Algae identification method and system based on residual attention optimization network, processing equipment and storage medium

Abstract

The invention relates to an algae identification method and system based on a residual attention optimization network, a processing device and a storage medium, and the method is characterized in that the method comprises the steps: obtaining a to-be-identified algae image; the to-be-recognized algae image is input into a pre-constructed algae recognition model, an algae recognition result of the to-be-recognized algae image is obtained, the algae recognition result indicates the position and the type of algae in the to-be-recognized algae image, and the algae recognition model is obtained by training a residual attention optimization network through a pre-constructed algae data set; the method can realize automatic and efficient algae identification, and can be widely applied to the field of algae identification.

Description

Technical Field

[0001] The present invention relates to the field of algae recognition, and in particular to an algae recognition method, system, processing equipment and storage medium based on a residual attention optimization network. Background Art

[0002] Water environment quality plays a vital role in modern industry, agriculture, public health and safety. Poor water environment conditions can cause ecosystem disorder, which in turn affects people's production and life. As the main primary producer of aquatic ecosystems, algae are sensitive to changes in the external water environment. Their species, density and community structure are closely related to water quality. Climate change and excessive nitrogen, phosphorus and other elements in water can cause explosive reproduction of algae, resulting in eutrophication events such as algal blooms and red tides, which seriously damage the balance of aquatic ecosystems. In addition, the dominant algae in lakes and reservoirs will release some toxic and odorous substances into the water, including algae toxins, geosmin (GSM) and 2-methylisoborneol (2-MIB), which are harmful to drinking water safety and fisheries. Therefore, the rapid and accurate identification, classification and monitoring of algae are of great significance.

[0003] At present, the traditional practice is for professional technicians to use microscopes to identify the cell morphology of algae samples to complete algae identification and monitoring. This method is currently the most widely used, but it is time-consuming and labor-intensive. One sample requires 2-3 hours of manual identification, and complex samples are even more troublesome. When algae blooms appear, it is impossible to quickly identify the algae in the entire water sample. Moreover, algae have a variety of morphologies, including single cells, colony structures, and filaments. Different species have different cell or colony structure characteristics, which are highly complex. Therefore, the current practice of manual algae identification has low accuracy and stability. Summary of the invention

[0004] In view of the above problems, the object of the present invention is to provide an algae recognition method, system, processing equipment and storage medium based on a residual attention optimization network with high accuracy and stability.

[0005] To achieve the above objectives, the present invention adopts the following technical solutions: In a first aspect, a method for identifying algae based on a residual attention optimization network is provided, comprising: Acquire an image of algae to be identified; The algae image to be identified is input into a pre-built algae recognition model to obtain an algae recognition result of the algae image to be identified. The algae recognition result indicates the position and type of algae in the algae image to be identified. The algae recognition model is obtained by training a residual attention optimization network using a pre-built algae dataset.

[0006] Furthermore, it also includes: After obtaining the algae recognition result of the algae image to be recognized, the algae recognition result of the algae image to be recognized is corrected and updated to the constructed algae data set, and the algae recognition model is trained using the updated algae data set to correct the algae recognition model.

[0007] Furthermore, the construction process of the algae recognition model is as follows: Constructing an algae dataset, including a number of algae sample images of different types, each algae sample image having annotation information of the location and type of each algae included; Construct a residual attention optimization network; According to the algae dataset, the constructed residual attention optimization network is trained to obtain the algae recognition model.

[0008] Furthermore, the constructing of the algae dataset comprises: Using microscopic imaging equipment, photograph water slide samples to obtain images of several different types of primitive algae; Annotating algae in the acquired original algae image; Perform image preprocessing on the annotated original algae images to obtain several algae sample images; The algae sample images are subjected to image enhancement processing to obtain the algae sample images after image enhancement processing, and an algae dataset is constructed.

[0009] Furthermore, the labeled original algae image is subjected to image preprocessing to obtain a number of algae sample images, including: According to the labeling results, each original algae image is cut into a number of single algae images, and each of the cut single algae images includes the same algae species; The image preprocessing is performed on a single algae image to obtain several algae sample images.

