Underwater video robot data acquisition method and device

Abstract

The present invention discloses a data acquisition method and device for an underwater video robot, relating to the technical field of video surveillance processing equipment. The method comprises: installing a video acquisition device and multiple sensors on the underwater video robot; acquiring video data and environmental data, as well as corresponding time information, in real time; performing target detection on the video data using a convolutional neural network (CNN) and calculating a first confidence level for the target detection result; calculating a first comprehensive environmental score at the current moment based on the environmental data; calculating the data priority at the current moment based on the first confidence level and the first comprehensive environmental score; and transmitting the corresponding data to a ground control station based on the data priority. The present invention has the following advantages: real-time target detection and classification, improved data acquisition efficiency, dynamic adjustment of weight coefficients, and adaptable data transmission.

Description

Technical Field

[0001] The present invention belongs to the technical field of video monitoring and processing equipment, and in particular relates to a method and device for collecting data of an underwater video robot. Background Art

[0002] With the advancement of underwater exploration technology, underwater robots (AUVs) are increasingly being used in fields such as ocean research, environmental monitoring, resource exploration, and underwater construction. These fields place higher demands on the performance and intelligence of AUVs, especially in data acquisition and processing. Traditional underwater data acquisition methods rely heavily on manual operation, resulting in low efficiency, high risks, and difficulty processing large amounts of data in real time. Therefore, developing AUVs capable of automated and efficient data acquisition has become a current research hotspot.

[0003] Underwater robots are typically equipped with a variety of sensors, such as high-definition cameras, sonar, temperature sensors, salinity sensors, and turbidity sensors. These sensors provide information about the underwater environment. However, effectively integrating data from these different sensors and processing and analyzing them in real time for rapid decision-making remains a technical challenge. In complex underwater environments, factors such as insufficient light, high flow, and fluctuating water turbidity can affect sensor performance, reducing the accuracy of data collection.

[0004] When transmitting data, underwater robots must consider network bandwidth and latency. Transmitting large amounts of data in real time can lead to information overload, impacting system response time. Therefore, optimizing data transmission strategies based on data importance and priority is crucial for improving underwater data collection efficiency. Summary of the Invention

[0005] In view of the above-mentioned defects in the prior art, the present invention provides a method for collecting data of an underwater video robot, comprising the following steps:

[0006] Step S101: Install a video acquisition device and multiple sensors on the underwater video robot;

[0007] Step S103: real-time acquisition of video data and environmental data, as well as corresponding time information, wherein the environmental data includes turbidity data, temperature, and salinity;

[0008] Step S105: performing target detection on the video data using a convolutional neural network (CNN), and calculating a first confidence level of the target detection result;

[0009] Step S107: Calculate a first comprehensive environment score at the current moment based on the environment data;

[0010] Step S109: Calculate the data priority at the current moment based on the first confidence level and the first comprehensive environment score;

[0011] Step S1011: Transmitting corresponding data to a ground control station based on the data priority.

[0012] Wherein, the step S105 includes the following steps:

[0013] Step S1051: Calculate the classification probability of targets of different categories;

[0014] Step S1053: Calculate the loss between the predicted box and the true box using the smooth L1 function;

[0015] Step S1055: Calculate the first confidence level of the target detection result.

[0016] Wherein, the step S1051 includes: assuming there are 𝑁 categories, the CNN outputs the classification probability 𝑃(𝑐 𝑖 ), where 𝑐 𝑖 Represents the 𝑖th category, and the Softmax function is used to calculate the classification probability of each category:

[0017] , 1≤j≤N, j is an integer, 𝑧 𝑖 It is the logits output by the last layer of the CNN.

[0018] Wherein, the step S1053 includes:

[0019] Use the smooth L1 loss function to calculate the loss between the predicted box and the true box: hour, ,otherwise, , where B pred is the predicted bounding box, 𝐵 𝑡𝑟𝑢𝑒 is the true bounding box;

[0020] Calculate the box regression results, .

