A measurement system for the drag coefficient of marine biological currents

Abstract

The present invention discloses a measurement system for the drag coefficient of marine biological flows, including a regional acquisition module, a model simulation module, a test calculation module, an operation analysis module, a management platform, a mobile terminal, and a cloud server. The mobile terminal is used for users to log in to the management platform, select each group of information, and issue relevant operation commands. The present invention can ensure that the tracking frame can be fully cross-camera matched with the newly added identified marine organisms, avoid missing matches, improve the calculation accuracy of the subsequent drag coefficient of marine biological flows, can real-time feedback the accuracy rate, detection rate, and false alarm rate of the simulation model to the staff, facilitate the maintenance of the staff, and can automatically recycle and clean the data stored in the cloud server to improve the data transmission efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of ecological environment simulation, and particularly to a measurement system for the drag coefficient of marine biological flow. Background Art

[0002] The ocean is the largest body of water on the earth. The ocean itself is the largest heat storage body on the earth's surface, and ocean currents are the largest heat energy conveyor belts on the earth's surface. The gas exchange between the ocean and the atmosphere has a particularly great impact on the change and development of climate. With the continuous enhancement of people's environmental protection awareness, the research on marine organisms has also become one of the important ways to protect the ocean. The biodiversity of marine organisms is not only an important indicator of the ocean situation, but also the key to protecting the marine environment.

[0003] In the existing measurement system for the drag coefficient of marine biological flow, the tracking frame and the newly added determination of marine organisms cannot be cross-camera matched, and missing matches are likely to occur; in addition, the existing measurement system for the drag coefficient of marine biological flow is inconvenient for staff to maintain, and the data transmission efficiency is poor. For this reason, we propose a measurement system for the drag coefficient of marine biological flow. Summary of the Invention

[0004] The purpose of the present invention is to solve the defects existing in the prior art, and to propose a measurement system for the drag coefficient of marine biological flow.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A measurement system for the drag coefficient of marine biological flow includes a regional acquisition module, a model simulation module, a test calculation module, an operation analysis module, a management platform, a mobile terminal, and a cloud server;

[0007] Among them, the mobile terminal is used for users to log in to the management platform, select each group of information, and issue relevant operation commands;

[0008] The management platform is used to receive the data sent by each sub-module, display it in a graphical form, receive the operation commands sent by the mobile terminal, and issue relevant control instructions;

[0009] The regional acquisition module is used to collect the image information and water flow information of the marine area selected by the user;

[0010] The model simulation module is used to perform image processing on the collected regional image information, and then perform cascade analysis on the processed image to construct a simulation model;

[0011] The test calculation module is used to calculate and record the drag coefficient of the marine biological flow in the corresponding area according to the data generated by the simulation model;

[0012] The operation analysis module is used to collect the operation status of the simulation model in real time and provide repair feedback for it;

[0013] The cloud server is used to store the calculated drag coefficients of marine biological flows in each region, and regularly clean and recycle the stored data;

[0014] The cascade analysis of the model simulation module is specifically as follows:

[0015] Step (1): The model simulation module calculates and records the interval time of the actual video frames of each group of processed environmental pictures, constructs a first-level backbone convolutional neural network to extract the feature data of each group of environmental pictures and sends it to the bidirectional feature pyramid for feature fusion, and performs classification regression on the fusion result to output detection boxes, categories, and scores;

[0016] Step (2): Then construct a second-level object detection network, take the resolution r of the input environmental picture, the width w and depth d of the object detection network as parameters to be optimized, and conduct a large number of searches on the object detection network architecture. When the number of parameters of the object detection network is less than a certain value, find the parameters (r, w, d) that make its precision the highest, collect the biological detection box information in the environmental picture, generate the corresponding detection box coordinates, and perform enlarged cropping on the relevant environmental pictures to obtain biological pictures;