[0010] Furthermore, the residual attention optimization network includes a convolution module, a first residual module, a first attention module, a second residual module, a second attention module, a third attention module, a third residual module, a fourth attention module, a fourth residual module, a fifth residual module, a pooling layer and a fully connected layer connected in sequence, the first residual module is also connected to the second attention module, the second residual module is also connected to the fourth attention module, and the third residual module is also connected to the fifth residual module; the output of the first residual module and the output of the second residual module are spliced ​​by channel and input into the second attention module, the output of the second residual module and the output of the third residual module are spliced ​​by channel and input into the fourth attention module, and the output of the third residual module and the output of the fourth residual module are spliced ​​by channel and input into the fifth residual module; each attention module includes a trunk branch and a soft mask branch, the trunk branch is used to extract image features, and the soft mask branch is used to perform weighted fusion on the extracted image features.

[0011] Furthermore, the input size of the convolution module is 224*224*3, the convolution kernel size is 3*3, the number of channels is 64, and the output size is 112*112*64; The input size of the first residual module is 112*112*64, and the output size is 56*56*128; The input size of the first attention module is 56*56*128, and the output size is 56*56*128; The input size of the second residual module is 56*56*128, and the output size is 28*28*256; The input size of the second attention module and the third attention module are both 28*28*384, and the output size is 28*28*384; The input size of the third residual module is 28*28*384, and the output size is 14*14*512; The input size of the fourth attention module is 14*14*768, and the output size is 14*14*768; The input size of the fourth residual module is 14*14*768, and the output size is 7*7*1024; The input size of the fifth residual module is 7*7*1536, and the output size is 7*7*1536.

[0012] In a second aspect, an algae recognition system based on a residual attention optimization network is provided, comprising: An image acquisition module, used for acquiring an image of algae to be identified; The algae recognition module is used to input the algae image to be identified into a pre-built algae recognition model to obtain the algae recognition result of the algae image to be identified. The algae recognition result indicates the position and type of algae in the algae image to be identified. The algae recognition model is obtained by training the residual attention optimization network using a pre-built algae dataset.

[0013] In a third aspect, a processing device is provided, comprising computer program instructions, wherein the computer program instructions, when executed by the processing device, are used to implement the steps corresponding to the above-mentioned algae recognition method based on the residual attention optimization network.

[0014] In a fourth aspect, a computer-readable storage medium is provided, on which computer program instructions are stored, wherein the computer program instructions, when executed by a processor, are used to implement the steps corresponding to the above-mentioned algae recognition method based on the residual attention optimization network.

[0015] Due to the adoption of the above technical scheme, the present invention has the following advantages: the present invention optimizes the structure of the residual attention network model according to the characteristics of small algae targets and high similarity between algae, and builds a residual attention optimization network. The improved network structure mainly transforms the dimension of the feature map by adjusting the attention module and the residual module parameters to realize the extraction of global information and original feature information. The algae recognition model obtained by training based on the network structure has a high recognition accuracy and a fast reasoning speed, and can realize automatic and efficient algae recognition. The present invention can be widely used in the field of algae recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Throughout the accompanying drawings, the same reference numerals are used to represent the same components. In the accompanying drawings: Figure 1 It is a schematic diagram of a method flow provided by an embodiment of the present invention; Figure 2 is a schematic diagram of four single algae images including different types of algae provided by an embodiment of the present invention, wherein: Figure 2 (a) is a schematic diagram of a single algae image including only Microcystis. Figure 2 (b) is a schematic diagram of a single algae image including only Longisporium. Figure 2 (c) is a schematic diagram of a single algae image including only Cyclotella. Figure 2 (d) is a schematic diagram of a single algae image including only Elbow algae; Figure 3 It is a schematic diagram of a residual attention optimization network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0018] It should be understood that the terms used in the text are only for the purpose of describing specific example embodiments, and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used in the text may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0019] Although the terms first, second, third, etc. can be used in the text to describe multiple elements, components, regions, layers and / or sections, these elements, components, regions, layers and / or sections should not be limited by these terms. These terms can only be used to distinguish an element, component, region, layer or section from another region, layer or section. Unless the context clearly indicates, terms such as "first", "second" and other numerical terms do not imply order or sequence when used in the text. Therefore, the first element, component, region, layer or section discussed below can be referred to as the second element, component, region, layer or section without departing from the teaching of the example embodiments.