[0021] Wherein, the step S1055 includes: based on the classification probability and the box regression result, the first confidence calculation formula is: .

[0022] Wherein, the step S107 includes: calculating the first comprehensive environment score by normalization processing :

[0023] ;

[0024] Among them, W t 、Ws and W c is the weight coefficient of temperature, salinity and turbidity in the environmental score, T represents the temperature of the underwater environment, T min Indicates the minimum value of underwater ambient temperature, T max represents the maximum temperature of the underwater environment, S represents the salinity of the underwater environment, S min Indicates the minimum value of underwater environmental salinity, S max represents the maximum salinity of the underwater environment, C represents the turbidity of the underwater environment, and C min Indicates the minimum value of underwater turbidity, C max Indicates the maximum turbidity value of the underwater environment.

[0025] The step S109 includes calculating the value of the data priority using the following formula:

[0026] ,in Represents the dynamic weight coefficient of video data, represents the dynamic weight coefficient of sensor data, Indicates the first confidence of the target detection result, Indicates the first comprehensive environment score.

[0027] The dynamic weight coefficient is calculated by the first comprehensive environment score and the first confidence level of target detection, specifically:

[0028] ;

[0029] ;

[0030] Among them, W target and W environment The weight coefficient for the first confidence and first comprehensive environment score of target detection.

[0031] Wherein, the step S1011 includes setting the data priority The data is compared with a first threshold, and data corresponding to a data priority greater than the first threshold is transmitted to a ground control station.

[0032] In addition, the present invention also proposes an underwater video robot data acquisition device, comprising

[0033] A video acquisition module, which is used to obtain video data of the underwater environment in real time;

[0034] Turbidity sensor, which is used to obtain the turbidity of the underwater environment in real time;

[0035] A temperature sensor is used to obtain the temperature of the underwater environment in real time;

[0036] Salinity sensor, which is used to obtain the salinity of the underwater environment in real time;

[0037] A target detection result module is used to perform target detection on the video data using a convolutional neural network (CNN) and calculate a first confidence level of the target detection result;

[0038] An environment score calculation module, configured to calculate a first comprehensive environment score at a current moment based on the environment data;

[0039] a priority calculation module, configured to calculate the data priority at a current moment based on the first confidence level and the first comprehensive environment score;

[0040] A data transmission module is used to transmit corresponding data to a ground control station based on the data priority.

[0041] Compared with the prior art, the present invention has the following advantages:

[0042] By using a convolutional neural network (CNN), the present invention can identify and classify underwater organisms, obstacles, and other environmental features in real time within complex underwater environments. This real-time capability ensures the robot can quickly respond to environmental changes, improving the efficiency and accuracy of data collection.

[0043] A comprehensive evaluation mechanism combining target detection results and environmental scores prioritizes the transmission of important data and reduces unnecessary data redundancy. This prioritization strategy ensures the timely transmission of critical data and optimizes the data collection process.

[0044] By using dynamic weight coefficients, the present invention can flexibly optimize data transmission strategies based on real-time environmental changes and target detection results. This adaptability makes the system more stable and reliable in different underwater environments.

[0045] This method integrates multiple environmental data points, such as temperature, salinity, and turbidity, and generates an environmental score through normalization. This multi-dimensional data analysis capability enables the robot to more comprehensively understand and adapt to the underwater environment, improving the system's intelligence.

[0046] By combining sonar data with environmental characteristics, the present invention enables the underwater robot to have the ability to self-navigate and automatically avoid obstacles. This feature not only improves the robot's safety but also enhances its ability to operate in complex underwater environments.

[0047] The dynamic data transmission module adjusts the data transmission rate based on real-time network conditions and data priority to ensure the timely transmission of important data. This adaptive mechanism improves the reliability and effectiveness of data transmission and reduces the risk of information loss.

