Plankton in-situ online imaging observation method and system based on image recognition and storage medium

Abstract

The invention discloses a plankton in-situ online imaging observation method and system based on image recognition and a storage medium, and relates to the field of plankton image recognition. Environment feature information and position feature information in a target water area are collected, and a time sequence abundance image is collected through an in-situ imager; respectively storing the time sequence abundance images to communication nodes corresponding to the transmission block chain; on a communication node, image preprocessing is carried out on a time sequence abundance image by using an image labeling and classification algorithm training strategy based on active learning, and plankton and particulate matter images are subjected to high-accuracy fine classification and identification on the preprocessed image by using a double-convolutional neural network cascade algorithm; and obtaining time sequence abundance images of different time periods corresponding to different communication nodes and a fine classification identification result of the time sequence abundance images according to different private keys. The method is of great significance to promotion of development of water area ecological monitoring, environmental protection and scientific research.

Description

Technical Field

[0001] The present invention relates to the field of plankton image recognition, and more specifically to an in-situ online imaging observation method, system and storage medium for plankton based on image recognition. Background Art

[0002] The lake ecosystem is a complex and changeable system, which is composed of the interaction between lake organisms and the surrounding environment. Some of its functions and values ​​play an important role in the life system and ensure the sustainable development of society and the environment. Lake plankton is the most basic and important component of the lake ecosystem. They are diverse in type and form and huge in number. Changes in their number, structure and spatial distribution may break the stability of the ecosystem, destroy the normal ecological balance, and even cause global climate change. Obviously, the number and distribution of plankton in lake water bodies are of great significance in the biogeochemical cycle of biogenic elements (such as C cycle, N cycle) and global climate research. Therefore, it is of great significance to monitor the population, number and spatial distribution of plankton in lakes, and the intelligent and rapid identification of plankton is a prerequisite for improving the quality of monitoring. Thanks to the rapid development of computer technology and the improvement of hardware integration quality, deep learning technology has soared and developed rapidly, and has made great breakthroughs in many fields. Especially in the field of image recognition, deep learning relies on its powerful feature extraction ability and is gradually replacing traditional machine learning methods. How to achieve automatic in situ detection of the abundance and biomass distribution of key species in large lakes, combine image recognition with deep learning, and improve the recognition ability and accuracy of lake plankton is an important issue that needs to be urgently addressed in lake detection. Summary of the invention

[0003] In view of this, the present invention provides an in-situ online imaging observation method, system and storage medium for plankton based on image recognition, which is of great significance for promoting the development of aquatic ecological monitoring, environmental protection and scientific research.

[0004] In order to achieve the above object, the present invention adopts the following technical solution:

[0005] An in-situ online imaging observation method for plankton based on image recognition comprises the following steps:

[0006] Collect environmental characteristic information and location characteristic information in the target waters, and use an in-situ imager to collect plankton abundance images corresponding to different times of plankton abundance changes in the target waters as time series abundance images;

[0007] The time series abundance images are stored in the communication nodes corresponding to the transmission blockchain according to the different target waters. Each communication node stores the time series abundance images according to the preset time period and sets different public keys;

[0008] On the communication node, the time series abundance images are preprocessed using active learning-based image annotation and classification algorithm training strategies. The preprocessed images are then subjected to a dual convolutional neural network cascade algorithm to perform high-precision fine classification and recognition of plankton and particulate matter images.

[0009] The remote user obtains the time series abundance images of different time periods corresponding to different communication nodes and the fine classification and recognition results of the time series abundance images according to different private keys.

[0010] Optionally, the remote user may predict the plankton growth and water quality of the current target waters based on the time series abundance image and the fine classification and recognition results of the time series abundance image, specifically including the following steps:

[0011] Based on the time series abundance images of each target water area and the refined classification and identification results, the plankton growth of each target water area is predicted based on the LSTM algorithm;

[0012] The plankton growth conditions of each target water area are summarized, and the similarity between the plankton growth summary results and the plankton growth conditions corresponding to different water quality conditions is calculated. Based on the similarity results, the water quality conditions corresponding to the plankton growth summary results are found, which are the water quality conditions of the current water area.