[0020] At present, the traditional practice is for professional technicians to use microscopes to identify the cell morphology of algae samples to complete algae identification monitoring. This method is currently the most widely used, but it is time-consuming and labor-intensive. One sample requires 2-3 hours of manual identification, and complex samples are more troublesome. When algae blooms appear, it is impossible to quickly achieve algae identification of the entire water sample. Moreover, algae morphology is diverse, including single cells, colony structures, and filaments. Different types have different cell or colony structure characteristics, and the complexity is high. Therefore, the current manual algae identification method has low accuracy and stability. The embodiment of the present invention provides an algae identification method based on a residual attention optimization network, including: obtaining an algae image to be identified; inputting the algae image to be identified into a pre-constructed algae recognition model to obtain an algae recognition result of the algae image to be identified, the algae recognition result indicates the position and type of algae in the algae image to be identified, and the algae recognition model is obtained by training a residual attention optimization network using a pre-constructed algae data set. The present invention optimizes the structure of the residual attention network according to the characteristics of small algae targets and high similarity between algae. The structure of the constructed residual attention optimization network transforms the dimension of the feature map by adjusting the attention module and the residual module parameters, so as to realize the extraction of global information and original feature information. The output of the residual module after the attention module will be spliced ​​with the output of the residual module before the attention module according to the channel and then forward propagated. The algae recognition model trained based on the network structure has high recognition accuracy and fast reasoning speed, and can realize automatic and efficient algae recognition.

[0021] Example 1 like Figure 1 As shown, this embodiment provides an algae recognition method based on a residual attention optimization network, comprising the following steps: 1) Construct an algae dataset.

[0022] Specifically, the constructed algae dataset includes several different types of algae sample images, each of which carries the annotation information of the location and type of each algae included. For example, the algae species of the constructed algae dataset include Microcystis, Longispora, Chlorella, Cyclotella, Cryptomonas, Cryptomonas, Nitzschia, Pseudoanabaena, Chain-shaped algae and Elbow-shaped algae. The specific process of this step is: 1.1) Use a microscope to photograph water slides and obtain images of several different types of primitive algae.

[0023] 1.2) Use labelImg annotation software and refer to the VOC dataset format to annotate the algae in the original algae image.

[0024] 1.3) Perform image preprocessing on the labeled original algae images to obtain several algae sample images: 1.3.1) Since the original algae images obtained are microscopic images with large size and high pixel density, and one original algae image often includes more algae and impurities, in order to meet the requirements of subsequent network model training, each original algae image needs to be cut into several single algae images according to the annotation results, and each single algae image contains the same algae species.

[0025] 1.3.2) Perform image preprocessing on a single algae image to obtain several algae sample images.

[0026] Specifically, the single algae image is scaled into an algae sample image with a resolution of 224*224 to obtain a preliminary algae sample image. Figure 2 Shown are four algae sample images. Figure 2 The algae sample image in (a) includes only Microcystis, Figure 2 The algae sample image in (b) includes only Longispora algae. Figure 2 The algae sample image in (c) includes only Cyclotella, Figure 2 The algae sample image in (d) includes only Elbow algae.

[0027] 1.4) Since the dominant species account for a large proportion in the dataset and it is difficult to obtain algae sample images, the number of species is unbalanced, which easily leads to underfitting during model training. Therefore, after obtaining the preliminary algae sample images, it is necessary to perform image enhancement processing on the algae sample images to obtain the algae sample images after image enhancement processing, and finally construct the algae dataset. Specifically, the image enhancement processing includes at least one of rotation, random cropping, scaling and adding noise. In actual operation, it can be completed using Python's imgaug image data enhancement library.

[0028] 2) Construct a residual attention optimization network.

[0029] Specifically, the size of algae targets is generally small. For example, the diameter of a single cell of Microcystis is about 5 microns. In addition, the similarity between algae is high, and the distinguishing features between different species of the same genus are very small. Therefore, the feature extraction effect of the residual attention network is limited. Therefore, the residual attention optimization network constructed in the embodiment of the present invention is optimized based on Attention-56 and Attention-92. Figure 3The network structure shown: The constructed residual attention optimization network includes a convolution module, a first residual module, a first attention module, a second residual module, a second attention module, a third attention module, a third residual module, a fourth attention module, a fourth residual module, a fifth residual module, a pooling layer and a fully connected layer connected in sequence. The first residual module is also connected to the second attention module, the second residual module is also connected to the fourth attention module, and the third residual module is also connected to the fifth residual module. The output of the first residual module and the output of the second residual module are spliced ​​by channel and input to the second attention module, the output of the second residual module and the output of the third residual module are spliced ​​by channel and input to the fourth attention module, and the output of the third residual module and the output of the fourth residual module are spliced ​​by channel and input to the fifth residual module.

[0030] The convolution module is used for feature extraction, dimensionality reduction, denoising and enhancement of images.

[0031] Each attention module consists of a trunk branch and a soft mask branch. The trunk branch is used to extract image features, and the soft mask branch is used to perform weighted fusion of the extracted image features.

