deep-sea fish trapping methods, traps or cages are a common method, using a single line or buoys to fish from the bottom. Typically, only one cage is placed on each line, or multiple cages are attached to a long hook line and marked with buoys. This method is particularly suitable for catching lobsters and most crabs. Another effective method is scooping, using a scoop net equipped with a metal rake that is dragged back and forth across the seabed. This method allows the scooped shellfish to be easily collected and bagged for later processing. Traps and scoops facilitate the collection of lobsters, crabs, and shellfish, increasing catch efficiency. This fishing method plays a vital role in marine fisheries.

[0004] However, no matter which deep-sea fish trapping method is adopted, the specific placement of the trapping device needs to be paid special attention. If the trapping device is placed exactly in the sub-water area with the most deep-sea fish in the deep-sea waters, the fish trapping effect of the trapping device will be twice as effective with half the effort. On the contrary, if the trapping device is placed exactly in the sub-water area with the least deep-sea fish in the deep-sea waters, the fish trapping effect of the trapping device will be half as effective with twice the effort. Therefore, how to determine the sub-water area with the most deep-sea fish in each deep-sea water area is the key to customizing deep-sea fish trapping strategies. Since the sub-water areas where deep-sea fish are active in deep-sea waters are flexible and changeable, and the water flow in different sub-water areas of deep-sea waters is also random and changeable, the location of the sub-water area with the least deep-sea fish in each time segment may be different, and it is difficult to determine the location of the sub-water area with the least deep-sea fish in each time segment. Summary of the Invention

[0005] In order to solve the technical problems in related fields, the present invention provides a method for customizing deep-sea fish trapping strategies based on partitioned visual data. By adopting a deep-sea fish intelligent prediction model designed for the current deep-sea water area customization structure, before the arrival of the future time segment after the current moment, the method intelligently predicts the number of fish appearing in each sub-sea area in the target deep-sea water area in the future time segment after the current moment according to multiple layers of sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area in the target deep-sea water area, the number of deep-sea fish species existing in the target deep-sea water area, the shallowest depth and the deepest depth of the target deep-sea water area, and the occupancy time of each time segment, thereby providing valuable reference information for the subsequent customization of deep-sea fish trapping strategies for the target deep-sea water area, and ensuring the trapping effect and trapping efficiency of different deep-sea fish trapping strategies in different time segments.

[0006] According to the present invention, a method for customizing a deep-sea fish trapping strategy based on partitioned visual data is provided, the method comprising:

[0007] Each of the multiple past time segments before the current moment is used as a target time segment, and each deep-sea waterproof camera mechanism is placed at the top position of each sub-sea area arranged at equal depths from top to bottom in the target deep-sea water area in each target time segment, so as to perform visual data acquisition with the same resolution for each sub-sea area, so as to obtain visual images of each sub-sea area corresponding to the target time segment, and use the L channel value, A channel value, B channel value, depth of field value, vertical coordinate value and horizontal coordinate value of each pixel point of each sub-sea area visual image and the number of fish bodies appearing in each sub-sea area visual image as associated visual information of the sub-sea area visual image, and output the associated visual information of each sub-sea area visual image corresponding to each target time segment as the layer-by-layer sea area visual data corresponding to the target time segment;

[0008] An intelligent prediction model for deep-sea fish corresponding to the target deep-sea waters is used to intelligently predict the number of fish appearing in each sub-sea area in the target deep-sea waters in future time segments after the current moment based on multiple layers of sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species present in the target deep-sea waters, the shallowest and deepest depths of the target deep-sea waters, and the occupancy duration of each time segment;

[0009] The sub-sea area with the largest number of fish bodies appearing in the future time segment after the current moment is output as the key trapping sub-sea area of ​​the target deep sea water area in the future time segment after the current moment.

