Plankton correction method, device, equipment and storage medium
By constructing a detailed verification category dataset and using a detailed verification model for feature extraction and fusion processing, the problem of low accuracy caused by non-biological interference individuals in plankton detection was solved, achieving higher detection accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN OASIS LIGHT BIOTECHNOLOGY CO LTD
- Filing Date
- 2023-02-23
- Publication Date
- 2026-05-29
Smart Images
Figure CN116186559B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence, and in particular to a planktonic calibration method, apparatus, device, and storage medium. Background Technology
[0002] The underwater environment is complex and variable, with a variety of plankton species mixed with various interfering substances. Therefore, the in-situ images acquired by the imager will inevitably show different levels of turbidity, different densities of plankton, and different densities of interfering substances in the field of view.
[0003] Currently, existing planktonic detection and classification technologies mainly employ a single neural network architecture. This architecture typically uses multiple convolutional layers, pooling layers, and activation layers to extract features from images, then fuses these features using a feature pyramid to form Regions of Interest (ROIs). Finally, it analyzes metrics such as Intersection over Union (IOU) and confidence scores of the ROIs to obtain the detection results. However, after detection, many non-planktonic individuals are found in the results, leading to low detection accuracy in current methods. Summary of the Invention
[0004] This invention provides a plankton calibration method, apparatus, device, and storage medium for improving the accuracy of plankton target detection.
[0005] The first aspect of the present invention provides a plankton calibration method, the plankton calibration method comprising: obtaining the total number of biological categories to be calibrated, and constructing a detailed calibration category data set based on the total number of biological categories; obtaining plankton target detection results, and generating a calibration result set based on the plankton target detection results and the detailed calibration category data set; and performing a difference comparison between the plankton target detection results and the calibration result set to obtain a target result set.
[0006] Optionally, in a first implementation of the first aspect of the present invention, the step of obtaining the planktonic target detection result and generating a calibration result set based on the planktonic target detection result and the detail calibration category data set includes: obtaining the planktonic target detection result; inputting the planktonic target detection result into a preset detail calibration model for set processing to obtain an initial result set; and generating a calibration result set based on the detail calibration category data set and the initial result set.
[0007] Optionally, in a second implementation of the first aspect of the present invention, the step of inputting the planktonic target detection results into a preset detail correction model for set processing to obtain an initial result set includes: inputting the planktonic target detection results into a preset detail correction model, wherein the detail correction model includes: a first convolutional network, a feature extraction network, a second convolutional network, and a fully connected network; and performing set fusion processing on the planktonic target detection results through the detail correction model to obtain an initial result set.
[0008] Optionally, in a third implementation of the first aspect of the present invention, the step of performing set fusion processing on the plankton target detection results through the detail correction model to obtain an initial result set includes: performing convolution operations on the plankton target detection results through a first convolutional network in the detail correction model to obtain a first convolutional feature; inputting the first convolutional feature into the feature extraction network for feature extraction processing to obtain target feature data; inputting the target feature data into a second convolutional network for convolution operations to obtain a second convolutional feature; and inputting the second convolutional feature into the fully connected network for set fusion processing to output the initial result set.
[0009] Optionally, in a fourth implementation of the first aspect of the present invention, the step of inputting the first convolutional feature into the feature extraction network for feature extraction processing to obtain target feature data includes: inputting the first convolutional feature into the feature extraction network, wherein the feature extraction network includes: a first feature extraction layer and a second feature extraction layer; performing convolutional feature operations on the first convolutional feature through the first feature extraction layer to obtain initial feature data; and performing feature normalization processing on the initial feature data through the second feature extraction layer to obtain target feature data.
[0010] Optionally, in a fifth implementation of the first aspect of the present invention, the step of performing convolutional feature operations on the first convolutional feature through the first feature extraction layer to obtain initial feature data includes: inputting the first convolutional feature into the first feature extraction layer, wherein the first feature extraction layer includes: a first feature processing block and a second feature processing block; and processing the first convolutional feature through the first feature processing block and the second feature processing block to obtain initial feature data.