[0032] Each residual module is used to enable the deep neural network to learn more hierarchical features, thereby improving the performance of the deep neural network.

[0033] The pooling layer is used to effectively reduce the size of the parameter matrix, remove redundant information, compress features, and simplify the complexity of the network, thereby reducing the number of parameters in the last connection layer.

[0034] The fully connected layer is used for output, mapping the feature representation of the previous layer to the final algae image classification result.

[0035] Specifically, three different attention modules are used in the structure of the residual attention optimization network of the present invention: The first attention module adopts the first attention structure (A1 structure), the second and third attention modules adopt the second attention structure (A2 structure), and the fourth attention module adopts the third attention structure (A3 structure). The hierarchical structures of the soft mask branches in the three attention structures are different, that is, the number of shortcut connections is different. The hierarchical structure of the soft mask branches in each attention structure can be customized as needed.

[0036] Specifically, the structure of the residual attention optimization network constructed by the present invention optimizes the network structure by adjusting the number of attention modules and residual modules, splicing the output of the residual module after the attention module and the output of the residual module before the corresponding attention module according to the channel and then forward propagating, combining the bottom-level features and top-level features on the channel alignment fusion backbone network, which can improve the accuracy and reasoning speed of the model. In addition, the residual attention optimization network adopts the Leaky ReLU activation function, which can further improve the accuracy and reasoning speed of the model.

[0037] Specifically, the input of the constructed residual attention optimization network is an algae sample image in RGB format of 224*224, so the input size is 224*224*3. The size and parameters of each module in the residual attention optimization network are: The input size of the convolution module is 224*224*3, the convolution kernel size of the convolution module is 3*3, the number of channels is 64, and the output size is 112*112*64.

[0038] The input size of the first residual module is 112*112*64; the convolution kernel size of the first convolution layer connected in sequence in the first residual module is 1*1, and the number of channels is 32, the convolution kernel size of the second convolution layer connected in sequence is 3*3, and the number of channels is 32, the convolution kernel size of the third convolution layer connected in sequence is 1*1, and the number of channels is 128, and the three convolution layers are connected in series; the output size of the first residual module is 56*56*128.

[0039] The input size of the first attention module is 56*56*128 and the output size is 56*56*128.

[0040] The input size of the second residual module is 56*56*128; the convolution kernel size of the first convolution layer connected in the second residual module is 1*1 and the number of channels is 64, the convolution kernel size of the second convolution layer connected in sequence is 3*3 and the number of channels is 64, and the convolution kernel size of the third convolution layer connected in sequence is 1*1 and the number of channels is 256; the output size of the second residual module is 28*28*256.

[0041] The input size of the second attention module and the third attention module are both 28*28*384, and the output size is 28*28*384.

[0042] The input size of the third residual module is 28*28*384; the convolution kernel size of the first convolution layer connected in the third residual module is 1*1, and the number of channels is 128, the convolution kernel size of the second convolution layer connected in sequence is 3*3, and the number of channels is 128, and the convolution kernel size of the third convolution layer connected in sequence is 1*1, and the number of channels is 512; the output size of the third residual module is 14*14*512.

[0043] The input size of the fourth attention module is 14*14*768 and the output size is 14*14*768.

[0044] The input size of the fourth residual module is 14*14*768; the convolution kernel size of the first convolution layer connected in the fourth residual module is 1*1 and the number of channels is 256, the convolution kernel size of the second convolution layer connected in sequence is 3*3 and the number of channels is 256, and the convolution kernel size of the third convolution layer connected in sequence is 1*1 and the number of channels is 1024; the output size of the fourth residual module is 7*7*1024.

[0045] The input size of the fifth residual module is 7*7*1536; the convolution kernel size of the first convolution layer connected in the fifth residual module is 1*1 and the number of channels is 384, the convolution kernel size of the second convolution layer connected in sequence is 3*3 and the number of channels is 384, and the convolution kernel size of the third convolution layer connected in sequence is 1*1 and the number of channels is 1536; the output size of the fifth residual module is 7*7*1536.

[0046] The input size of the pooling layer pair is 7*7*1536 and the output size is 1*1*1536.

[0047] The input size of the fully connected layer is 1*1*1536, and the output size is N, where N is the number of algae species included in the constructed algae dataset. The following Table 1 takes N=10 as an example:

[0048] 3) According to the algae dataset, the constructed residual attention optimization network is trained to obtain the algae recognition model.

[0049] 4) Obtain the algae image to be identified.