[0010] It can be seen that the present invention has at least the following four outstanding substantive features:

[0011] Substantive Feature A: Before the arrival of a future time segment after the current moment, the system intelligently predicts the number of fish present in each sub-sea area in the target deep-sea area in the future time segment after the current moment based on multiple layers of sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area in the target deep-sea area, the number of deep-sea fish species present in the target deep-sea area, the shallowest and deepest depths of the target deep-sea area, and the occupancy duration of each time segment, thereby providing valuable reference information for the subsequent customization of deep-sea fish trapping strategies for the target deep-sea area;

[0012] Substantive Feature B: Outputting the sub-sea area with the largest number of fish in the future time segment after the current moment as the key trapping sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment, and burying a deep-sea fish trapping device in the key trapping sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment at the current moment, wherein the deep-sea fish trapping device is buried at a water level at an intermediate depth of the key trapping sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment, thereby completing the execution of the deep-sea fish trapping strategy based on the partitioned visual data;

[0013] Substantive Feature C: Each of the multiple past time segments before the current moment is used as a target time segment, and each deep-sea waterproof camera unit is placed at the top position of each sub-sea area arranged at equal depths from top to bottom in the target deep-sea water area in each target time segment, so as to perform visual data acquisition with the same resolution for each sub-sea area respectively, so as to obtain each sub-sea area visual picture corresponding to the target time segment respectively, and use the L channel value, A channel value, B channel value, depth of field value, vertical coordinate value and horizontal coordinate value of each pixel point of each sub-sea area visual picture and the number of fish bodies appearing in each sub-sea area visual picture as the associated visual information of the sub-sea area visual picture, and use the associated visual information of each sub-sea area visual picture corresponding to each target time segment as the layer-by-layer sea area visual data corresponding to the target time segment, thereby completing the targeted acquisition of the most critical basic data for intelligent prediction;

[0014] Substantive Feature D: It is an intelligent prediction of the number of fish species appearing in each sub-sea area in the target deep-sea water area in the future time segment after the current moment. A deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is designed with a customized structure. The deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is a deep neural network that has been trained multiple times. The deep neural network includes multiple hidden layers, an output layer and an input layer. The number of times the deep neural network has been trained is positively correlated with the number of deep-sea fish species existing in the target deep-sea water area, and the number of hidden layers in the deep neural network is proportional to the depth of the target deep-sea water area, thereby customizing deep-sea fish intelligent prediction models with different structures for different deep-sea water areas. DETAILED DESCRIPTION

[0015] The following is a detailed description of an embodiment of the method for customizing deep-sea fish trapping strategies based on partitioned visual data of the present invention.

[0016] Embodiment 1 of the present invention

[0017] The method for customizing a deep-sea fish trapping strategy based on partitioned visual data according to the first embodiment of the present invention specifically includes the following steps:

[0018] Each of the multiple past time segments before the current moment is used as a target time segment, and each deep-sea waterproof camera mechanism is placed at the top position of each sub-sea area arranged at equal depths from top to bottom in the target deep-sea water area in each target time segment, so as to perform visual data acquisition with the same resolution for each sub-sea area, so as to obtain visual images of each sub-sea area corresponding to the target time segment, and use the L channel value, A channel value, B channel value, depth of field value, vertical coordinate value and horizontal coordinate value of each pixel point of each sub-sea area visual image and the number of fish bodies appearing in each sub-sea area visual image as associated visual information of the sub-sea area visual image, and output the associated visual information of each sub-sea area visual image corresponding to each target time segment as the layer-by-layer sea area visual data corresponding to the target time segment;

[0019] For example, the target deep sea area may be waters from 600 meters to 2700 meters below sea level, that is, the depth of the target deep sea area is 2100 meters, and the duration of each time segment may be selected as 45 minutes;

[0020] An intelligent prediction model for deep-sea fish corresponding to the target deep-sea waters is used to intelligently predict the number of fish appearing in each sub-sea area in the target deep-sea waters in future time segments after the current moment based on multiple layers of sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species present in the target deep-sea waters, the shallowest and deepest depths of the target deep-sea waters, and the occupancy duration of each time segment;

[0021] Specifically, when the target deep-sea area is waters ranging from 600 meters to 2700 meters below sea level, the shallowest depth and the deepest depth of the target deep-sea area are 600 meters and 2700 meters respectively;

[0022] Outputting the sub-sea area with the largest number of fish bodies in the future time segment after the current moment as the key trapping sub-sea area of ​​the target deep sea water area in the future time segment after the current moment;

[0023] The deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is a deep neural network that has been trained multiple times. The deep neural network includes multiple hidden layers, an output layer, and an input layer. The number of times the deep neural network has been trained is positively correlated with the number of deep-sea fish species present in the target deep-sea water area.