[0017] Step (3): Filter out the simple negative samples belonging to the background in each group of biological pictures, establish a motion model through the Kalman filtering theory, and simultaneously obtain the motion states of each tracking target in real time through the constructed motion model. The motion model defines the motion states of the tracking targets in the video frames based on the linear motion assumption of the tracking targets, collects the motion states of the tracking targets in the current video frame, and constructs a prediction equation to estimate the motion states of each tracking target in the next video frame;

[0018] Step (4): According to the detection results of all marine organisms in the current video frame of the i-th region acquisition module calculated by the multi-target real-time detection algorithm, then calculate the cosine distance between the detection results in the current video frame of the i-th region acquisition module and the tracking targets, and then use the Hungarian algorithm to perform bipartite matching on each group of cosine distances;

[0019] Step (5): Update the Kalman gain and covariance matrix in the motion models of the tracking targets that finally fail to match. After the update is completed, re-match the groups of tracking targets that finally fail to match. After the matching is completed, sequentially process the video frame data of each video stream in parallel, and sequentially execute target marking, estimate the motion states of the tracking targets, matching association, and cross-module multi-target real-time tracking on the video frames obtained after downsampling each video stream.

[0020] As a further solution of the present invention, the mobile terminal specifically includes a smart phone, a tablet computer, and a laptop computer.

[0021] As a further solution of the present invention, the specific image processing steps of the model simulation module are as follows:

[0022] Step 1: Extract the collected image data frame by frame to obtain environmental pictures of different time frames, determine the number of blocks according to the display ratio of each group of environmental pictures, perform block processing on the environmental pictures, and at the same time perform blurring processing on the block-completed environmental pictures through low-pass filtering;

[0023] Step 2: Calculate the threshold value for each pixel point in the processed environmental pictures one by one, find the feature area through the gray threshold value. After obtaining the feature area, perform the mutual transformation of image space conversion to the frequency space through the forward and inverse Fourier transforms, and perform filtering processing on the high-frequency components in the environmental pictures converted to the frequency space to reduce noise interference.

[0024] As a further solution of the present invention, the specific transformation formulas of the forward and inverse Fourier transforms are as follows:

[0025]

[0026] In the formula, both u and v are frequency variables, x and y are the coordinates of each pixel point of the environmental picture. Formula (1) is the forward Fourier transform, and formula (2) is the inverse Fourier transform.

[0027] As a further solution of the present invention, the specific calculation and recording steps of the test calculation module are as follows:

[0028] The first step: The test calculation module extracts the depth and movement speed of the marine organisms in the simulation model, calculates the acceleration of each marine organism in the water, and at the same time receives the water flow information collected by the area collection module;

[0029] The second step: Based on the collected data of each group, calculate the logarithmic distribution of the average external flow velocity of each marine organism, and calculate the surface friction of each marine organism according to its logarithmic distribution. Then, calculate the drag coefficient of each marine organism through the calculated data of each group, and generate a coefficient record table to record the name, depth, and drag coefficient of each marine organism. The specific calculation formula is as follows:

[0030]

[0031] D = ρC D (z)U 2 (z) (4)

[0032] It can be obtained from formula (3) and formula (4)

[0033] C D (z) = [u * / U(z)] 2 (5)

[0034] In the formula, z represents the water depth, U(z) represents the flow velocity at a distance z from the bottom of the marine organism, u * represents the friction velocity, K represents the von Kármán constant, the usual value is 0.4, z0 represents the friction depth, D represents the frictional resistance of the marine organism, ρ represents the fluid density, C D (z) represents the drag coefficient of the marine organism at z, where, u * and z0 can be obtained by least squares fitting.

[0035] As a further solution of the present invention, the specific steps of the repair feedback of the operation analysis module are as follows:

[0036] S1: When the operation analysis module collects the operation data of the simulation model in real time, and calculates the loss value during the simulation process of the simulation model through the focal loss function, the specific calculation formula of the focal loss function is as follows:

[0037] FL(pi) = -α(1 - pi) γ log(pi) (6)

[0038] In the formula, pi represents the predicted value, α represents the weight factor, and γ represents the focusing parameter;

[0039] S2: The operation analysis module determines whether the loss value meets the default or manually set expected value of the system. If not, it analyzes the accuracy rate, detection rate, and false alarm rate of the screening network model, and feeds back the analysis results to the staff for viewing and adjustment.