[0050] 5) The algae image to be identified is input into the constructed algae identification model to obtain an algae identification result of the algae image to be identified, and the algae identification result indicates the position and type of the algae in the algae image to be identified.

[0051] 6) Considering the difficulty of acquiring algae images, after obtaining the algae recognition results of the algae images to be identified, the algae recognition results of the algae images to be identified are corrected and updated to the constructed algae dataset to supplement the algae dataset. In addition, after running for a predetermined period of time, the updated algae dataset can be used to train the algae recognition model and correct the algae recognition model.

[0052] The following is a comparison of the algae recognition results of the algae recognition model obtained by training the residual attention optimization network constructed by the present invention and the algae recognition results of the algae recognition model obtained by training the residual attention network model Attention-56 and the residual attention network model Attention-92 as shown in Table 2 below: Table 2: Comparison of algae identification results

[0053] Comparing the algae recognition results of the residual attention network models Attention-56 and Attention-92, it can be seen that Attention-92 achieves a higher accuracy than Attention-56 by increasing the number of attention modules, but the reasoning speed is reduced, and the FPS drops from 38.5 to 21.7. The algae recognition model obtained by training the residual attention optimization network constructed by the present invention not only achieves the highest recognition accuracy of 94.2%, but also has a reasoning speed of 38.3, which is comparable to the reasoning speed of the fastest Attention-56 model. That is, the network structure of the present invention achieves improved recognition accuracy without sacrificing reasoning speed, has a better comprehensive recognition effect, and has higher recognition efficiency.

[0054] Example 2 This embodiment provides an algae recognition system based on a residual attention optimization network, including: The image acquisition module is used to acquire the image of algae to be identified.

[0055] The algae recognition module is used to input the algae image to be recognized into the pre-built algae recognition model to obtain the algae recognition result of the algae image to be recognized, and the algae recognition result indicates the position and type of the algae in the algae image to be recognized.

[0056] In a preferred embodiment, a model building module is further included, which is used to train the residual attention optimization network using a pre-built algae dataset to obtain an algae recognition model, including: An algae dataset is constructed, including several algae sample images of different types, each of which carries annotation information of the location and type of each algae included.

[0057] Construct a residual attention optimization network.

[0058] According to the algae dataset, the constructed residual attention optimization network is trained to obtain the algae recognition model.

[0059] In a preferred embodiment, it also includes a model correction module, which is used to correct the algae recognition result of the algae image to be identified and update it to the constructed algae dataset after obtaining the algae recognition result of the algae image to be identified, and use the updated algae dataset to train the algae recognition model to correct the algae recognition model.

[0060] The system provided in this embodiment is used to execute the above-mentioned method embodiments. Please refer to the above-mentioned embodiments for specific processes and detailed contents, which will not be repeated here.

[0061] Example 3 This embodiment provides a processing device corresponding to the algae recognition method based on the residual attention optimization network provided in Embodiment 1. The processing device can be suitable for a client processing device, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the method of Embodiment 1.

[0062] The processing device includes a processor, a memory, a communication interface and a bus, and the processor, the memory and the communication interface are connected through the bus to complete mutual communication. The memory stores a computer program that can be run on the processing device, and when the processing device runs the computer program, the algae recognition method based on the residual attention optimization network provided in this embodiment 1 is executed.

[0063] In some implementations, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk storage.

[0064] In some other implementations, the processor may be a central processing unit (CPU), a digital signal processor (DSP), or other general-purpose processors of various types, which are not limited herein.

[0065] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0066] Those skilled in the art will understand that the structure of the above-mentioned computing device is only a partial structure related to the solution of the present invention, and does not constitute a limitation on the computing device to which the solution of the present invention is applied. The specific computing device may include more or fewer components, or combine certain components, or have a different arrangement of components.

[0067] Example 4 This embodiment provides a computer program product corresponding to the algae identification method based on the residual attention optimization network provided in Example 1. The computer program product may include a computer-readable storage medium on which computer-readable program instructions for executing the algae identification method based on the residual attention optimization network described in Example 1 are loaded.

[0068] Computer readable storage media can be tangible devices that hold and store instructions used by instruction execution devices. Computer readable storage media can be, for example, but not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any combination thereof.

[0069] The above embodiment provides a computer-readable storage medium, whose implementation principle and technical effect are similar to those of the above method embodiment, and will not be repeated here.