[0011] Optionally, in a sixth implementation of the first aspect of the present invention, the step of performing a difference comparison between the plankton target detection result and the calibration result set to obtain a target result set includes: performing probability calculations on the plankton target detection result and the calibration result set to obtain a first element probability and a second element probability; comparing the first element probability and the second element probability according to a preset calibration function to obtain abnormal objects in the plankton target detection result; and standardizing the abnormal objects in the plankton target detection result to obtain the target result set.
[0012] A second aspect of the present invention provides a plankton calibration device, the plankton calibration device comprising: an acquisition module, configured to acquire the total number of biological categories to be calibrated, and construct a detailed calibration category data set based on the total number of biological categories; a processing module, configured to acquire plankton target detection results, and generate a calibration result set based on the plankton target detection results and the detailed calibration category data set; and a comparison module, configured to perform a difference comparison between the plankton target detection results and the calibration result set to obtain a target result set.
[0013] Optionally, in a first implementation of the second aspect of the present invention, the processing module further includes: an acquisition submodule for acquiring planktonic target detection results; a processing submodule for inputting the planktonic target detection results into a preset detail correction model for set processing to obtain an initial result set; and a generation submodule for generating a correction result set based on the detail correction category data set and the initial result set.
[0014] Optionally, in a second implementation of the second aspect of the present invention, the processing submodule further includes: an input unit, used to input the plankton target detection results into a preset detail correction model, wherein the detail correction model includes: a first convolutional network, a feature extraction network, a second convolutional network, and a fully connected network; and a fusion unit, used to perform set fusion processing on the plankton target detection results through the detail correction model to obtain an initial result set.
[0015] Optionally, in a third implementation of the second aspect of the present invention, the fusion unit further includes: a first computation subunit, configured to perform convolution operations on the plankton target detection results through a first convolutional network in the detail correction model to obtain first convolutional features; a feature extraction subunit, configured to input the first convolutional features into the feature extraction network for feature extraction processing to obtain target feature data; a second computation subunit, configured to input the target feature data into the second convolutional network for convolution operations to obtain second convolutional features; and a fusion processing subunit, configured to input the second convolutional features into the fully connected network for set fusion processing to output an initial result set.
[0016] Optionally, in a fourth implementation of the second aspect of the present invention, the feature extraction subunit is specifically used to: input the first convolutional feature into the feature extraction network, wherein the feature extraction network includes: a first feature extraction layer and a second feature extraction layer; perform convolutional feature operations on the first convolutional feature through the first feature extraction layer to obtain initial feature data; and perform feature normalization processing on the initial feature data through the second feature extraction layer to obtain target feature data.
[0017] Optionally, in a fifth implementation of the second aspect of the present invention, the feature extraction subunit is further configured to: input the first convolutional feature into the first feature extraction layer, wherein the first feature extraction layer includes: a first feature processing block and a second feature processing block; process the first convolutional feature through the first feature processing block and the second feature processing block to obtain initial feature data.
[0018] Optionally, in a sixth implementation of the second aspect of the present invention, the comparison module is specifically used for: performing probability calculations on the plankton target detection results and the calibration result set to obtain a first element probability and a second element probability; comparing the first element probability and the second element probability according to a preset calibration function to obtain abnormal objects in the plankton target detection results; and standardizing the abnormal objects in the plankton target detection results to obtain a target result set.
[0019] A third aspect of the present invention provides a plankton calibration device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the plankton calibration device to perform the plankton calibration method described above.
[0020] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the above-described plankton calibration method.