[0013] Optionally, establish a channel transmission encryption mechanism for the transmission blockchain, monitor the pairing of public and private keys and the status of the channel in real time, and allocate channels based on information retrieval:

[0014] Each channel corresponds to a random matrix and a public key vector. The values ​​in the random matrix are randomly generated according to the cached data of each communication node. The cached data of each communication node is encrypted based on the random matrix and the public key vector, and the encrypted data is transmitted through the channel.

[0015] Calculate the utilization rate of the transmission channel used to complete the pairing of the public key and the private key, determine the threshold for the channel load and delay based on the utilization rate, and when the utilization rate exceeds the threshold of the current channel, increase the channel share to take on the transmission pressure of the current channel.

[0016] Optionally, the image is preprocessed based on active learning image annotation and classification algorithm training strategy, specifically: using active learning to annotate pixel points of the time series abundance image, and based on the classification algorithm, dividing the pixel levels into plankton content points and non-plankton content points.

[0017] Optionally, a dual convolutional neural network cascade algorithm uses two convolutional neural networks to complement each other, where the first convolutional neural network extracts pixel features and edges in the image, and the second convolutional neural network completes classification and recognition based on the pixel features and edges.

[0018] Optionally, the plankton growth and water quality in the current target waters can be predicted to observe the diurnal vertical migration of zooplankton, the dynamic changes of dominant species, and complete biodiversity surveys and biological disaster warnings.

[0019] An in-situ online imaging observation system for plankton based on image recognition, comprising:

[0020] Time series abundance image acquisition module: used to collect environmental characteristic information and location characteristic information in the target waters, and collect plankton abundance images corresponding to different times of plankton abundance changes in the target waters through an in-situ imager as time series abundance images;

[0021] Time series abundance image storage module: used to store the time series abundance images on the communication nodes corresponding to the transmission blockchain according to the different target waters. Each communication node stores the time series abundance images according to the preset time period and sets different public keys.

[0022] Plankton and particulate matter image recognition module: used on the communication node to pre-process the time series abundance images using active learning-based image annotation and classification algorithm training strategies, and to use a dual convolutional neural network cascade algorithm on the pre-processed images to perform high-precision and fine classification and recognition of plankton and particulate matter images;

[0023] Remote retrieval module: used to obtain time series abundance images of different time periods corresponding to different communication nodes and fine classification and recognition results of time series abundance images according to different private keys.

[0024] Optionally, it also includes a plankton growth and water quality prediction module: used to predict the plankton growth and water quality of the current target waters based on the time series abundance image and the fine classification and recognition results of the time series abundance image, specifically including the following steps:

[0025] Based on the time series abundance images of each target water area and the refined classification and identification results, the plankton growth of each target water area is predicted based on the LSTM algorithm;

[0026] The plankton growth conditions of each target water area are summarized, and the similarity between the plankton growth summary results and the plankton growth conditions corresponding to different water quality conditions is calculated. Based on the similarity results, the water quality conditions corresponding to the plankton growth summary results are found, which are the water quality conditions of the current water area.

[0027] A computer storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of any one of the above-mentioned methods for in-situ online imaging observation of plankton based on image recognition are implemented.

[0028] It can be seen from the above technical solutions that, compared with the prior art, the present invention provides an in-situ online imaging observation method for plankton based on image recognition, which has the following beneficial effects:

[0029] 1. Collecting plankton abundance images directly in the waters through in-situ imaging can reflect the dynamic changes of plankton in real time, avoiding the delays and errors of traditional sampling methods. Collecting plankton abundance images at different time points to form time series abundance images will help analyze the changing trends of plankton over time and provide more comprehensive data support for scientific research.

[0030] 2. The image annotation and classification algorithm based on active learning can automatically select high-information images for annotation, reduce the workload of manual annotation, and improve annotation efficiency. At the same time, the application of the dual convolutional neural network cascade algorithm further improves the accuracy of fine classification and recognition of plankton and particulate matter images.