[0048] By uploading data to a cloud computing platform for analysis and storage, and leveraging big data technology to mine historical data, this invention can provide even more powerful support for future underwater data collection and analysis. This capability enables the system to move beyond real-time operations and also enable long-term data trend analysis.

[0049] Automated data collection and processing reduces reliance on manual operations, minimizing the errors and risks that can result from human intervention. This makes underwater robots more reliable during long-term operations and in harsh environments.

[0050] Since the invention can monitor underwater organisms and environmental changes in real time, it provides an effective technical means for ecological protection and environmental monitoring, which has important practical significance for the research and protection of marine ecosystems. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present disclosure are shown in an illustrative and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0052] Figure 1 The figure shows a flow chart of a method for collecting data of an underwater video robot according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] To make the objectives, technical solutions, and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings. It is apparent that the embodiments described are only some, not all, of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0054] The terms used in the embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a," "an," "the," and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two.

[0055] It should be understood that although the terms "first," "second," "third," etc. may be used to describe "...," these "..." should not be limited to these terms. These terms are merely used to distinguish "...." For example, "first..." could also be referred to as "second...", and similarly, "second..." could also be referred to as "first..." without departing from the scope of the present invention.

[0056] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0057] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0058] It should also be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or device. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the product or device comprising the element.

[0059] The optional embodiments of the present invention are described in detail below with reference to the accompanying drawings. Example

[0060] like Figure 1 As shown, the present invention discloses a method for collecting and evaluating underwater video robot data, comprising the following steps:

[0061] Step S101: Install a video acquisition device and multiple sensors on the underwater video robot;

[0062] Step S103: real-time acquisition of video data and environmental data, as well as corresponding time information, wherein the environmental data includes turbidity data, temperature, and salinity;

[0063] Step S105: performing target detection on the video data using a convolutional neural network (CNN), and calculating a first confidence level of the target detection result;

[0064] Step S107: Calculate a first comprehensive environment score at the current moment based on the environment data;

[0065] Step S109: Calculate the data priority at the current moment based on the first confidence level and the first comprehensive environment score;

[0066] Step S1011: Transmitting corresponding data to a ground control station based on the data priority.

[0067] Step S105 includes but is not limited to the following steps.

[0068] Before feeding the video frames into the CNN, the images are first preprocessed, including: denoising, which uses filters to reduce noise in underwater environments; normalization, which standardizes the image pixel values ​​to a specific range to improve the convergence speed of the model; and data augmentation, which rotates, scales, flips, and other operations on the images to increase the diversity of training samples.

[0069] In one embodiment, a convolutional neural network (CNN) is typically composed of multiple layers, including:

[0070] The convolution layer extracts image features and generates feature maps through convolution operations. These features can be low-level features such as edges and textures, which can be further combined to form more complex features.

[0071] Activation layer, which uses nonlinear activation functions (such as ReLU) to enhance the expressiveness of the model.

[0072] The pooling layer reduces the dimension of the feature map, retains important information, and reduces the amount of computation and the risk of overfitting.

[0073] The fully connected layer maps high-dimensional features to classification labels and outputs the final classification results.

[0074] In this paper, CNN performs object detection through a specific architecture (such as YOLO and Faster R-CNN), which can simultaneously identify multiple objects and generate bounding boxes. Each bounding box corresponds to a detected object and is accompanied by a corresponding classification label.

[0075] CNN generates a confidence score C for each detected object target , which represents the probability that the target belongs to a specific category. The calculation of confidence is usually based on the following factors:

[0076] Classification probability, the probability value of each category output by the model; box regression, that is, the accuracy of the bounding box, ensures that the target is correctly located.

[0077] Confidence values ​​typically range from 0 to 1, with higher values ​​indicating greater confidence in the model's recognition of the target.

[0078] Finally, the CNN outputs the identified target category, bounding box coordinates, and its confidence score (C_target). The robot uses this information to determine the important objects in the current environment and optimize subsequent data collection and transmission strategies.