[0024] For example, the positive correlation between the number of times the deep neural network has been trained and the number of deep-sea fish species present in the target deep-sea waters includes: when the number of deep-sea fish species present in the target deep-sea waters is 10, the number of times the deep neural network has been trained is 600 times; when the number of deep-sea fish species present in the target deep-sea waters is 15, the number of times the deep neural network has been trained is 700 times; when the number of deep-sea fish species present in the target deep-sea waters is 20, the number of times the deep neural network has been trained is 800 times, and so on;

[0025] Among them, the deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is a deep neural network that has been trained multiple times. The deep neural network includes multiple hidden layers, an output layer and an input layer. The number of times the deep neural network has been trained is positively correlated with the number of deep-sea fish species existing in the target deep-sea water area. It also includes: the number of hidden layers in the deep neural network is proportional to the depth of the target deep-sea water area.

[0026] Embodiment 2 of the present invention

[0027] Compared with the first embodiment of the present invention, the method for customizing a deep-sea fish trapping strategy based on partitioned visual data according to the second embodiment of the present invention further includes the following steps:

[0028] burying deep-sea fish trapping devices in a key trapping sub-sea area of ​​the target deep-sea waters in a future time segment after the current moment;

[0029] Among them, burying the deep-sea fish trapping device at the current moment in the key trapping sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment includes: burying the deep-sea fish trapping device at the middle depth of the key trapping sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment.

[0030] Embodiment 3 of the present invention

[0031] Compared with the first embodiment of the present invention, the method for customizing a deep-sea fish trapping strategy based on partitioned visual data shown in the third embodiment of the present invention further includes the following steps:

[0032] receiving the sub-sea area numbers corresponding to the key trapping sub-sea areas in the target deep-sea waters in future time segments after the current moment, and wirelessly transmitting the sub-sea area numbers corresponding to the key trapping sub-sea areas in the target deep-sea waters in future time segments after the current moment to a sea area monitoring server in a deep-sea fishing vessel on the coast via a wireless communication link;

[0033] For example, receiving the sub-sea area number corresponding to the key trapping sub-sea area in the future time segment of the target deep-sea water area after the current moment, and wirelessly transmitting the sub-sea area number corresponding to the key trapping sub-sea area in the future time segment of the target deep-sea water area after the current moment to the sea area monitoring server in the deep-sea fishing vessel on the coast through a wireless communication link includes: the sea area monitoring server in the deep-sea fishing vessel on the coast is a cloud computing monitoring node or a big data monitoring node.

[0034] Next, the specific steps of the method for customizing a deep-sea fish trapping strategy based on partitioned visual data of the present invention will be further described.

[0035] In the method for customizing deep-sea fish trapping strategies based on partitioned visual data according to any embodiment of the present invention:

[0036] The deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is a deep neural network that has been trained multiple times, the deep neural network including multiple hidden layers, an output layer, and an input layer. The number of times the deep neural network has been trained is positively correlated with the number of deep-sea fish species present in the target deep-sea water area. The method further includes: using a frequency analytical function to represent a numerical conversion relationship of the positive correlation between the number of times the deep neural network has been trained and the number of deep-sea fish species present in the target deep-sea water area;

[0037] For example, using a frequency analytical function to represent the numerical conversion relationship of the positive correlation between the number of times the deep neural network has been trained and the number of deep-sea fish species present in the target deep-sea waters includes: programmable logic devices can be selected to implement simulation and simulation of the frequency analytical function.