[0040] As a further solution of the present invention, the specific steps of the cleaning and recycling of the cloud server are as follows:

[0041] P1: After the cloud server receives the operation instruction issued by the management platform, it regularly calculates and updates the recovery rate of each group of stored data according to the default or manually set cycle time value, and at the same time feeds back the updated recovery rate value to the management platform for the staff to view;

[0042] P2: Then the cloud server recovers each group of stored data according to the calculated recovery rate ratio. At the same time, the cloud server visualizes the recovery information and feeds it back to the management platform for the staff to view.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] 1. Compared with the previous measurement system for the drag coefficient, in this system, after the model simulation module processes the collected marine area image information frame by frame, based on each group of processed environmental pictures, the interval time of the actual video frames of each group of environmental pictures is calculated and recorded. Then, through the backbone convolutional neural network and the object detection network, each group of environmental pictures is detected to obtain pictures of each marine organism. After that, a motion model is established through the Kalman filter theory. At the same time, the motion states of each marine organism are obtained in real time through the constructed motion model, and a prediction equation is constructed to estimate the motion states of each tracking target in the next video frame. Then, according to the detection results of all marine organisms in the current video frame of the i-th area acquisition module calculated by the multi-object real-time detection algorithm, the Hungarian algorithm is used for bipartite matching. After the matching is completed, the video frame data of each video stream is processed in parallel in turn, and for each video frame obtained after downsampling of each video stream, target marking, estimating the motion state of the tracking target, matching association, and cross-module multi-object real-time tracking are performed in sequence, which can ensure that the tracking box can be fully cross-camera matched with the newly determined marine organisms, avoid missing matches, and improve the calculation accuracy of the drag coefficient of the subsequent marine organism flow.

[0045] 2. When the running analysis module collects the running data of the simulation model in real time, the focal loss function is used to calculate the loss value during the simulation process of the simulation model. Then, the running analysis module determines whether the loss value meets the expected value set by the system default or manually. If not, the accuracy rate, detection rate, and false alarm rate of the screening network model are analyzed, and the analysis results are fed back to the staff for viewing and adjustment. At the same time, the cloud server automatically calculates and updates the recovery rate of each group of stored data according to the cycle time value set by the system default or manually. Then, the cloud server recovers each group of stored data according to the calculated recovery rate ratio, which can feedback the accuracy rate, detection rate, and false alarm rate of the simulation model to the staff in real time, facilitating the staff to maintain, and at the same time can automatically recycle and clean the data stored in the cloud server, improving the data transmission efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention.

[0047] Figure 1 It is a system block diagram of a measurement system for the drag coefficient of a marine organism flow proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0049] Embodiment 1

[0050] Referring to Figure 1 , a measurement system for the drag coefficient of marine biological flow includes a regional acquisition module, a model simulation module, a test calculation module, an operation analysis module, a management platform, a mobile terminal, and a cloud server.

[0051] The mobile terminal is used for the user to log in to the management platform, select each group of information, and issue relevant operation commands at the same time.

[0052] The management platform is used to receive the data sent by each sub-module, display it in a graphical form, receive the operation commands sent by the mobile terminal, and issue relevant control instructions at the same time.

[0053] The regional acquisition module is used to collect the image information and water flow information of the marine area selected by the user.

[0054] The model simulation module is used to perform image processing on the collected regional image information, and then perform cascade analysis on the processed image to construct a simulation model.

[0055] Specifically, the collected image data is extracted frame by frame to obtain environmental pictures of different time frames. The number of blocks is determined according to the display ratio of each group of environmental pictures, and the environmental pictures are block-processed. At the same time, the block-completed environmental pictures are blurred by low-pass filtering. Then, the threshold value is calculated point by point for each pixel in the processed environmental pictures, and the feature area is found through the gray threshold. After obtaining the feature area, the image space is transformed to the frequency space through the forward and inverse Fourier transforms, and the high-frequency components in the environmental pictures transformed to the frequency space are filtered to reduce noise interference.

