A target determination system for a continuous plankton sampling net

Abstract

The present invention provides a target determination system for a plankton continuous sampling network, which belongs to the technical field of underwater collection equipment, and includes: a control chip for controlling the operation of the entire system; three pan-tilts (first, second, and third pan-tilts), respectively equipped with searchlights, cameras, and plankton samplers; a plankton laser sampler, including a sealed black box, a transparent capillary, two sets of laser transceivers, and a micro pump. Through laser signal detection and micro pump water pumping, plankton samples in the water can be continuously collected; the control chip is provided with two modules: one is a control module, which is responsible for the control of the pan-tilt, lamps, cameras, and pumps; the other is a plankton continuous sampling target determination module, which determines whether it is a target sampling area based on image and laser signal analysis. The solution of the present invention solves the technical problem that the existing technology for plankton identification sampling often relies on naked eye observation and microscope observation to determine the sampling target area, and requires high experience of the operator.

Description

Technical Field

[0001] The invention belongs to the technical field of underwater collection equipment, and in particular relates to a target determination system for a plankton continuous sampling net. Background Art

[0002] Plankton refers to tiny organisms floating in water bodies, including phytoplankton, protozoa, small crustaceans, etc., which play a key role in aquatic ecosystems. Plankton produces a large amount of oxygen through photosynthesis, providing essential dissolved oxygen for other aquatic organisms; at the same time, plankton is also an important part of the aquatic ecological chain and the main food source for many fish and shellfish. Therefore, accurate monitoring and analysis of the species and quantity distribution of plankton in waters is of great significance for the assessment of water environment, protection and management of aquatic resources. There are two main traditional methods for sampling plankton: one is to collect surface water samples using a water sampler, and then observe and identify plankton in the laboratory; the other is to drag a plankton net to collect plankton samples, and then classify and count them. Both methods have some limitations: the former can only obtain information on the surface waters and cannot reflect the vertical distribution of plankton in the water body; the latter requires manual towing of the plankton net, which is labor-intensive and takes a long time to sample, making it difficult to achieve continuous monitoring. At the same time, the identification of plankton by these traditional methods also mainly relies on naked eye observation and microscopic examination, which has certain subjectivity and limitations. Summary of the invention

[0003] In view of this, the present invention provides a target determination system for a plankton continuous sampling network, which can solve the technical problem that the existing technology for plankton identification and sampling often relies on naked eye observation and microscope observation to determine the sampling target area, and requires high experience from the operator.

[0004] The present invention is achieved in that:

[0005] A first aspect of the present invention provides a target determination system for a plankton continuous sampling network, which includes a control chip and a first pan-tilt, a second pan-tilt, a third pan-tilt, a searchlight, a camera, and a plankton laser sampler electrically connected to the control chip. The searchlight is arranged on the first pan-tilt, the camera is arranged on the second pan-tilt, and the plankton laser sampler is arranged on the third pan-tilt. The searchlight is used to provide a light source, and the camera is used to collect underwater images. The plankton laser sampler includes a sealed black box, a transparent capillary, a first laser transceiver group, a second laser transceiver group, and a micro pump, wherein the black box is in a cube shape, the first laser transceiver group includes a first laser emitter and a first laser sensor, the first laser emitter and the first laser sensor are respectively arranged on opposite side walls of the sealed black box, the first laser beam emitted by the first laser emitter passes through the geometric center of the black box, the first laser sensor is used to receive the first laser beam, the second laser transceiver group includes a second laser emitter and a second laser sensor, and the second laser emitter and the second laser sensor are respectively arranged on opposite side walls of the sealed black box, the first laser beam emitted by the first laser emitter passes through the geometric center of the black box, the first laser sensor is used to receive the first laser beam, and ... and the first laser beam emitted by the first laser emitter passes through the geometric center of the black box, and the first laser sensor is used to receive the first laser beam. The second laser sensor is arranged at the opposite vertex of the enclosed black box, the second laser beam emitted by the second laser emitter passes through the geometric center of the black box, and the second laser sensor is used to receive the second laser beam; on the plane where the first laser beam and the second laser beam are located, two through holes are opened at the intersection of the central axis of the black box perpendicular to the first laser beam and the black box shell, the two through holes are connected by the transparent capillary, and the transparent capillary and the first laser beam and the second laser beam are on the same plane; a filter screen and a micro pump are installed at both ends of the transparent capillary respectively; a control module and a plankton continuous sampling target determination module are arranged in the control chip, the control module is used to control the first pan-tilt, the second pan-tilt, the third pan-tilt, the searchlight, the camera, the micro pump, the first laser emitter and the second laser emitter, and the plankton continuous sampling target determination module is used to analyze according to the underwater image collected by the camera, the water flow velocity formed by the micro pump sucking water, the first laser signal collected by the first laser sensor, and the second laser signal collected by the second laser sensor to determine whether the area is the target water area for continuous sampling of microorganisms.

[0006] The plankton continuous sampling target determination module specifically performs the following steps:

[0007] S10, controlling the first pan / tilt and the second pan / tilt through the control module so that the camera follows the searchlight, and obtains an underwater image of the current water area collected by the camera, and performs image preprocessing and feature extraction to obtain image features;

[0008] S20, obtaining the water flow velocity generated by the micro pump;

[0009] S30, controlling the third pan / tilt via the control module according to the image features so that the filter of the plankton optical sampler is aligned with the target area;

[0010] S40, acquiring a first laser beam signal collected by a first laser sensor and a second laser beam signal collected by a second laser sensor;

[0011] S50, calculating the size and quantity of plankton passing through the transparent capillary according to the water flow velocity, the first laser beam signal and the second laser beam signal;

[0012] S60, clustering the plankton based on the size of the plankton to obtain multiple cluster centers, calculating the density of the plankton corresponding to each cluster, and forming a plankton size density matrix;

[0013] S70, inputting the image features and the plankton size density matrix into a pre-trained plankton recognition model to obtain the types and densities of plankton;

[0014] S80, comparing the type and density of the plankton with the preset type and density requirements of the plankton to be collected, determining whether the current water area is a target water area for continuous microbial sampling and outputting the determination result.