[0079] Categories include: underwater life, which identifies and classifies fish, corals, and other underwater organisms to provide data support for ecological research; obstacle detection, which identifies potential collision risks such as rocks and sunken ships in real time to ensure safe navigation for the robot; and environmental feature monitoring, which identifies changes in the underwater environment, such as aquatic plants and sediments, to assist in environmental protection and monitoring. Classification can be further refined or broadly categorized into the three above, depending on the needs. Example

[0080] The present invention proposes an underwater video robot data acquisition method, comprising the following steps:

[0081] Step S101: Install a video acquisition device and multiple sensors on the underwater video robot;

[0082] Step S103: real-time acquisition of video data and environmental data, as well as corresponding time information, wherein the environmental data includes turbidity data, temperature, and salinity;

[0083] Step S105: performing target detection on the video data using a convolutional neural network (CNN), and calculating a first confidence level of the target detection result;

[0084] Step S107: Calculate a first comprehensive environment score at the current moment based on the environment data;

[0085] Step S109: Calculate the data priority at the current moment based on the first confidence level and the first comprehensive environment score;

[0086] Step S1011: Transmitting corresponding data to a ground control station based on the data priority.

[0087] In this invention, real-time target detection not only identifies underwater targets but also assesses the confidence level of the target, which directly influences the priority of that frame of video data. Environmental factors (such as water quality and temperature) also influence priority assessment. By dynamically adjusting weights, important information (such as safety and target detection results) is prioritized for processing and transmission under varying environmental conditions.

[0088] In one embodiment, a convolutional neural network (CNN) is used to detect and track underwater targets (such as marine life and sunken objects) based on real-time video streams. The loss function used is:

[0089] , where L represents the loss value, which is used to measure the gap between the model prediction and the true label, 𝑦 𝑖 Indicates the true label, indicating whether the target exists (1 for existence, 0 for non-existence). Represents the model prediction value, and represents the probability of the target existence.

[0090] The convolutional neural network updates the network weights using the following optimization algorithm:

[0091] , where θ represents the network weight, which represents the adjustable parameter in the model, 𝜂 represents the learning rate, which controls the weight update step size, and ∇L(θ) represents the gradient of the loss function, which represents the rate of change of the loss value with respect to the weight.

[0092] Wherein, the step S105 includes the following steps:

[0093] Step S1051: Calculate the classification probability of targets of different categories;

[0094] Step S1053: Calculate the loss between the predicted box and the true box using the smooth L1 function;

[0095] Step S1055: Calculate the first confidence level of the target detection result.

[0096] Wherein, the step S1051 includes: assuming there are 𝑁 categories, the CNN outputs the classification probability 𝑃(𝑐 𝑖 ), where 𝑐 𝑖 Represents the 𝑖th category, and the Softmax function is used to calculate the classification probability of each category:

[0097] , 1≤j≤N, j is an integer, 𝑧 𝑖 It is the logits output by the last layer of the CNN.

[0098] Wherein, the step S1053 includes:

[0099] Use the smooth L1 loss function to calculate the loss between the predicted box and the true box: hour, ,otherwise, , where B pred is the predicted bounding box, 𝐵 𝑡𝑟𝑢𝑒 is the true bounding box;

[0100] Calculate the box regression results, .

[0101] Wherein, the step S1055 includes: based on the classification probability and the box regression result, the first confidence calculation formula is: .

[0102] Wherein, the step S107 includes: calculating the first comprehensive environment score by normalization processing :

[0103] ;

[0104] Among them, W t 、W s and W c is the weight coefficient of temperature, salinity and turbidity in the environmental score, T represents the temperature of the underwater environment, T min Indicates the minimum value of underwater ambient temperature, T max represents the maximum temperature of the underwater environment, S represents the salinity of the underwater environment, S min Indicates the minimum value of underwater environmental salinity, S max represents the maximum salinity of the underwater environment, C represents the turbidity of the underwater environment, and C min Indicates the minimum value of underwater turbidity, C max Indicates the maximum turbidity value of the underwater environment.