[0056] It should be further noted that the specific transformation formulas of the forward and inverse Fourier transforms are as follows:

[0057]

[0058] In the formula, u and v are both frequency variables, x and y are the coordinates of each pixel point of the environmental picture. Formula (1) is the forward Fourier transform, and formula (2) is the inverse Fourier transform.

[0059] Specifically, the model simulation module calculates and records the interval time of the actual video frames of each group of processed environmental pictures, constructs a first-level backbone convolutional neural network to extract the feature data of each group of environmental pictures and sends it to a bidirectional feature pyramid for feature fusion, classifies and regresses the fusion result to output detection boxes, categories, and scores. Then, a second-level object detection network is constructed, and the resolution r of the input environmental pictures, the width w, and the depth d of the object detection network are used as parameters to be optimized. Then, a large number of searches are performed on the object detection network architecture. When the number of parameters of the object detection network is less than a certain value, the parameters (r, w, d) that maximize its precision are found. The biological detection box information in the environmental pictures is collected, the corresponding detection box coordinates are generated, and the relevant environmental pictures are enlarged and cropped to obtain biological pictures. Then, the simple negative samples belonging to the background in each group of biological pictures are filtered out, and a motion model is established through the Kalman filtering theory. At the same time, the motion states of each tracking target are obtained in real time through the constructed motion model. The motion model defines the motion state of the tracking target in the video frame based on the linear motion assumption of the tracking target, collects the motion state of the tracking target in the current video frame, and constructs a prediction equation to estimate the motion state of each tracking target in the next video frame. Then, according to the detection results of all marine organisms in the current video frame of the i-th area acquisition module calculated by the multi-object real-time detection algorithm, the cosine distance between the detection results in the current video frame of the i-th area acquisition module and the tracking target is calculated. Then, the Hungarian algorithm is used to perform bipartite matching on each group of cosine distances. For the Kalman gain and covariance matrix in the motion model of the tracking target that finally fails to match, they are updated. After the update is completed, the groups of tracking targets that finally fail to match are rematched. After the matching is completed, the video frame data of each video stream is processed in parallel in turn, and the video frames obtained after downsampling each video stream are sequentially subjected to target marking, estimating the motion state of the tracking target, matching association, and cross-module multi-object real-time tracking.

[0060] Embodiment 2

[0061] Refer to Figure 1 , a measurement system for the drag coefficient of a marine biological flow, including an area acquisition module, a model simulation module, a test calculation module, an operation analysis module, a management platform, a mobile terminal, and a cloud server.

[0062] The test calculation module is used to calculate and record the drag coefficient of the marine biological flow in the corresponding area according to each group of data generated by the simulation model.

[0063] Specifically, the test calculation module extracts the depth and movement speed of marine organisms in the simulation model, calculates the acceleration of each marine organism in water, and simultaneously receives the water flow information collected by the area collection module. Then, based on the collected data groups, it calculates the logarithmic distribution of the average external flow velocity of each marine organism, calculates the surface friction force of each marine organism based on its logarithmic distribution, calculates the drag coefficient of each marine organism through the calculated data groups, and generates a coefficient record table to record the names, depths, and drag coefficients of each marine organism.

[0064] It should be further noted that the specific calculation formula is as follows:

[0065]

[0066] D = ρC D (z)U 2 (z) (4)

[0067] From formula (3) and formula (4), it can be obtained that

[0068] C D (z) = [u * / U(z)] 2 (5)

[0069] In the formula, z represents the water depth, U(z) represents the flow velocity at a distance z from the bottom of the marine organism, u * represents the friction velocity, K represents the von Kármán constant, whose typical value is 0.4, z0 represents the friction depth, D represents the friction resistance of the marine organism, ρ represents the fluid density, C D (z) represents the drag coefficient of the marine organism at z, where, u * and z0 can be obtained through least squares fitting.

[0070] The operation analysis module is used to collect the operation status of the simulation model in real time and provide repair feedback.