[0015] Furthermore, feature extraction is performed on the image, including extraction of color features, texture features, and shape features.

[0016] The step S10 specifically includes:

[0017] Step 101, controlling the first pan / tilt and the second pan / tilt through a control module, so that the camera follows the searchlight to collect underwater images of the current water area in real time;

[0018] Step 102: preprocessing the acquired underwater image, including image denoising, histogram equalization, and color space conversion, to improve image quality;

[0019] Step 103: extracting visual features of the underwater image, including color features, texture features, shape features, etc., and representing these features with mathematical vectors to provide a basis for subsequent plankton identification;

[0020] Step 104, wherein the color feature can adopt the statistics of color space such as RGB, HSV, etc., the texture feature can adopt the gray level co-occurrence matrix, wavelet transform and other methods, and the shape feature can adopt the Hu invariant moment, Zernike moment and other methods;

[0021] By extracting these visual features, the visual information of underwater images can be fully portrayed, laying the foundation for subsequent plankton identification.

[0022] Optionally, the visual feature extraction in step S10 also includes: converting the underwater image into a grayscale image; calculating the Sobel operator of the grayscale image to obtain the gradient features of the image; calculating the Laplacian operator of the grayscale image to obtain the edge features of the image; and combining the gradient features and the edge features as the visual features of the underwater image.

[0023] The specific steps of step S20 include:

[0024] Step 201, by measuring the rotation speed of the micro pump, combined with the geometric parameters of the pump, such as the impeller radius, the number of blades, etc., the water flow rate is calculated; for example, the rotation speed of the micro pump is measured, combined with the geometric parameters of the pump, and the water flow velocity is calculated using the formula V=2πRNω; accurate measurement of the water flow velocity is critical for the subsequent estimation of the size of the plankton based on the time it takes to pass through the laser beam.

[0025] Optionally, the water flow velocity measurement in step S20 further includes:

[0026] Step 211, using an ultrasonic flow meter to directly measure the flow rate of the water flow in the transparent capillary tube;

[0027] Step 212: Install an ultrasonic flow meter at the inlet and outlet of the transparent capillary tube to measure the water flow velocity at the inlet and outlet;

[0028] Step 213: Calculate the average flow velocity of the water in the pipeline based on the difference in water flow velocity at the inlet and outlet.

[0029] Furthermore, the step of aligning the filter of the plankton optical sampler with the target area specifically adopts a target tracking algorithm to estimate the position of the plankton according to the image features and determine the target area.

[0030] The specific steps of step S30 are: according to the image features collected by the camera, the third pan-tilt is controlled by the control module to align the filter of the plankton optical sampler with the target area; a target tracking algorithm, such as Kalman filtering, particle filtering, etc., can be used to estimate the position of the plankton in real time according to the image features; through the precise control of the third pan-tilt, the filter is accurately aligned with the target area to maximize the capture of the plankton samples of interest;

[0031] Optionally, the target tracking algorithm needs to be combined with the kinematic model of the target, such as the uniform linear motion model, the uniformly accelerated motion model, etc., and set appropriate process noise and measurement noise parameters, so as to stably track the movement trajectory of plankton in a complex underwater environment.

[0032] The specific steps of step S40 are: obtaining a first laser beam signal collected by a first laser sensor and a second laser beam signal collected by a second laser sensor; when the plankton passes through the first laser beam, the first laser sensor detects a temporary drop in laser intensity; when the plankton passes through the second laser beam, the second laser sensor also detects a temporary drop in laser intensity;

[0033] By analyzing the laser intensity variation curves detected by the two laser sensors, the time difference Δt between the plankton passing through the two laser beams can be obtained.

[0034] Furthermore, the specific steps of step S50 include:

[0035] Step 501, calculating the size of plankton passing through the transparent capillary according to the water flow velocity V, the first laser beam signal and the second laser beam signal;

[0036] Step 502: for each detected plankton, the size L = VΔt of the plankton can be calculated according to the time difference Δt of the plankton passing through the two laser beams and the water flow velocity V;

[0037] Step 503: Count the size distribution of all plankton that pass through the transparent capillary within a predetermined time to form plankton size statistics.

[0038] Generally, the predetermined time is 15 seconds to 60 seconds.

[0039] The plankton size statistics in step S50 also include: for all plankton that pass through the transparent capillary within a certain period of time, a size frequency histogram is drawn according to the calculated size L; based on the peak value of the histogram, the size range of the main plankton species that appear can be preliminarily determined; further, the number of plankton within each size range is counted to calculate the corresponding plankton density.

[0040] Furthermore, the clustering method is the binary K-Means algorithm.

[0041] The specific steps of step S60 include:

[0042] Step 601: Perform cluster analysis based on the size of the plankton, and use the K-Means algorithm to divide the plankton into several size classes;

[0043] Step 602, setting the number of clusters K=5, indicating that the plankton is divided into 5 different size classes;

[0044] Step 603: Count the number of plankton corresponding to each cluster center, and calculate the density of plankton in each size class by combining the volume of the transparent capillary; form a plankton size density matrix, which includes density information of plankton of different sizes, and provides basic data for the next step of plankton species identification.