[0038] And in the method for customizing deep-sea fish trapping strategies based on partitioned visual data according to any embodiment of the present invention:

[0039] Each of the multiple past time segments before the current moment is used as a target time segment, and each deep-sea waterproof camera mechanism is placed at the top position of each sub-sea area arranged at equal depths from top to bottom in the target deep-sea water area in each target time segment, so as to perform visual data acquisition with the same resolution for each sub-sea area respectively, so as to obtain each sub-sea area visual picture corresponding to the target time segment respectively, and use the L channel value, A channel value, B channel value, depth of field value, vertical coordinate value and horizontal coordinate value of each pixel point of each sub-sea area visual picture and the number of fish bodies appearing in each sub-sea area visual picture as the associated visual information of the sub-sea area visual picture, and use the associated visual information of each sub-sea area visual picture corresponding to each target time segment as the layer-by-layer sea area visual data output corresponding to the target time segment, including: performing recognition of fish bodies appearing in each sub-sea area visual picture based on a baseline contour pattern of deep-sea fish bodies;

[0040] wherein, each of the multiple past time segments before the current moment is taken as a target time segment, and each deep-sea waterproof camera mechanism is placed at the top position of each sub-sea area arranged at equal depths from top to bottom in the target deep-sea water area in each target time segment, so as to perform visual data acquisition with the same resolution for each sub-sea area respectively, so as to obtain each sub-sea area visual picture corresponding to the target time segment respectively, and take the L channel value, A channel value, B channel value, depth of field value, vertical coordinate value and horizontal coordinate value of each pixel point of each sub-sea area visual picture and the number of fish bodies appearing in each sub-sea area visual picture as the associated visual information of the sub-sea area visual picture, and take the associated visual information of each sub-sea area visual picture corresponding to each target time segment as the layer-by-layer sea area visual data output corresponding to the target time segment, further comprising: the L channel value, A channel value and B channel value of each pixel point are the L channel value, A channel value and B channel value of the pixel point in the LAB color space;

[0041] Among them, the L channel value, A channel value, and B channel value of each pixel point are the L channel value, A channel value, and B channel value of the pixel point in the LAB color space, including: the value range of any channel value among the L channel value, A channel value, and B channel value of each pixel point is between 0-255.

[0042] And in the method for customizing deep-sea fish trapping strategies based on partitioned visual data according to any embodiment of the present invention:

[0043] An intelligent prediction model for deep-sea fish corresponding to the target deep-sea waters is used to intelligently predict the number of fish appearing in each sub-sea area in the target deep-sea waters in future time segments after the current moment based on multiple layers of sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species present in the target deep-sea waters, the shallowest depth and the deepest depth of the target deep-sea waters, and the occupancy time of each time segment. The depth of each water area is the difference between the shallowest depth and the deepest depth of the water area.

[0044] wherein, using the deep-sea fish intelligent prediction model corresponding to the target deep-sea waters to intelligently predict the number of each fish body appearing in each sub-sea area in the future time segment after the current moment based on multiple copies of layer-by-layer sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species existing in the target deep-sea waters, the shallowest depth and the deepest depth of the target deep-sea waters, and the occupancy time of each time segment also includes: synchronously inputting the multiple copies of layer-by-layer sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species existing in the target deep-sea waters, the shallowest depth and the deepest depth of the target deep-sea waters, and the occupancy time of each time segment into the deep-sea fish intelligent prediction model corresponding to the target deep-sea waters;

[0045] And wherein, the deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is used to intelligently predict the number of each fish body appearing in each sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment based on multiple layers of sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species existing in the target deep-sea water area, the shallowest depth and the deepest depth of the target deep-sea water area, and the occupancy time of each time segment, which also includes: executing the deep-sea fish intelligent prediction model corresponding to the target deep-sea water area to obtain the number of each fish body appearing in each sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment output by the deep-sea fish intelligent prediction model corresponding to the target deep-sea water area.

[0046] In addition, in the method for customizing deep-sea fish trapping strategies based on partitioned visual data, the deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is a deep neural network that has been trained multiple times, and the deep neural network includes multiple hidden layers, an output layer and an input layer. The number of times the deep neural network has been trained is positively correlated with the number of deep-sea fish species present in the target deep-sea water area, and also includes: in the deep neural network, the multiple hidden layers are located between the output layer and the input layer.