[0031] 3. Remote users can use private keys to access time series abundance images and their fine classification and identification results on different communication nodes, realizing remote access and sharing of data, making it easier for researchers to conduct collaborative research in different locations. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0033] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0034] Figure 2 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] Example 1

[0037] The embodiment of the present invention discloses an in-situ online imaging observation method of plankton based on image recognition, such as Figure 1 As shown, the following steps are included:

[0038] S1: Collect environmental characteristic information and location characteristic information in the target waters, and collect plankton abundance images corresponding to different times of plankton abundance changes in the target waters through an in-situ imager as time series abundance images;

[0039] S2: The time series abundance images are stored in the communication nodes corresponding to the transmission blockchain according to the different target waters. Each communication node stores the time series abundance images according to the preset time period and sets different public keys;

[0040] S3: On the communication node, the time series abundance images are preprocessed using active learning-based image annotation and classification algorithm training strategies. The preprocessed images are then subjected to a dual convolutional neural network cascade algorithm to perform high-precision fine classification and recognition of plankton and particulate matter images.

[0041] S4: The remote user obtains the time series abundance images of different time periods corresponding to different communication nodes and the fine classification and recognition results of the time series abundance images according to different private keys.

[0042] Furthermore, in this embodiment, the remote user predicts the plankton growth and water quality of the current target waters based on the time series abundance image and the fine classification and recognition results of the time series abundance image, which specifically includes the following steps:

[0043] S51: Based on the time series abundance images of each target water area and the refined classification and recognition results, the plankton growth of each target water area is predicted based on the LSTM algorithm;

[0044] S52: Summarize the plankton growth conditions of each target water area, calculate the similarity between the plankton growth summary result and the plankton growth conditions corresponding to different water quality conditions, and find the water quality condition corresponding to the plankton growth summary result based on the similarity result, which is the water quality condition of the current water area.

[0045] Furthermore, the plankton growth and water quality in the current target waters are predicted to observe the diurnal vertical migration of zooplankton and the dynamic changes of dominant species, and to complete biodiversity surveys and biological disaster warnings.

[0046] Furthermore, in S2, a channel transmission encryption mechanism is established for the transmission blockchain, the pairing of public keys and private keys and the status of the channel are monitored in real time, and the channel is allocated according to the information retrieval situation:

[0047] S21: Each channel corresponds to a random matrix and a public key vector. The values ​​in the random matrix are randomly generated according to the cache data of each communication node. The cache data of each communication node is encrypted based on the random matrix and the public key vector, and the encrypted data is transmitted through the channel.

[0048] S22: Calculate the usage rate of the transmission channel used to complete the pairing of the public key and the private key, determine the threshold for the channel load and delay based on the usage rate, and when the usage rate exceeds the threshold of the current channel, increase the channel share to take on the transmission pressure of the current channel.

[0049] The blockchain in this embodiment has the following functions:

[0050] Dealing with data overload: As blockchain networks develop, the amount of data continues to increase, which may lead to a decrease in transmission speed and information processing efficiency. The cache function can store part of the data and reduce frequent access to the main network, thereby dealing with the problem of data overload.

[0051] Improve transaction speed: In a blockchain network, transactions need to be verified and recorded. Cache can store recent transaction records for quick retrieval and verification, thereby improving transaction speed.

[0052] Optimize resource utilization: The cache function reduces repeated access to storage resources by storing frequently used data, thereby optimizing the utilization of storage resources.

[0053] Furthermore, in S3, the image is preprocessed based on active learning image annotation and classification algorithm training strategy, specifically: active learning is used to annotate pixel points of the time series abundance image, and based on the classification algorithm, the level of the pixel points is divided into plankton content points and non-plankton content points.

[0054] Furthermore, in S3, the dual convolutional neural network cascade algorithm uses two convolutional neural networks to complement each other. The first convolutional neural network extracts pixel features and edges in the image, and the second convolutional neural network completes classification and recognition based on pixel features and edges.

[0055] More specifically, the first convolutional neural network: This network is used to extract basic features in an image, such as color, edges, etc.

[0056] It is usually composed of multiple convolutional layers, pooling layers, and activation functions. Through the operation of convolutional layers and pooling layers, the network can extract the basic features of the image and reduce the size of the feature map. At the same time, the activation function can limit the network output to a certain range to avoid the problem of gradient disappearance or explosion.

[0057] Second convolutional neural network: This network performs more accurate image classification and object detection tasks based on the feature maps extracted by the first network.