[0071] Specifically, when the operation analysis module collects the operation data of the simulation model in real time, it calculates the loss value during the simulation process of the simulation model through the focal loss function. Then, the operation analysis module determines whether the loss value meets the default or manually set expected value of the system. If it does not meet, it analyzes the accuracy rate, detection rate, and false alarm rate of the screening network model and feeds back the analysis results to the staff for viewing and adjustment.

[0072] It should be further noted that the specific calculation formula of its focal loss function is as follows:

[0073] FL(pi) = -α(1 - pi) γ log(pi) (6)

[0074] Wherein, pi represents the predicted value, α represents the weight factor, and γ represents the focusing parameter.

[0075] The cloud server is used to store the calculated drag coefficients of marine biological flows in each area, and regularly clean and recycle the stored data.

[0076] Specifically, after receiving the operation instruction issued by the management platform, the cloud server calculates and updates the recovery rate of each group of stored data at regular intervals according to the loop time value set by default in the system or manually. At the same time, the cloud server feeds back the updated recovery rate value to the management platform for the staff to view. After that, the cloud server recovers each group of stored data according to the calculated recovery rate ratio. At the same time, the cloud server processes the recovery information into images and feeds it back to the management platform for the staff to view.

Claims

1. A measurement system for the drag coefficient of marine biological flow, characterized in that: It includes a regional acquisition module, a model simulation module, a test calculation module, an operation analysis module, a management platform, a mobile terminal, and a cloud server; Among them, the mobile terminal is used for users to log in to the management platform, select each group of information, and issue relevant operation commands at the same time; The management platform is used to receive the data sent by each sub-module, display it in a graphical form, receive the operation commands sent by the mobile terminal, and issue relevant control instructions at the same time; The regional acquisition module is used to acquire the marine area image information and water flow information selected by the user; The model simulation module is used to perform image processing on the acquired regional image information, and then perform cascade analysis on the processed image to construct a simulation model; The test calculation module is used to calculate and record the drag coefficient of the marine biological flow in the corresponding area according to the data generated by the simulation model; The operation analysis module is used to collect the operation status of the simulation model in real time and perform repair feedback on it; The cloud server is used to store the calculated drag coefficients of the marine biological flows in each area, and clean up and recycle the stored data regularly; The specific steps of the cascade analysis of the model simulation module are as follows: Step (1): The model simulation module calculates and records the interval time of the actual video frames of each group of processed environmental pictures, constructs a first-level backbone convolutional neural network to extract the feature data of each group of environmental pictures, sends it to a bidirectional feature pyramid for feature fusion, and performs classification regression on the fusion result to output detection boxes, categories, and scores; Step (2): Then construct a second-level object detection network, use the resolution r of the input environmental picture, the width w and depth d of the object detection network as parameters to be optimized, search the object detection network architecture a large number of times, and find the parameters (r, w, d) that make its precision the highest when the number of parameters of the object detection network is less than a certain value, collect the biological detection box information in the environmental picture, generate the corresponding detection box coordinates, and perform enlarged cropping on the relevant environmental pictures to obtain biological pictures; Step (3): Filter out the simple negative samples belonging to the background in each group of biological pictures, establish a motion model through the Kalman filtering theory, and obtain the motion state of each tracking target in real time through the constructed motion model. The motion model defines its motion state in the video frame based on the linear motion assumption of the tracking target, collects the motion state of the tracking target in the current video frame, and constructs a prediction equation to estimate the motion state of each tracking target in the next video frame; Step (4): According to the detection results of all marine organisms in the current video frame of the i-th regional acquisition module calculated by the multi-target real-time detection algorithm, then calculate the cosine distance between the detection results in the current video frame of the i-th regional acquisition module and the tracking target, and then use the Hungarian algorithm to perform bipartite matching on each group of cosine distances; Step (5): Update the Kalman gain and covariance matrix in the motion model of the tracking targets that finally fail to match. After the update is completed, re-match each group of tracking targets that finally fail to match. After the matching is completed, process the video frame data of each video stream in parallel in turn, and sequentially perform target marking, estimate the motion state of the tracking targets, matching association, and cross-module multi-target real-time tracking on the video frames obtained after downsampling each video stream.