[0021] The technical solution provided by this invention involves obtaining the total number of biological categories to be calibrated and constructing a detailed calibration category data set based on the total number of biological categories; obtaining the planktonic target detection results and generating a calibration result set based on the planktonic target detection results and the detailed calibration category data set; and performing a difference comparison between the planktonic target detection results and the calibration result set to obtain a target result set. This invention reduces the interference of non-biological factors on the identification results by calibrating and correcting the target detection results, removes the non-biological result part, ensures the accuracy of planktonic identification, and solves the problem of high false positive rate in planktonic identification. Attached Figure Description
[0022] Figure 1 This is a schematic diagram of one embodiment of the plankton calibration method in this invention;
[0023] Figure 2 This is a schematic diagram of another embodiment of the plankton calibration method in this invention;
[0024] Figure 3 This is a schematic diagram of one embodiment of the plankton calibration device in this invention;
[0025] Figure 4 This is a schematic diagram of another embodiment of the plankton calibration device in this invention;
[0026] Figure 5 This is a schematic diagram of one embodiment of the plankton calibration device in this invention. Detailed Implementation
[0027] This invention provides a plankton calibration method, apparatus, device, and storage medium to improve the accuracy of plankton target detection. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0028] It should be noted that the vocabulary High-Conc, Noctiluca, Phaeocystis, Pteropoda, Shrimp, and Low-den represent high turbidity scenes, Noctiluca scenes, Phaeocystis scenes, pteropod scenes, shrimp scenes, and low turbidity scenes, respectively; and the vocabulary appendicularia, chaetognatha, copepoda, creseis, echinodermata, fish, larval_Fish, lyngbya, medusae, noctiluca, phaeocystis, polychaete, scylla, shrimp, skeletononema, spiral_diatom, and thaliacea represent planktonic organisms such as appendicularia, arrow worms, copepods, pencapsule snails, echinoderm larvae, fish, juvenile fish, filamentous algae, hydroids, Noctiluca, Phaeocystis, polychaetes, blue crabs, shrimp, ribbed algae, spirulina, and sea salp, respectively.
[0029] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the plankton calibration method in this invention includes:
[0030] 101. Obtain the total number of biological categories to be proofread, and construct a detailed proofreading category data set based on the total number of biological categories;
[0031] It is understood that the executing entity of this invention can be a plankton calibration device, a terminal, or a server; no specific limitation is made here. This embodiment of the invention will be described using a server as an example.
[0032] Specifically, a detailed proofreading category dataset D = (D1, D2, ..., Dk) is constructed. It should be noted that k is the total number of all biological categories that need to be proofread. In this application, there are 7 biological categories: chaetognatha, creseis, fish, medusae, noctiluca, phaeocystis, and shrimp. The total number of data for these 7 biological categories is 21,267, including 475 data for chaetognatha, 3,244 for creseis, 3,092 for fish, 3,872 for medusae, 3,053 for noctiluca, 3,646 for phaeocystis, and 3,885 for shrimp.
[0033] 102. Obtain the planktonic target detection results, and generate a set of correction results based on the planktonic target detection results and the detailed correction category data set;
[0034] Specifically, the server acquires the planktonic target detection results and constructs a set of calibration results based on these results. Specifically, the server performs information detection on the planktonic target detection results, using the detected planktonic target detection results as information data. The information data is then calibrated using an information data calibration function. Simultaneously, the server constructs a text recognition function and uses it to perform text recognition on the calibrated information data and the planktonic target detection results that did not detect information, thereby generating the set of calibration results.
[0035] 103. Compare the differences between the detection results and the calibration results set of planktonic targets to obtain the target result set.
[0036] Specifically, upon receiving a difference comparison instruction, the system obtains the data format information of the plankton target detection results and the calibration result set. Based on the format information, it determines whether the text information can be directly obtained according to the corresponding encoding format. If so, the data content is read directly; otherwise, the system calls the optical character recognition model to obtain the data content, forming data content data. The system then compares the data content of the plankton target detection results and the calibration result set to obtain the target result set.
[0037] In this embodiment of the invention, the total number of biological categories to be calibrated is obtained, and a detailed calibration category data set is constructed based on the total number of biological categories; the planktonic target detection results are obtained, and a calibration result set is generated based on the planktonic target detection results and the detailed calibration category data set; the planktonic target detection results and the calibration result set are compared for differences to obtain the target result set. This invention reduces the interference of non-biological factors on the recognition results by calibrating and correcting the target detection results, removes the non-biological result part, ensures the accuracy of planktonic identification, and solves the problem of high false positive rate in planktonic identification.
[0038] Please see Figure 2 Another embodiment of the plankton calibration method in this invention includes:
[0039] 201. Obtain the total number of biological categories to be proofread, and construct a detailed proofreading category dataset based on the total number of biological categories;
[0040] 202. Obtain the detection results of planktonic targets;
[0041] 203. Input the planktonic target detection results into the preset detail correction model for set processing to obtain the initial result set;
[0042] Specifically, the planktonic target detection results are input into a pre-set detail correction model, which includes a first convolutional network, a feature extraction network, a second convolutional network, and a fully connected network. The detail correction model is used to perform set fusion processing on the planktonic target detection results to obtain an initial result set.