[0058] It can also be composed of convolutional layers, pooling layers, and fully connected layers. Through further convolution and pooling operations, the network can focus on more complex features, such as shape, texture, etc., thereby improving the accuracy of classification and object detection.

[0059] Cascade connection: The output of the first convolutional neural network is used as the input of the second convolutional neural network, and the two networks are complemented and collaborated with each other through cascading.

[0060] and Figure 1 Corresponding to the method shown, the present invention also discloses an in-situ online imaging observation system for plankton based on image recognition for Figure 1 The implementation of the method, the specific structure is as follows Figure 2 As shown, including:

[0061] Time series abundance image acquisition module: used to collect environmental characteristic information and location characteristic information in the target waters, and collect plankton abundance images corresponding to different times of plankton abundance changes in the target waters through an in-situ imager as time series abundance images;

[0062] Time series abundance image storage module: used to store the time series abundance images on the communication nodes corresponding to the transmission blockchain according to the different target waters. Each communication node stores the time series abundance images according to the preset time period and sets different public keys.

[0063] Plankton and particulate matter image recognition module: used on the communication node to pre-process the time series abundance images using active learning-based image annotation and classification algorithm training strategies, and to use a dual convolutional neural network cascade algorithm on the pre-processed images to perform high-precision and fine classification and recognition of plankton and particulate matter images;

[0064] Remote retrieval module: used to obtain time series abundance images of different time periods corresponding to different communication nodes and fine classification and recognition results of time series abundance images according to different private keys.

[0065] In this embodiment, a plankton growth and water quality prediction module is also included: it is used to predict the plankton growth and water quality of the current target water area based on the time series abundance image and the fine classification and recognition results of the time series abundance image, and specifically includes the following steps:

[0066] Based on the time series abundance images of each target water area and the refined classification and identification results, the plankton growth of each target water area is predicted based on the LSTM algorithm;

[0067] The plankton growth conditions of each target water area are summarized, and the similarity between the plankton growth summary results and the plankton growth conditions corresponding to different water quality conditions is calculated. Based on the similarity results, the water quality conditions corresponding to the plankton growth summary results are found, which are the water quality conditions of the current water area.

[0068] Finally, this embodiment discloses a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, any one of the steps of the in-situ online imaging observation method of plankton based on image recognition is implemented.

[0069] Example 2

[0070] The only difference between this embodiment and embodiment 1 is that this embodiment discloses an in-situ imager, which is equipped with an image annotation and classification algorithm training strategy based on active learning, making full use of human intelligence and machine intelligence to achieve the purpose of image annotation, classifier training and classification result correction. On this basis, a dual convolutional neural network cascade algorithm is adopted, which has a large-scale image data set containing nearly 100 types of plankton and particulate matter images, effectively eliminating the interference of particulate matter in nearshore waters on plankton identification, and can achieve high-accuracy and fine classification and identification of plankton images.

[0071] The imaging system is integrated into surface buoys and deployed in rivers, lakes, reservoirs and other water bodies. It can successfully obtain time series data on the abundance changes of plankton in the waters, observe the diurnal vertical migration of zooplankton, and the dynamic changes of dominant species. The monitoring data can be used for biodiversity surveys and biological disaster warnings.

[0072] The system can also integrate conventional monitoring parameters such as water temperature, pH, dissolved oxygen, algae, waves and meteorology.

[0073] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0074] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for in-situ online imaging observation of plankton based on image recognition, characterized in that: The following steps are involved: Collect environmental characteristic information and location characteristic information in the target waters, and use an in-situ imager to collect plankton abundance images corresponding to different times of plankton abundance changes in the target waters as time series abundance images; The time series abundance images are stored in the communication nodes corresponding to the transmission blockchain according to the different target waters. Each communication node stores the time series abundance images according to the preset time period and sets different public keys; On the communication node, the time series abundance images are preprocessed using active learning-based image annotation and classification algorithm training strategies. The preprocessed images are then subjected to a dual convolutional neural network cascade algorithm to perform high-precision fine classification and recognition of plankton and particulate matter images. The remote user obtains the time series abundance images of different time periods corresponding to different communication nodes and the fine classification and recognition results of the time series abundance images according to different private keys.