[0045] Furthermore, the training steps of the plankton recognition model specifically include:

[0046] Establishing a training data set, including multiple groups of historical data, each group of historical data includes the size density matrix obtained according to steps S10-S60 and the manually annotated plankton species and density of the corresponding underwater image;

[0047] The training model specifically adopts a convolutional neural network and uses the training data set for training to obtain the plankton recognition model.

[0048] The specific steps of step S80 include:

[0049] Step 801: comparing the plankton species and density with the pre-set plankton species and density requirements to be collected;

[0050] Step 802: setting a plankton species and density threshold matrix to indicate the plankton species and density ranges that are expected to be collected;

[0051] Step 803, comparing the plankton species and density information obtained in step S70 with the threshold matrix one by one;

[0052] Step 804: If the plankton species and density in the current water area fully meet the threshold requirements, it is determined to be the target water area for continuous sampling, and a judgment result of "yes" is output;

[0053] Step 805: Otherwise, output a judgment result of "No";

[0054] This threshold matrix can be adjusted according to specific application needs to provide different plankton indicator requirements.

[0055] Furthermore, the steps of the binary K-Means algorithm are specifically as follows: first, all dimension data are regarded as a whole and their centroid is calculated; then, the cluster is divided into two sub-clusters, and the centroids of the two sub-clusters are calculated respectively; and the iteration is continued until the number of sub-clusters reaches the preset K=5.

[0056] Wherein, a flow meter is arranged at the water outlet of the micro pump.

[0057] Wherein, the first laser beam and the second laser beam form an angle of 45°.

[0058] Compared with the prior art, the target determination system for a plankton continuous sampling net provided by the present invention has the following beneficial effects:

[0059] 1. It realizes comprehensive perception and intelligent analysis of the underwater environment. The system integrates multiple sensing devices such as underwater cameras, laser sensors, and micro pumps. Through a series of algorithms such as image feature extraction, target tracking, and size measurement, it can fully perceive the underwater environment and obtain information on the types, quantity, and size of plankton. Compared with a single laser-induced fluorescence method or image analysis method, the system integrates multiple sensing technologies and has stronger environmental adaptability and information perception capabilities.

[0060] 2. Intelligent target determination for continuous sampling of plankton has been realized. The system can not only detect and identify plankton in the waters, but also intelligently determine whether the current waters are suitable for continuous sampling based on pre-set sampling requirements. Through comprehensive analysis of image features, size distribution, density and other information, the system can accurately determine whether the current waters are the target sampling area, greatly improving the pertinence and effectiveness of continuous sampling.

[0061] 3. Provides detailed information on the structure of plankton populations. In addition to obtaining information on the types and quantity of plankton, the system can also measure and analyze the size distribution of plankton, providing richer data support for ecological assessment and resource management. By analyzing the laser signal and combining it with the calculation of water flow velocity, the system can accurately estimate the size of each individual plankton and form comprehensive size statistics. This information helps to better understand the structural characteristics of plankton populations in waters.

[0062] 4. Improved the automation level and accuracy of plankton monitoring. Compared with the traditional manual sampling method, the system realizes the automatic detection, identification and sampling of plankton in the waters, greatly reducing manpower input and improving the efficiency and accuracy of monitoring. At the same time, through the synergy of multiple sensors, the system can stably perceive and analyze plankton information in a complex and changeable underwater environment, providing reliable data support for subsequent ecological monitoring and resource assessment.

[0063] In summary, the plankton continuous sampling target determination system proposed by the present invention integrates a variety of cutting-edge perception and monitoring technologies, realizes comprehensive perception and intelligent analysis of the underwater environment, and provides a reliable decision-making basis for the continuous and automatic sampling of plankton. It solves the technical problem that the existing technology for plankton identification and sampling often relies on naked eye observation and microscope observation to determine the sampling target area, and requires high operator experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0065] Figure 1 A schematic diagram of the composition of a target determination system for a plankton continuous sampling net provided by the present invention;

[0066] Figure 2 This is a schematic diagram of the structure of the plankton laser sampler;

[0067] Figure 3 A flow chart of the steps executed by the module for determining the target for continuous sampling of swimming organisms;

[0068] In the accompanying drawings, the components represented by the reference numerals are listed as follows:

[0069] 10. Black box; 11. Transparent capillary; 111. Micro pump; 112. Filter; 12. First laser emitter; 13. First laser sensor; 14. Second laser emitter; 15. Second laser sensor; 21. First laser beam; 22. Second laser beam. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0071] like Figure 1 As shown in FIG. 1 , it is a schematic diagram of the composition of a target determination system for a plankton continuous sampling network provided by the present invention. The system includes a control chip and a first pan-tilt, a second pan-tilt, a third pan-tilt, a searchlight, a camera, and a plankton laser sampler electrically connected to the control chip. The searchlight is arranged on the first pan-tilt, the camera is arranged on the second pan-tilt, the plankton laser sampler is arranged on the third pan-tilt, the searchlight is used to provide a light source, and the camera is used to collect underwater images, such as Figure 2The plankton laser sampler shown includes a sealed black box 10, a transparent capillary 11, a first laser transceiver group, a second laser transceiver group and a micro pump 111, wherein the black box 10 is in the shape of a cube, the first laser transceiver group includes a first laser emitter 12 and a first laser sensor 13, the first laser emitter and the first laser sensor are respectively arranged on opposite side walls of the sealed black box, the first laser beam emitted by the first laser emitter passes through the geometric center of the black box, and the first laser sensor is used to receive the first laser beam, the second laser transceiver group includes a second laser emitter 14 and a second laser sensor 15, the second laser emitter and the second laser sensor are respectively arranged at opposite vertices of the sealed black box, the second laser beam emitted by the second laser emitter passes through the geometric center of the black box, and the second laser sensor is used to receive the second laser beam; On the plane where the laser beam is located, two through holes are opened at the intersection of the central axis of the black box perpendicular to the first laser beam and the black box shell, and the two through holes are connected by a transparent capillary, and the transparent capillary and the first laser beam 21 and the second laser beam 22 are on the same plane; a filter 112 and a micro pump are installed at both ends of the transparent capillary, respectively; a control module and a plankton continuous sampling target determination module are arranged in the control chip, the control module is used to control the first pan-tilt head, the second pan-tilt head, the third pan-tilt head, the searchlight, the camera, the micro pump, the first laser emitter and the second laser emitter, and the plankton continuous sampling target determination module is used to analyze according to the underwater image captured by the camera, the water flow velocity formed by the micro pump sucking water, the first laser signal collected by the first laser sensor, and the second laser signal collected by the second laser sensor, so as to determine whether the area is the target water area for continuous sampling of microorganisms.

[0072] like Figure 3 As shown, the plankton continuous sampling target determination module specifically performs the following steps:

[0073] S10, controlling the first pan / tilt and the second pan / tilt through the control module so that the camera follows the searchlight, and obtains an underwater image of the current water area collected by the camera, and performs image preprocessing and feature extraction to obtain image features;

[0074] S20, obtaining the water flow velocity generated by the micro pump;

[0075] S30, controlling the third pan / tilt via the control module according to the image features so that the filter of the plankton optical sampler is aligned with the target area;

[0076] S40, acquiring a first laser beam signal collected by a first laser sensor and a second laser beam signal collected by a second laser sensor;

[0077] S50, calculating the size and quantity of plankton passing through the transparent capillary according to the water flow velocity, the first laser beam signal and the second laser beam signal;

[0078] S60, clustering the plankton based on the size of the plankton to obtain multiple cluster centers, calculating the density of the plankton corresponding to each cluster, and forming a plankton size density matrix;

[0079] S70, inputting the image features and the plankton size density matrix into a pre-trained plankton recognition model to obtain the types and densities of plankton;

[0080] S80, comparing the type and density of the plankton with the preset type and density requirements of the plankton to be collected, determining whether the current water area is a target water area for continuous microbial sampling and outputting the determination result.

[0081] The specific implementation methods of the above steps are described in detail below:

[0082] The specific implementation method of step S10 is:

[0083] First, the control module controls the first pan-tilt and the second pan-tilt so that the camera follows the searchlight to collect underwater images of the current water area in real time. Then, the underwater images are preprocessed. Common preprocessing methods include image denoising, histogram equalization, color space conversion, etc. After preprocessing, the visual features in the image are extracted, such as color features, texture features, shape features, etc. These features can be represented by mathematical vectors to provide a basis for subsequent plankton identification. Among them, color features can use statistics of color spaces such as RGB and HSV; texture features can use gray-level co-occurrence matrix, wavelet transform, etc.; shape features can use Hu invariant moment, Zernike moment, etc. Methods. Through the extraction of these visual features, the visual information of underwater images can be fully characterized, laying the foundation for subsequent plankton identification.

[0084] The specific implementation method of step S20 is:

[0085] Obtain the water flow velocity generated by the micro pump. The water flow velocity can be calculated by measuring the speed of the micro pump and combining it with the geometric parameters of the pump. For example, a speed sensor can be installed to measure the speed of the micro pump in real time, and the water flow velocity V = 2πRNω can be calculated by combining the pump's impeller radius R, the number of blades N, and the speed ω. The measurement of water flow velocity is critical for the subsequent estimation of the size of plankton based on the time it takes to pass through the laser beam.

[0086] The specific implementation method of step S30 is:

[0087] According to the image features collected by the camera, the third pan-tilt is controlled by the control module to align the filter of the plankton optical sampler with the target area. Specifically, a target tracking algorithm, such as Kalman filtering, particle filtering, etc., can be used to estimate the position of plankton in real time according to the image features, and the filter can be accurately aligned with the target area through the precise control of the third pan-tilt. In this way, the plankton samples of interest can be captured to the maximum extent. The target tracking algorithm usually needs to be combined with the kinematic model of the target, such as the uniform linear motion model, the uniform acceleration motion model, etc., and set appropriate process noise and measurement noise parameters to stably track the movement trajectory of plankton in a complex underwater environment.

[0088] The specific implementation of step S40 is:

[0089] A first laser beam signal collected by a first laser sensor and a second laser beam signal collected by a second laser sensor are obtained. When the plankton passes through the first laser beam, the first laser sensor detects a brief drop in laser intensity; when the plankton passes through the second laser beam, the second laser sensor also detects a brief drop in laser intensity. By analyzing the laser intensity variation curves detected by the two laser sensors, the time difference Δt between the plankton passing through the two laser beams can be obtained. Combined with the water flow velocity V obtained in step S20, the size L of the plankton can be calculated according to the formula L=VΔt. This size estimation is very important for subsequent plankton classification and density calculation.