[0047] The method for customizing deep-sea fish trapping strategies based on partitioned visual data of the present invention is used to address the technical problem in the prior art that it is difficult to dynamically determine the sub-sea areas with the most fish appearing at different times in different deep-sea waters, resulting in the deep-sea fish trapping effect and efficiency failing to meet expectations. By adopting a deep-sea fish intelligent prediction model designed for current deep-sea waters with a customized structure, the number of fish appearing in each sub-sea area in the future time segments of the target deep-sea water area is intelligently predicted, and then the deep-sea fish trapping strategy of the matching target deep-sea water area is determined, thereby improving the deep-sea fish trapping effect and efficiency in different time segments and solving the above-mentioned technical problems.

[0048] Although some embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the claims and their equivalents.

Claims

1. A method for customizing deep-sea fish trapping strategies based on partitioned visual data, characterized in that: The method comprises: Each of the multiple past time segments before the current moment is used as a target time segment, and each deep-sea waterproof camera mechanism is placed at the top position of each sub-sea area arranged at equal depths from top to bottom in the target deep-sea water area in each target time segment, so as to perform visual data acquisition with the same resolution for each sub-sea area, so as to obtain visual images of each sub-sea area corresponding to the target time segment, and use the L channel value, A channel value, B channel value, depth of field value, vertical coordinate value and horizontal coordinate value of each pixel point of each sub-sea area visual image and the number of fish bodies appearing in each sub-sea area visual image as associated visual information of the sub-sea area visual image, and output the associated visual information of each sub-sea area visual image corresponding to each target time segment as the layer-by-layer sea area visual data corresponding to the target time segment; An intelligent prediction model for deep-sea fish corresponding to the target deep-sea waters is used to intelligently predict the number of fish appearing in each sub-sea area in the target deep-sea waters in future time segments after the current moment based on multiple layers of sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species present in the target deep-sea waters, the shallowest and deepest depths of the target deep-sea waters, and the occupancy duration of each time segment; The sub-sea area with the largest number of fish bodies appearing in the future time segment after the current moment is output as the key trapping sub-sea area of ​​the target deep sea water area in the future time segment after the current moment.

2. The method for customizing deep-sea fish trapping strategies based on partitioned visual data according to claim 1, wherein: The deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is a deep neural network that has been trained multiple times. The deep neural network includes multiple hidden layers, an output layer, and an input layer. The number of times the deep neural network has been trained is positively correlated with the number of deep-sea fish species present in the target deep-sea water area. Among them, the deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is a deep neural network that has been trained multiple times. The deep neural network includes multiple hidden layers, an output layer and an input layer. The number of times the deep neural network has been trained is positively correlated with the number of deep-sea fish species existing in the target deep-sea water area. It also includes: the number of hidden layers in the deep neural network is proportional to the depth of the target deep-sea water area.

3. The method for customizing deep-sea fish trapping strategies based on partitioned visual data according to claim 2, wherein: The method further comprises: burying deep-sea fish trapping devices in a key trapping sub-sea area of ​​the target deep-sea waters in a future time segment after the current moment; Among them, burying the deep-sea fish trapping device at the current moment in the key trapping sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment includes: burying the deep-sea fish trapping device at the middle depth of the key trapping sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment.

4. The method for customizing deep-sea fish trapping strategies based on partitioned visual data according to claim 2, wherein: The method further comprises: Receive the sub-sea area number corresponding to the key trapping sub-sea area in the future time segment of the target deep-sea water area after the current moment, and wirelessly transmit the sub-sea area number corresponding to the key trapping sub-sea area in the future time segment of the target deep-sea water area after the current moment to the sea area monitoring server in the deep-sea fishing vessel on the coast through a wireless communication link.

5. The method for customizing deep-sea fish trapping strategies based on partitioned visual data according to any one of claims 2 to 4, characterized in that: The deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is a deep neural network that has been trained multiple times, and the deep neural network includes multiple hidden layers, an output layer and an input layer. The number of times the deep neural network has been trained is positively correlated with the number of deep-sea fish species existing in the target deep-sea water area, and further includes: using a frequency analytical function to represent the numerical conversion relationship of the positive correlation between the number of times the deep neural network has been trained and the number of deep-sea fish species existing in the target deep-sea water area.