[0070] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0071] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0072] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0073] The above embodiments are only used to illustrate the present invention, wherein the structure, connection mode and manufacturing process of each component may be changed. Any equivalent transformations and improvements based on the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A method for algae recognition based on residual attention optimization network, characterized in that: include: Acquire an image of algae to be identified; The algae image to be identified is input into a pre-built algae recognition model to obtain an algae recognition result of the algae image to be identified. The algae recognition result indicates the position and type of algae in the algae image to be identified. The algae recognition model is obtained by training a residual attention optimization network using a pre-built algae dataset.

2. The algae recognition method based on residual attention optimization network according to claim 1, characterized in that: Also includes: After obtaining the algae recognition result of the algae image to be recognized, the algae recognition result of the algae image to be recognized is corrected and updated to the constructed algae data set, and the algae recognition model is trained using the updated algae data set to correct the algae recognition model.

3. The algae recognition method based on residual attention optimization network as claimed in claim 1, characterized in that: The construction process of the algae recognition model is as follows: Constructing an algae dataset, including a number of algae sample images of different types, each algae sample image having annotation information of the location and type of each algae included; Construct a residual attention optimization network; According to the algae dataset, the constructed residual attention optimization network is trained to obtain the algae recognition model.

4. The algae recognition method based on residual attention optimization network as claimed in claim 3, characterized in that: The algae dataset is constructed, comprising: Using microscopic imaging equipment, photograph water slide samples to obtain images of several different types of primitive algae; Annotating algae in the acquired original algae image; Perform image preprocessing on the annotated original algae images to obtain several algae sample images; The algae sample images are subjected to image enhancement processing to obtain the algae sample images after image enhancement processing, and an algae dataset is constructed.

5. The algae recognition method based on residual attention optimization network as claimed in claim 4, characterized in that: The image preprocessing is performed on the labeled original algae image to obtain a number of algae sample images, including: According to the labeling results, each original algae image is cut into a number of single algae images, and each of the cut single algae images includes the same algae species; The image preprocessing is performed on a single algae image to obtain several algae sample images.

6. The algae recognition method based on residual attention optimization network as claimed in claim 3, characterized in that: The residual attention optimization network includes a convolution module, a first residual module, a first attention module, a second residual module, a second attention module, a third attention module, a third residual module, a fourth attention module, a fourth residual module, a fifth residual module, a pooling layer and a fully connected layer connected in sequence, the first residual module is also connected to the second attention module, the second residual module is also connected to the fourth attention module, and the third residual module is also connected to the fifth residual module; the output of the first residual module and the output of the second residual module are spliced ​​by channel and input into the second attention module, the output of the second residual module and the output of the third residual module are spliced ​​by channel and input into the fourth attention module, and the output of the third residual module and the output of the fourth residual module are spliced ​​by channel and input into the fifth residual module; each attention module includes a trunk branch and a soft mask branch, the trunk branch is used to extract image features, and the soft mask branch is used to perform weighted fusion on the extracted image features.

7. The algae recognition method based on residual attention optimization network as claimed in claim 6, characterized in that: The input size of the convolution module is 224*224*3, the convolution kernel size is 3*3, the number of channels is 64, and the output size is 112*112*64; The input size of the first residual module is 112*112*64, and the output size is 56*56*128; The input size of the first attention module is 56*56*128, and the output size is 56*56*128; The input size of the second residual module is 56*56*128, and the output size is 28*28*256; The input size of the second attention module and the third attention module are both 28*28*384, and the output size is 28*28*384; The input size of the third residual module is 28*28*384, and the output size is 14*14*512; The input size of the fourth attention module is 14*14*768, and the output size is 14*14*768; The input size of the fourth residual module is 14*14*768, and the output size is 7*7*1024; The input size of the fifth residual module is 7*7*1536, and the output size is 7*7*1536.

8. An algae recognition system based on residual attention optimization network, characterized in that: include: An image acquisition module, used for acquiring an image of algae to be identified; The algae recognition module is used to input the algae image to be identified into a pre-built algae recognition model to obtain the algae recognition result of the algae image to be identified. The algae recognition result indicates the position and type of algae in the algae image to be identified. The algae recognition model is obtained by training the residual attention optimization network using a pre-built algae dataset.

9. A processing device, characterized in that: It includes computer program instructions, wherein when the computer program instructions are executed by a processing device, they are used to implement the steps corresponding to the algae recognition method based on the residual attention optimization network described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, wherein the computer program instructions, when executed by a processor, are used to implement the steps corresponding to the algae recognition method based on the residual attention optimization network described in any one of claims 1 to 7.

Images