[0090] The specific implementation method of step S50 is:

[0091] According to the water velocity V, the first laser beam signal and the second laser beam signal, the size and number of plankton passing through the transparent capillary are calculated. Specifically, for each detected plankton, the size L = VΔt of the plankton can be calculated according to the time difference Δt of its crossing the two laser beams and the water velocity V. Then, the size distribution of all plankton passing through the transparent capillary within a certain period of time is counted to form the plankton size statistics. These data lay the foundation for the subsequent cluster analysis and plankton density calculation.

[0092] The specific implementation of step S60 is:

[0093] Clustering is performed based on the size of the plankton to obtain multiple cluster centers, the density of the plankton corresponding to each cluster is calculated, and a plankton size density matrix is ​​formed. Specifically, the K-Means clustering algorithm can be used to perform cluster analysis on the size of the plankton to obtain several cluster centers, representing plankton species of different size levels. Then, the number of plankton corresponding to each cluster center is counted, and the density of plankton at each size level is calculated in combination with the volume of the transparent capillary to form a plankton size density matrix. This matrix contains the density information of plankton of different sizes, providing basic data for the next step of plankton species identification. When clustering, the number of clusters K=5 can be set, indicating that the plankton is divided into 5 different size levels.

[0094] The specific implementation of step S70 is:

[0095] The image features and the plankton size density matrix are input into a pre-trained plankton recognition model to obtain the types and density of plankton. Here, a supervised learning method can be used to prepare a plankton image data set with labeled types in advance, and use a deep learning algorithm such as a convolutional neural network (CNN) to train the model. During training, the image features and the size density matrix are input, and the predicted plankton species are output. During the training process, a cross entropy loss function can be used to optimize the model parameters using the stochastic gradient descent method until the accuracy of the model on the validation set reaches a satisfactory level, such as more than 90%. The trained plankton recognition model can be used to predict the types and density information of plankton in unknown waters.

[0096] The specific implementation method of step S80 is:

[0097] According to the comparison between the plankton species and density and the pre-set plankton species and density requirements to be collected, it is judged whether the current water area is a target water area for continuous microbial sampling and the judgment result is output. Specifically, a threshold matrix of plankton species and density can be set to indicate the plankton species and density range expected to be collected. Then, the plankton species and density information obtained in step S70 are compared with this threshold matrix one by one. If the plankton species and density of the current water area completely meet the threshold requirements, it is determined to be a target water area for continuous sampling, and a judgment result of "yes" is output; otherwise, a judgment result of "no" is output. This threshold matrix can be adjusted according to specific application requirements to give different plankton indicator requirements.

[0098] In general, this plankton continuous sampling target determination system makes full use of a variety of sensing devices such as underwater cameras, laser sensors, and micro pumps. Through a series of algorithms such as image feature extraction, plankton size measurement, density calculation, and species identification, it accurately determines whether the current water area is suitable for continuous sampling of plankton. The key algorithms involved include target tracking, K-Means clustering, convolutional neural networks, etc., which make full use of the relevant theories and methods of mathematical modeling and computer vision. The entire system realizes the comprehensive perception and intelligent analysis of underwater environmental information, providing a reliable basis for the judgment of target water areas for subsequent continuous sampling of plankton.

[0099] In order to better understand the present invention, the specific implementation of the above modules is described in more detail below in combination with specific formulas:

[0100] The specific implementation of step S10 is as follows:

[0101] In step S10, the control module first controls the first pan / tilt and the second pan / tilt so that the camera follows the searchlight to collect underwater images of the current water area in real time. The underwater image set recorded is Among them I ( i ) represents the i-th underwater image.

[0102] Then, the underwater image set is acquired Preprocessing is performed, including image denoising, histogram equalization, color space conversion and other operations, to improve image quality. Specifically, a Gaussian filter can be used to smooth the image and remove high-frequency noise:

[0103]

[0104] in, represents the Gaussian filter function, σ is the standard deviation of the Gaussian kernel, which can be adjusted according to actual conditions.

[0105] Next, use the Gamma correction algorithm to adjust the brightness and contrast of the image:

[0106]

[0107] Here, γ is the parameter of gamma correction, which can take values ​​in the interval (0,1] to enhance the dynamic range of underwater images.

[0108] Finally, the local histogram equalization method is used to further enhance the contrast of the image:

[0109]

[0110] in, represents the local histogram equalization function, which can adaptively adjust the brightness distribution of the image and highlight the underwater targets of interest.

[0111] Preprocessed underwater image set Lay the foundation for subsequent feature extraction and analysis.

[0112] Next, from the preprocessed underwater image set Visual features are extracted from the image, including color features, texture features, shape features, etc.

[0113] For color features, the image can be converted to the HSV color space and the statistics of each channel can be calculated, such as the mean μ H ,μ S ,μ V , standard deviation σ H ,σ S ,σ V etc., as color feature vectors:

[0114]

[0115] For texture features, the gray level co-occurrence matrix (GLCM) can be used to calculate the texture features of the image, such as contrast, correlation, entropy, etc.:

[0116]

[0117] For shape features, the Hu invariant moment can be used to calculate the geometric invariance features of the image, such as the first Hu moment φ1, the second Hu moment φ2, etc.:

[0118]

[0119] The above color features, texture features and shape features are concatenated to form the final underwater image feature vector:

[0120]

[0121] By extracting these visual features, the visual information of underwater images can be fully characterized, laying the foundation for subsequent plankton identification.

[0122] The specific implementation of step S20 is as follows:

[0123] In step S20, the water flow velocity generated by the micro pump is obtained. The water flow velocity can be calculated by measuring the rotation speed of the micro pump and combining the geometric parameters of the pump.