2. The method for in-situ online imaging observation of plankton based on image recognition according to claim 1, characterized in that: It also includes the remote user predicting the plankton growth and water quality of the current target waters based on the time series abundance images and the fine classification and recognition results of the time series abundance images, which specifically includes the following steps: Based on the time series abundance images of each target water area and the refined classification and identification results, the plankton growth of each target water area is predicted based on the LSTM algorithm; The plankton growth conditions of each target water area are summarized, and the similarity between the plankton growth summary results and the plankton growth conditions corresponding to different water quality conditions is calculated. Based on the similarity results, the water quality conditions corresponding to the plankton growth summary results are found, which are the water quality conditions of the current water area.

3. The method for in-situ online imaging observation of plankton based on image recognition according to claim 1, characterized in that: Establish a channel transmission encryption mechanism for the transmission blockchain, monitor the pairing of public and private keys and the status of the channel in real time, and allocate channels according to the information retrieval situation: Each channel corresponds to a random matrix and a public key vector. The values ​​in the random matrix are randomly generated according to the cached data of each communication node. The cached data of each communication node is encrypted based on the random matrix and the public key vector, and the encrypted data is transmitted through the channel. Calculate the utilization rate of the transmission channel used to complete the pairing of the public key and the private key, determine the threshold for the channel load and delay based on the utilization rate, and when the utilization rate exceeds the threshold of the current channel, increase the channel share to take on the transmission pressure of the current channel.

4. The method for in-situ online imaging observation of plankton based on image recognition according to claim 1, characterized in that: The image is preprocessed based on active learning image annotation and classification algorithm training strategy, specifically: active learning is used to annotate the pixels of the time series abundance image, and the classification algorithm is used to divide the content points of plankton and non-plankton according to the level of the pixels.

5. The method for in-situ online imaging observation of plankton based on image recognition according to claim 1, characterized in that: The dual convolutional neural network cascade algorithm uses two complementary convolutional neural networks. The first convolutional neural network extracts pixel features and edges in the image, and the second convolutional neural network completes classification and recognition based on pixel features and edges.

6. The method for in-situ online imaging observation of plankton based on image recognition according to claim 2, characterized in that: Predict the growth of plankton and water quality in the current target waters, observe the diurnal vertical migration of zooplankton, the dynamic changes of dominant species, and complete biodiversity surveys and biological disaster warnings.

7. An in-situ online imaging observation system for plankton based on image recognition, characterized in that: include: Time series abundance image acquisition module: used to collect environmental characteristic information and location characteristic information in the target waters, and collect plankton abundance images corresponding to different times of plankton abundance changes in the target waters through an in-situ imager as time series abundance images; Time series abundance image storage module: used to store the time series abundance images on the communication nodes corresponding to the transmission blockchain according to the different target waters. Each communication node stores the time series abundance images according to the preset time period and sets different public keys. Plankton and particulate matter image recognition module: used on the communication node to pre-process the time series abundance images using active learning-based image annotation and classification algorithm training strategies, and to use a dual convolutional neural network cascade algorithm on the pre-processed images to perform high-precision and fine classification and recognition of plankton and particulate matter images; Remote retrieval module: used to obtain time series abundance images of different time periods corresponding to different communication nodes and fine classification and recognition results of time series abundance images according to different private keys.

8. The in-situ online imaging observation system for plankton based on image recognition according to claim 7, characterized in that: It also includes a plankton growth and water quality prediction module: it is used to predict the plankton growth and water quality of the current target waters based on the time series abundance image and the fine classification and recognition results of the time series abundance image, and specifically includes the following steps: Based on the time series abundance images of each target water area and the refined classification and identification results, the plankton growth of each target water area is predicted based on the LSTM algorithm; The plankton growth conditions of each target water area are summarized, and the similarity between the plankton growth summary results and the plankton growth conditions corresponding to different water quality conditions is calculated. Based on the similarity results, the water quality conditions corresponding to the plankton growth summary results are found, which are the water quality conditions of the current water area.

9. A computer storage medium, characterized in that The computer storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the in-situ online imaging observation method of plankton based on image recognition as described in any one of claims 1 to 6 are implemented.

Images