6. The method for customizing deep-sea fish trapping strategies based on partitioned visual data according to any one of claims 2 to 4, characterized in that: Each of the multiple past time segments before the current moment is used as a target time segment, and each deep-sea waterproof camera mechanism is placed at the top position of each sub-sea area arranged at equal depths from top to bottom in the target deep-sea water area in each target time segment, so as to perform visual data acquisition with the same resolution for each sub-sea area respectively, so as to obtain each sub-sea area visual picture corresponding to the target time segment respectively, and use the L channel value, A channel value, B channel value, depth of field value, vertical coordinate value and horizontal coordinate value of each pixel point of each sub-sea area visual picture and the number of fish bodies appearing in each sub-sea area visual picture as the associated visual information of the sub-sea area visual picture, and use the associated visual information of each sub-sea area visual picture corresponding to each target time segment as the layer-by-layer sea area visual data output corresponding to the target time segment, including: performing recognition of fish bodies appearing in each sub-sea area visual picture based on the baseline contour pattern of the deep-sea fish body.

7. The method for customizing deep-sea fish trapping strategies based on zoned visual data according to claim 6, wherein: Each of the multiple past time segments before the current moment is used as a target time segment, and each deep-sea waterproof camera mechanism is placed at the top position of each sub-sea area arranged at equal depths from top to bottom in the target deep-sea water area in each target time segment, so as to perform visual data acquisition with the same resolution for each sub-sea area respectively, so as to obtain each sub-sea area visual picture corresponding to the target time segment respectively, and use the L channel value, A channel value, B channel value, depth of field value, vertical coordinate value and horizontal coordinate value of each pixel point of each sub-sea area visual picture and the number of fish bodies appearing in each sub-sea area visual picture as the associated visual information of the sub-sea area visual picture, and use the associated visual information of each sub-sea area visual picture corresponding to each target time segment as the layer-by-layer sea area visual data output corresponding to the target time segment, further comprising: the L channel value, A channel value and B channel value of each pixel point are the L channel value, A channel value and B channel value of the pixel point in the LAB color space.

8. The method for customizing deep-sea fish trapping strategies based on zoned visual data according to claim 7, wherein: The L channel value, A channel value, and B channel value of each pixel point are the L channel value, A channel value, and B channel value of the pixel point in the LAB color space, including: the value range of any channel value among the L channel value, A channel value, and B channel value of each pixel point is between 0-255.

9. The method for customizing deep-sea fish trapping strategies based on partitioned visual data according to any one of claims 2 to 4, characterized in that: An intelligent prediction model for deep-sea fish corresponding to the target deep-sea waters is used to intelligently predict the number of fish appearing in each sub-sea area in the target deep-sea waters in future time segments after the current moment based on multiple layers of sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species present in the target deep-sea waters, the shallowest depth and the deepest depth of the target deep-sea waters, and the occupancy time of each time segment. The depth of each water area is the difference between the shallowest depth and the deepest depth of the water area. wherein, using the deep-sea fish intelligent prediction model corresponding to the target deep-sea waters to intelligently predict the number of each fish body appearing in each sub-sea area in the future time segment after the current moment based on multiple copies of layer-by-layer sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species existing in the target deep-sea waters, the shallowest depth and the deepest depth of the target deep-sea waters, and the occupancy time of each time segment also includes: synchronously inputting the multiple copies of layer-by-layer sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species existing in the target deep-sea waters, the shallowest depth and the deepest depth of the target deep-sea waters, and the occupancy time of each time segment into the deep-sea fish intelligent prediction model corresponding to the target deep-sea waters; Among them, the deep-sea fish intelligent prediction model corresponding to the target deep-sea water area is used to intelligently predict the number of each fish body appearing in each sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment based on multiple layers of sea area visual data corresponding to multiple past time segments before the current moment, the depth of each sub-sea area, the number of deep-sea fish species existing in the target deep-sea water area, the shallowest depth and the deepest depth of the target deep-sea water area, and the occupancy time of each time segment. It also includes: executing the deep-sea fish intelligent prediction model corresponding to the target deep-sea water area to obtain the number of each fish body appearing in each sub-sea area of ​​the target deep-sea water area in the future time segment after the current moment output by the deep-sea fish intelligent prediction model corresponding to the target deep-sea water area.