[0124] Specifically, let the impeller radius of the micro pump be R, the number of blades be N, and the speed be ωrad / s. According to the basic principles of fluid mechanics, the velocity v of the water flow can be expressed as:

[0125] v=2πRNω

[0126] In order to measure the water flow velocity in real time, a speed sensor can be installed to obtain the speed ω of the micro pump in real time, and then the water flow velocity v is calculated according to the above formula. This water flow velocity measurement is very critical for the subsequent estimation of the size of plankton based on the time it takes to pass through the laser beam.

[0127] The specific implementation of step S30 is as follows:

[0128] In step S30, according to the underwater image features collected by the camera, the third pan / tilt is controlled by the control module so that the filter of the plankton optical sampler is aimed at the target area.

[0129] Specifically, the Kalman filter algorithm can be used to track plankton targets in underwater images in real time.

[0130] Assume that the state vector of the plankton target is Where (x, y) represents the position coordinates of the target. Represents the velocity component of the target. The state transfer equation and observation equation of Kalman filter are:

[0131] x k+1 =Fx k +w k

[0132] z k =Hx k +v k

[0133] Where F is the state transfer matrix, H is the observation matrix, and w k and v k represent state noise and observation noise respectively, and obey zero-mean Gaussian distribution.

[0134] Continuously update the target state estimate through Kalman filtering and covariance P k , the predicted position of the plankton target in the current frame can be obtained. Then, the predicted position coordinates are fed back to the control module of the third gimbal to adjust the position of the sampler so that it is accurately aligned with the target area.

[0135] Such a target tracking method can stably track the movement trajectory of plankton in complex underwater environments and provide reliable positioning support for subsequent sampling work.

[0136] The specific implementation of step S40 is as follows:

[0137] In step S40, a first laser beam signal collected by the first laser sensor and a second laser beam signal collected by the second laser sensor are acquired.

[0138] Specifically, when the plankton passes through the first laser beam, the first laser sensor detects a brief drop in laser intensity. This laser intensity change can be described by the following formula:

[0139]

[0140] Among them, I 1,0 is the initial value of the laser intensity, ΔI1 is the decrease in laser intensity, t1 is the time point when the plankton passes through the first laser beam, and σ1 is the time constant of the decrease in laser intensity.

[0141] Similarly, when plankton passes through the second laser beam, the second laser sensor will also detect a brief drop in laser intensity, which can be described by a similar model:

[0142]

[0143] The meanings of the parameters are similar to those in the above formula.

[0144] By analyzing the laser intensity variation curves detected by the two laser sensors, the time difference Δt=t2-t1 between the plankton passing through the two laser beams can be obtained. This time difference information is critical for the subsequent calculation of the plankton size based on the water flow velocity.

[0145] The specific implementation of step S50 is as follows:

[0146] In step S50, the size of the plankton passing through the transparent capillary is calculated based on the water flow velocity v, the first laser beam signal I1(t) and the second laser beam signal I2(t).

[0147] For each detected plankton, the size L of the plankton can be calculated based on the time difference Δt of its crossing the two laser beams and the water velocity v:

[0148] L=v·Δt

[0149] By counting the sizes of all plankton that pass through the transparent capillary within a certain period of time, a plankton size distribution histogram H(L) can be obtained. This size distribution histogram lays the foundation for subsequent cluster analysis and plankton density calculation.

[0150] The specific implementation of step S60 is as follows:

[0151] In step S60, based on the obtained plankton size distribution H(L), the K-Means clustering algorithm is used to divide it into K size classes.

[0152] Specifically, the K-Means algorithm iteratively optimizes the following objective function:

[0153]

[0154] in, represents the kth cluster, μ k is the centroid of the kth cluster. By continuously iteratively updating the centroid of each cluster and the sample attribution, the final K clustering results can be obtained.

[0155] Assume that the number of clusters K = 5, which means that the plankton is divided into 5 different size classes. Count the number of plankton N it contains k , combined with the volume V of the transparent capillary, the density ρ of the plankton of this size class can be calculated k :

[0156]

[0157] The density values ​​of these different size classes of plankton are combined into a density matrix Provide basic data for subsequent plankton species identification.

[0158] The specific implementation of step S70 is as follows:

[0159] In step S70, the extracted underwater image features x and the plankton size density matrix D are input into a pre-trained plankton recognition model to obtain the type and density information of the plankton.

[0160] Here we use the supervised learning method to prepare a plankton image dataset with labeled species in advance. where y (i) is the category label of the i-th sample. Using this dataset, train a multi-category classification model, such as a convolutional neural network (CNN).

[0161] Specifically, the input of the CNN model is the image feature x and the size density matrix D, and the output is the predicted probability of each category. Where C is the number of plankton categories. During training, the cross entropy loss function is used:

[0162]

[0163] The stochastic gradient descent method is used to optimize the model parameters until the classification accuracy on the validation set reaches a satisfactory level, such as above 90%.

[0164] The trained plankton recognition model can be used to predict the types and density information of plankton in unknown waters.

[0165] The specific implementation of step S80 is as follows:

[0166] In step S80, the plankton species and density information p obtained in step S70 are compared with the pre-set plankton species and density requirements to be collected to determine whether the current water area is a target water area for continuous microbial sampling.

[0167] Specifically, a threshold matrix of plankton species and density can be set Indicates the lower and upper limits of the density of each type of plankton that you expect to collect.