[0105] The step S109 includes calculating the value of the data priority using the following formula:

[0106] ,in Represents the dynamic weight coefficient of video data, represents the dynamic weight coefficient of sensor data, Indicates the first confidence of the target detection result, Indicates the first comprehensive environment score.

[0107] The dynamic weight coefficient is calculated by the first comprehensive environment score and the first confidence level of target detection, specifically:

[0108] ;

[0109] ;

[0110] Among them, W target and W environment The weight coefficient for the first confidence and first comprehensive environment score of target detection.

[0111] Wherein, the step S1011 includes setting the data priority The data is compared with a first threshold, and data corresponding to a data priority greater than the first threshold is transmitted to a ground control station.

[0112] In one embodiment, data from multiple frames is compared, and D_total values ​​at different time points or frames are compared. Frames or time points with higher D_total values ​​or those with values ​​above a certain threshold are prioritized for transmission. The transmitted data includes video data at the current time or frame, environmental data at the current time or frame, and current time information.

[0113] In one embodiment, the environmental score Q score It includes a comprehensive assessment of temperature, salinity, and turbidity, where each environmental parameter is normalized using the following formula:

[0114] , X ∈ {T, S, C}.

[0115] The target detection adopts a multi-layer convolutional neural network structure and optimizes the target recognition capability through incremental learning to improve the detection accuracy and real-time performance in complex underwater environments.

[0116] The dynamic data transmission module adopts an adaptive algorithm to adjust the data transmission rate according to the real-time network status and data priority to ensure the timely transmission of important data.

[0117] The underwater robot has a self-navigation function, and by combining sonar data and environmental data, it can automatically avoid underwater obstacles and optimize the data collection path.

[0118] In one embodiment, the comprehensive evaluation formula further includes a time factor T time , the formula is:

[0119] , where β is the time weight coefficient, which is used to consider the timeliness of data collection.

[0120] The target detection results are post-processed by image processing algorithms to eliminate noise and improve the accuracy of target recognition, specifically including morphological processing and edge detection technology.

[0121] The underwater video robot performs data analysis and storage through a cloud computing platform and utilizes big data technology to mine historical data in order to optimize future acquisition strategies and algorithms. Example

[0122] The present invention also proposes an underwater video robot data acquisition device,

[0123] A video acquisition module, which is used to obtain video data of the underwater environment in real time;

[0124] Turbidity sensor, which is used to obtain the turbidity of the underwater environment in real time;

[0125] A temperature sensor is used to obtain the temperature of the underwater environment in real time;

[0126] Salinity sensor, which is used to obtain the salinity of the underwater environment in real time;

[0127] A target detection result module is used to perform target detection on the video data using a convolutional neural network (CNN) and calculate a first confidence level of the target detection result;

[0128] An environment score calculation module, configured to calculate a first comprehensive environment score at a current moment based on the environment data;

[0129] a priority calculation module, configured to calculate the data priority at a current moment based on the first confidence level and the first comprehensive environment score;

[0130] A data transmission module is used to transmit corresponding data to a ground control station based on the data priority. Example

[0131] An embodiment of the present disclosure provides a non-volatile computer storage medium, wherein the computer storage medium stores computer-executable instructions, and the computer-executable instructions can execute the method steps described in the above embodiment.

[0132] It should be noted that the computer-readable medium described above in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. Computer-readable storage media may include, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wire, optical cable, RF (radio frequency), or any suitable combination thereof.

[0133] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0134] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0135] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of 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 box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes 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 box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0136] The units involved in the embodiments described in this disclosure may be implemented in software or hardware, wherein the name of a unit does not necessarily limit the unit itself.