[0043] The planktonic target detection results are input into the detail correction model to obtain the correction result set DR = (DR1, DR2, ..., DRk). By comparing the differences between the sets, non-living things in the planktonic target detection results R are removed, and the incorrectly identified organisms in the planktonic target detection results are corrected.
[0044] The detail verification model employs a neural network classifier, whose input is the OR of the detection results from the object detection module. Its verification process involves comparing the differences between sets to correct them. Specifically, this application selects five organisms—copepoda, medusae, noctiluca, phaeocystis, and shrimp—for testing. A total of 240 test images were collected. Manually counted images showed 193 copepoda, 94 medusae, 1845 noctiluca, 968 phaeocystis, and 171 shrimp. The scene_model detected 376 copepoda, 98 medusae, 1926 noctiluca, 983 phaeocystis, and 617 shrimp. Of these, 178 were genuine copepoda, 198 were pseudo-copepoda, 90 were genuine medusae, 8 were pseudo-medusae, and the remaining 10 were pseudo-noctiluca. There were 1837 true *Iluca*, 89 false *Noctiluca*, 961 true *Phaeocystis*, 22 false *Phaeocystis*, 144 true *Shrimp*, and 473 false *Shrimp*. After adding the detail correction model, the `detail_model` was called to correct the planktonic target detection results. The corrected results showed 185 true *Copepoda*, 4 false *Copepoda*, 92 true *Medusa*, 3 false *Medusa*, 1840 true *Noctiluca*, 16 false *Noctiluca*, 960 true *Phaeocystis*, 10 false *Phaeocystis*, 164 true *Shrimp*, and 3 false *Shrimp*. In the data statistics, without the addition of the proofreading model, the copepoda and shrimp organisms had a high false positive rate, with the copepoda organism having a false positive rate as high as 103% and the shrimp organism having a false positive rate as high as 277%. However, after adding the detail proofreading model, the scene_model_detail complex neural network architecture had a false positive rate of only 2% for the copepoda organism and only 1.8% for the shrimp organism.
[0045] Specifically, the first convolutional network in the detail calibration model performs convolution operations on the plankton target detection results to obtain the first convolutional features; the first convolutional features are input into the feature extraction network for feature extraction processing to obtain target feature data; the target feature data are input into the second convolutional network for convolution operations to obtain the second convolutional features; the second convolutional features are input into the fully connected network for set fusion processing to output the initial result set.
[0046] The process involves obtaining planktonic target detection results, performing comprehensive preprocessing on fault data, extracting features based on convolutional neural networks in a convolutional neural network autoencoder to obtain the first convolutional feature, extracting temporal features based on expert knowledge in an artificial temporal feature extraction module to obtain target feature data, concatenating the first convolutional feature and the target feature data, and performing convolution operations to obtain the second convolutional feature. The second convolutional feature is then input into a fully connected network for set fusion processing to output the initial result set.
[0047] Specifically, the first convolutional feature is input into the feature extraction network, which includes a first feature extraction layer and a second feature extraction layer. The first feature extraction layer performs convolutional feature operations on the first convolutional feature to obtain initial feature data. The second feature extraction layer performs feature normalization processing on the initial feature data to obtain target feature data.
[0048] Optionally, the server obtains the first convolutional features, establishes initial attribute categories based on the initial attribute cluster centers, performs merging and separation operations on all initial attribute categories to obtain clustered attribute categories, calculates the dependency of each attribute data in the clustered attribute categories relative to other attribute data, selects the highest dependency value as the feature attribute value, and selects and aggregates the feature attribute values of all attribute categories to obtain the target feature data.
[0049] Specifically, the first convolutional features are input into the first feature extraction layer, which includes a first feature processing block and a second feature processing block; the first convolutional features are processed by the first feature processing block and the second feature processing block to obtain initial feature data.