[0168]

[0169] Among them, t low ,t high are the lower and upper limits of the threshold matrix, is an indicator function. If the plankton species distribution p in the current water area fully meets the threshold requirement, it means that the water area is the target water area for continuous sampling, and the judgment result of "yes" is output; otherwise, the judgment result of "no" is output.

[0170] This threshold matrix T can be adjusted according to specific application needs to give different plankton index requirements. For example, for coastal waters, some phytoplankton density requirements can be increased; while for inland lakes, more attention may be paid to zooplankton indicators. By flexibly adjusting the threshold matrix T, water area judgment can be more in line with actual monitoring needs.

[0171] In summary, this plankton continuous sampling target determination system makes full use of a variety of sensing devices such as underwater cameras, laser sensors, and micro pumps. Through a series of algorithms such as image feature extraction, plankton size measurement, density calculation, and species identification, it accurately determines whether the current water area is suitable for continuous sampling of plankton. The key algorithms involved include Kalman filtering, K-Means clustering, convolutional neural networks, etc., which make full use of the relevant theories and methods of mathematical modeling and computer vision. The entire system realizes the comprehensive perception and intelligent analysis of underwater environmental information, providing a reliable basis for the judgment of target water areas for subsequent continuous sampling of plankton.

[0172] The specific implementation of step S90 is as follows:

[0173] In step S90, according to the water area judgment result of step S80, if the current water area is confirmed as the target water area for continuous sampling of plankton, the continuous sampling system is started to continuously sample the plankton in the water area.

[0174] Specifically, the third pan / tilt can be controlled by the control module to accurately align the filter of the plankton optical sampler with the target area. At the same time, the micro pump is started to suck the water sample containing plankton into the sampling system.

[0175] In order to ensure the sampling quality, some monitoring indicators can be set, such as water flow speed, plankton density, etc., which can be monitored in real time and fed back to the control module. If some indicators are monitored to be beyond the preset range, the parameters of the sampling system can be adjusted in time to ensure that the sampling process is stable and effective.

[0176] After sampling is completed, the collected plankton samples will be sent to the laboratory for further identification and analysis to provide valuable data support for related research.

[0177] Through the continuous sampling process of step S90 above, combined with the intelligent judgment of the target water area in the previous steps S10-S80, efficient and accurate sampling of plankton in the target water area can be achieved, laying a solid foundation for subsequent ecological monitoring, resource assessment and other work.

[0178] Specifically, the principle of the present invention is:

[0179] The basic principle is: when the micro pump absorbs water, the water will flow through the transparent capillary, the first laser beam will be irradiated vertically at the center of the transparent capillary, and the second laser beam will be irradiated at the center of the transparent capillary at 45 degrees. The plankton that follows the water flow through the transparent capillary will block or partially block the laser beam, causing the laser beam collected by the laser sensor to temporarily disappear or weaken. According to the water flow speed, the size of the plankton can be calculated to form a large amount of plankton size statistics. Combined with underwater images, the plankton species and density in the current water area can be determined to determine whether it is the collection target water area of ​​the plankton continuous sampling network. The specific principle is as follows:

[0180] 1. Comprehensive perception of underwater environment information

[0181] The system integrates a variety of sensing devices such as underwater cameras, laser sensors, and micro pumps, and can fully perceive the underwater environment from different angles. Among them, the underwater camera is used to collect image information in the water area in real time, providing visual data for subsequent identification of plankton species; the laser sensor is used to measure the time information of plankton passing through the laser beam, and the size of the plankton can be estimated in combination with the water flow speed; the micro pump is responsible for sucking water samples to provide support for the continuous sampling of plankton. Through the synergistic effect of multiple sensors, the system can fully perceive the various characteristic indicators of plankton in the water area.

[0182] 2. Intelligent analysis of plankton information

[0183] Based on the above perception data, the system uses a series of intelligent algorithms to analyze and process underwater environmental information. First, the collected underwater images are preprocessed, including denoising, brightness adjustment and other operations to improve image quality; then, visual features such as color, texture, and shape in the image are extracted to lay the foundation for subsequent plankton identification. At the same time, the system uses target tracking algorithms such as Kalman filtering to accurately locate the position of plankton, providing support for the precise positioning of the sampler. In terms of measuring the size of plankton, the system analyzes the signals detected by the laser sensor, calculates the size of each plankton based on the time difference between the plankton passing through the two laser beams and the water flow velocity. Further, the system performs cluster analysis on these size data to determine the density distribution of plankton of different size levels. Finally, the system inputs the image features and size density information into a pre-trained plankton identification model, such as a convolutional neural network, to obtain the types and density information of plankton in the water area. Combining these analysis results, the system can fully grasp important indicators such as the types, quantity and size distribution of plankton in the current water area.

[0184] 3. Intelligent judgment and continuous sampling of target waters

[0185] Based on the above comprehensive plankton information analysis, the system can intelligently determine whether the current water area meets the requirements of continuous sampling. Specifically, the system pre-sets a threshold matrix of plankton species and density, which represents the plankton indicators expected to be collected. Then, the actual monitored plankton information is compared with this threshold matrix. If the plankton species and density of the current water area fully meet the preset requirements, it is determined to be the target water area for continuous sampling; otherwise, it is determined to be a non-target water area. Such an intelligent judgment mechanism ensures that the continuous sampling system can accurately locate the best sampling area, effectively improving the pertinence and efficiency of sampling. At the same time, this comprehensive analysis method based on multiple indicators also enhances the system's adaptability and robustness in complex and changeable underwater environments.