[0137] The above introduces the preferred embodiments of the present invention, which is intended to make the spirit of the present invention clearer and easier to understand, and is not intended to limit the present invention. Any modifications, replacements, and improvements made within the spirit and principles of the present invention should be included in the scope of protection outlined by the claims attached to the present invention.

Claims

1. A method for collecting data from an underwater video robot, characterized in that: The following steps are involved: Step S101: installing a video acquisition module and multiple sensors on the underwater video robot; Step S103: real-time acquisition of video data and environmental data, as well as corresponding time information, wherein the environmental data includes turbidity data, temperature, and salinity; Step S105: performing target detection on the video data using a convolutional neural network (CNN), and calculating a first confidence level of the target detection result; Step S107: Calculate the first comprehensive environmental score at the current moment through normalization processing based on the environmental data : ； Among them, W t 、W s and W c is the weight coefficient of temperature, salinity and turbidity in the environmental score, T represents the temperature of the underwater environment, T min Indicates the minimum value of underwater ambient temperature, T max represents the maximum temperature of the underwater environment, S represents the salinity of the underwater environment, S min Indicates the minimum value of underwater environmental salinity, S max represents the maximum salinity of the underwater environment, C represents the turbidity of the underwater environment, and C min Indicates the minimum value of underwater turbidity, C max Indicates the maximum value of underwater turbidity; Step S109: Calculate the data priority at the current moment based on the first confidence level and the first comprehensive environment score; Step S1011: Transmitting corresponding data to a ground control station based on the data priority.

2. The method according to claim 1, wherein: The step S105 includes the following steps: Step S1051: Calculate the classification probability of targets of different categories; Step S1053: Calculate the loss between the predicted box and the true box using the smooth L1 function; Step S1055: Calculate the first confidence level of the target detection result.

3. The method according to claim 2, wherein: The step S1051 includes: assuming there are 𝑁 categories, the CNN outputs the classification probability 𝑃(𝑐 𝑖 ), where 𝑐 𝑖 Represents the 𝑖th category, and the Softmax function is used to calculate the classification probability of each category: , 1≤j≤N, j is an integer, 𝑧 𝑖 It is the logits output by the last layer of the CNN.

4. The method according to claim 3, wherein: The step S1053 includes: Use the smooth L1 loss function to calculate the loss between the predicted box and the true box: hour, ,otherwise, , where B pred is the predicted bounding box, 𝐵 𝑡𝑟𝑢𝑒 is the true bounding box; Calculate the box regression results, .

5. The method according to claim 4, wherein: The step S1055 includes: based on the classification probability and the box regression result, the first confidence calculation formula is: .

6. The method according to claim 5, wherein: The step S109 includes calculating the value of the data priority by the following formula: ,in Represents the dynamic weight coefficient of video data, represents the dynamic weight coefficient of sensor data, Indicates the first confidence of the target detection result, represents the first comprehensive environmental score; The dynamic weight coefficient is calculated by the first comprehensive environment score and the first confidence level of target detection, specifically: ； ； Among them, W target and W environment The weight coefficient for the first confidence and first comprehensive environment score of target detection.

7. The method according to claim 6, wherein: The step S1011 includes setting the data priority The data is compared with a first threshold, and data corresponding to a data priority greater than the first threshold is transmitted to a ground control station.

8. A device for implementing the method of claim 1, comprising: A video acquisition device, which is used to acquire video data of the underwater environment in real time; Turbidity sensor, which is used to obtain the turbidity of the underwater environment in real time; A temperature sensor is used to obtain the temperature of the underwater environment in real time; Salinity sensor, which is used to obtain the salinity of the underwater environment in real time; A target detection result module is used to perform target detection on the video data using a convolutional neural network (CNN) and calculate a first confidence level of the target detection result; An environment score calculation module, configured to calculate a first comprehensive environment score at a current moment based on the environment data; a priority calculation module, configured to calculate the data priority at a current moment based on the first confidence level and the first comprehensive environment score; A data transmission module is used to transmit corresponding data to a ground control station based on the data priority.

Images