2. The measuring system for the drag coefficient of marine biological flow according to claim 1, characterized in that: The mobile terminal specifically includes a smart phone, a tablet computer, and a laptop computer.

3. The measurement system for the drag coefficient of marine biological flow according to claim 1, characterized in that: The specific image processing steps of the model simulation module are as follows: Step 1: Extract the collected image data frame by frame to obtain environmental pictures of different time frames. Determine the number of blocks according to the display ratio of each group of environmental pictures, and perform block processing on the environmental pictures. At the same time, perform blurring processing on the block-completed environmental pictures through low-pass filtering. Step 2: Calculate the threshold value for each pixel in the processed environmental pictures point by point, and find the feature regions through the gray threshold. After obtaining the feature regions, perform the mutual transformation between the image space and the frequency space through the forward and inverse Fourier transforms, and filter the high-frequency components in the environmental pictures transformed into the frequency space to reduce noise interference.

4. The measurement system for the drag coefficient of marine biological flow according to claim 3, characterized in that: The specific transformation formulas of the forward and inverse Fourier transforms are as follows: In the formula, u and v are both frequency variables, x and y are the coordinates of each pixel point of the environmental picture. Formula (1) is the forward Fourier transform, and formula (2) is the inverse Fourier transform.

5. The measurement system for the drag coefficient of marine biological flow according to claim 1, wherein: The specific calculation and recording steps of the test calculation module are as follows: The first step: The test calculation module extracts the depth and movement speed of the marine organisms in the simulation model, calculates the acceleration of each marine organism in the water, and simultaneously receives the water flow information collected by the area collection module. The second step: Calculate the logarithmic distribution of the average external flow velocity of each marine organism based on the collected data groups, calculate the surface friction force of each marine organism based on its logarithmic distribution, and then calculate the drag coefficient of each marine organism through the calculated data groups. Generate a coefficient record table to record the names, depths, and drag coefficients of each marine organism. The specific calculation formulas are as follows: D = ρC D (z)U 2 (z) (4) From formula (3) and formula (4), it can be obtained that C D (z) = [u * / U(z)] 2 (5) Where z represents the water depth, U(z) represents the flow velocity at a distance z from the bottom of the marine organism, u * represents the friction velocity, K represents the von Kármán constant, with a typical value of 0.4, z0 represents the friction depth, D represents the frictional resistance of the marine organism, ρ represents the fluid density, C D (z) represents the drag coefficient of the marine organism at z, where, u * and z0 can be obtained by least squares fitting.

6. The measurement system for the drag coefficient of marine biological flow according to claim 1, wherein: The specific steps of the repair and feedback of the operation analysis module are as follows: S1: When the operation analysis module collects the operation data of the simulation model in real time, calculate the loss value during the simulation process of the simulation model through the focal loss function. The specific calculation formula of the focal loss function is as follows: FL(pi) = -α(1 - pi) γ log(pi) (6) In the formula, pi represents the predicted value, α represents the weight factor, and γ represents the focusing parameter; S2: The operation analysis module determines whether the loss value meets the default or manually set expected value of the system. If not, analyze the accuracy rate, detection rate, and false alarm rate of the screening network model, and feedback the analysis results to the staff for viewing and adjustment.

7. The measurement system for the drag coefficient of marine biological flow according to claim 1, characterized in that: The specific steps of the cloud server cleaning and recycling are as follows: P1: After the cloud server receives the operation instructions issued by the management platform, it regularly calculates and updates the recovery rates of each group of stored data according to the loop time value set by the system default or manually. At the same time, it feeds back the recovery rate value after each update to the management platform for the staff to view. P2: Then, the cloud server recovers each group of stored data according to the calculated recovery rate ratio. At the same time, the cloud server processes the recovery information graphically and feeds it back to the management platform for the staff to view.

Images