[0186] In summary, the plankton continuous sampling target determination system of the present invention fully integrates cutting-edge technologies such as multi-sensor acquisition, image analysis, and pattern recognition, and realizes comprehensive perception and intelligent analysis of underwater environmental information. The system can accurately measure the types, quantity, and size distribution of plankton, and intelligently determine whether the current waters are suitable for continuous sampling based on preset sampling requirements, greatly improving the automation level and accuracy of plankton monitoring.

[0187] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A target determination system for a plankton continuous sampling net, characterized in that: It includes a control chip and a first pan-tilt, a second pan-tilt, a third pan-tilt, a searchlight, a camera, and a plankton laser sampler electrically connected to the control chip, wherein the searchlight is arranged on the first pan-tilt, the camera is arranged on the second pan-tilt, and the plankton laser sampler is arranged on the third pan-tilt, the searchlight is used to provide a light source, and the camera is used to collect underwater images. The plankton laser sampler includes a sealed black box, a transparent thin tube, a first laser transceiver group, a second laser transceiver group, and a micro pump, wherein the black box is in a cube shape, the first laser transceiver group includes a first laser emitter and a first laser sensor, the first laser emitter and the first laser sensor are respectively arranged on opposite side walls of the sealed black box, the first laser beam emitted by the first laser emitter passes through the geometric center of the black box, the first laser sensor is used to receive the first laser beam, the second laser transceiver group includes a second laser emitter and a second laser sensor, the second laser emitter and the second laser sensor are respectively arranged at opposite vertices of the sealed black box , the second laser beam emitted by the second laser emitter passes through the geometric center of the black box, and the second laser sensor is used to receive the second laser beam; on the plane where the first laser beam and the second laser beam are located, two through holes are opened at the intersection of the central axis of the black box perpendicular to the first laser beam and the black box shell, and the transparent capillary connects the two through holes, and the transparent capillary and the first laser beam and the second laser beam are on the same plane; a filter screen and a micro pump are installed at both ends of the transparent capillary, respectively; a control module and a plankton continuous sampling target determination module are set in the control chip, and the control module is used to control the first pan-tilt head, the second pan-tilt head, the third pan-tilt head, the searchlight, the camera, the micro pump, the first laser emitter and the second laser emitter, and the plankton continuous sampling target determination module is used to analyze according to the underwater image collected by the camera, the water flow velocity formed by the micro pump sucking water, the first laser signal collected by the first laser sensor, and the second laser signal collected by the second laser sensor to determine whether the area is a target water area for continuous sampling of microorganisms; The plankton continuous sampling target determination module specifically performs the following steps: S10, controlling the first pan / tilt and the second pan / tilt through the control module so that the camera follows the searchlight, obtains the underwater image of the current water area collected by the camera, and performs image preprocessing and feature extraction to obtain image features; S20, obtaining the water flow rate generated by the micro pump; S30, controlling the third pan / tilt via the control module according to the image features so that the filter of the plankton optical sampler is aligned with the target area; S40, acquiring a first laser beam signal collected by a first laser sensor and a second laser beam signal collected by a second laser sensor; S50, calculating the size and quantity of plankton passing through the transparent capillary according to the water flow velocity, the first laser beam signal and the second laser beam signal; S60, clustering the plankton based on the size of the plankton to obtain multiple cluster centers, calculating the density of the plankton corresponding to each cluster, and forming a plankton size density matrix; S70, inputting the image features and the plankton size density matrix into a pre-trained plankton recognition model to obtain the types and densities of plankton; S80, comparing the type and density of the plankton with the pre-set type and density requirements of the plankton to be collected, determining whether the current water area is a target water area for continuous microbial sampling and outputting the determination result; The specific steps of step S50 include: Step 501, calculating the size of plankton passing through the transparent capillary according to the water flow velocity V, the first laser beam signal and the second laser beam signal; Step 502: for each detected plankton, the size L = VΔt of the plankton is calculated based on the time difference Δt of the plankton passing through the two laser beams and the water velocity V; Step 503: Count the size distribution of all plankton that pass through the transparent capillary within a predetermined time to form plankton size statistics.

2. A target determination system for a plankton continuous sampling net according to claim 1, characterized in that: Perform feature extraction on the image, including extracting color features, texture features and shape features.

3. A target determination system for a plankton continuous sampling net according to claim 1, characterized in that: The step of aligning the filter screen of the plankton optical sampler with the target area specifically adopts a target tracking algorithm to estimate the position of the plankton according to the image features and determine the target area.

4. The target determination system for a plankton continuous sampling net according to claim 1, characterized in that: The clustering method is the binary K-Means algorithm.

5. A target determination system for a plankton continuous sampling net according to claim 4, characterized in that: The training steps of the plankton recognition model specifically include: Establishing a training data set, including multiple groups of historical data, each group of historical data includes the size density matrix obtained according to steps S10-S60 and the manually annotated plankton species and density of the corresponding underwater image; The training model specifically adopts a convolutional neural network and uses the training data set for training to obtain the plankton recognition model.

6. A target determination system for a plankton continuous sampling net according to claim 5, characterized in that: The steps of the binary K-Means algorithm are specifically as follows: first, all dimension data are regarded as a whole and their centroid is calculated; then the cluster is divided into two sub-clusters, and the centroids of the two sub-clusters are calculated respectively; and iterate continuously until the number of sub-clusters reaches the preset K=5.

7. The target determination system for a plankton continuous sampling net according to claim 1, characterized in that: A flow meter is arranged at the water outlet of the micro pump.

8. The target determination system for a plankton continuous sampling net according to claim 1, characterized in that: The first laser beam and the second laser beam form an angle of 45°.

Images