[0050] 204. Generate a proofreading result set based on the detailed proofreading category data set and the initial result set;
[0051] 205. Compare the differences between the target detection results and the calibration results set of planktonic organisms to obtain the target result set.
[0052] Specifically, probability calculations are performed on the set of plankton target detection results and calibration results to obtain the first element probability and the second element probability; the first element probability and the second element probability are compared according to a preset calibration function to obtain the abnormal objects in the plankton target detection results; the abnormal objects in the plankton target detection results are standardized to obtain the target result set.
[0053] Specifically, the server performs probabilistic calculations on the plankton target detection results and the calibration result set. Specifically, the server queries the cache based on the probability query keyword. The cache stores intermediate calculation results from previous probability query processes. If there are intermediate calculation results in the cache that match the joint values of multiple variables, the intermediate calculation results are used as the query results for the probability query, obtaining the first element probability and the second element probability. The first element probability and the second element probability are compared according to a preset calibration function to obtain the abnormal objects in the plankton target detection results. The abnormal objects in the plankton target detection results are then standardized to obtain the target result set.
[0054] In this embodiment of the invention, the total number of biological categories to be calibrated is obtained, and a detailed calibration category data set is constructed based on the total number of biological categories; the planktonic target detection results are obtained, and a calibration result set is generated based on the planktonic target detection results and the detailed calibration category data set; the planktonic target detection results and the calibration result set are compared for differences to obtain the target result set. This invention reduces the interference of non-biological factors on the recognition results by calibrating and correcting the target detection results, removes the non-biological result part, ensures the accuracy of planktonic identification, and solves the problem of high false positive rate in planktonic identification.
[0055] The plankton calibration method in the embodiments of the present invention has been described above. The plankton calibration device in the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 3 One embodiment of the plankton calibration device in this invention includes:
[0056] The acquisition module 301 is used to acquire the total number of biological categories to be verified, and to construct a detailed verification category data set based on the total number of biological categories;
[0057] The processing module 302 is used to acquire the planktonic target detection results and generate a set of correction results based on the planktonic target detection results and the detailed correction category data set;
[0058] The comparison module 303 is used to compare the differences between the planktonic target detection results and the calibration result set to obtain the target result set.
[0059] In this embodiment of the invention, the total number of biological categories to be calibrated is obtained, and a detailed calibration category data set is constructed based on the total number of biological categories; the planktonic target detection results are obtained, and a calibration result set is generated based on the planktonic target detection results and the detailed calibration category data set; the planktonic target detection results and the calibration result set are compared for differences to obtain a target result set. This invention reduces the interference of non-biological factors on the recognition results by calibrating and correcting the target detection results, removes the non-biological result part, ensures the accuracy of planktonic identification, and solves the problem of high false positive rate in planktonic identification.
[0060] Please see Figure 4 Another embodiment of the plankton calibration device in this invention includes:
[0061] The acquisition module 301 is used to acquire the total number of biological categories to be verified, and to construct a detailed verification category data set based on the total number of biological categories;
[0062] The processing module 302 is used to acquire the planktonic target detection results and generate a set of correction results based on the planktonic target detection results and the detailed correction category data set;
[0063] The comparison module 303 is used to compare the differences between the planktonic target detection results and the calibration result set to obtain the target result set.
[0064] Optionally, the processing module 302 further includes:
[0065] The acquisition submodule 3021 is used to acquire the detection results of planktonic targets.
[0066] The processing submodule 3022 is used to input the planktonic target detection results into a preset detail correction model for set processing to obtain an initial result set;
[0067] The generation submodule 3023 is used to generate a set of proofreading results based on the detailed proofreading category data set and the initial result set.
[0068] Optionally, the processing submodule 3022 further includes:
[0069] The input unit is used to input the planktonic target detection results into a preset detail correction model, wherein the detail correction model includes: a first convolutional network, a feature extraction network, a second convolutional network, and a fully connected network;
[0070] The fusion unit is used to perform set fusion processing on the planktonic target detection results through the detailed correction model to obtain an initial result set.
[0071] Optionally, the fusion unit further includes:
[0072] The first computational subunit is used to perform convolution operations on the planktonic target detection results through the first convolutional network in the detail correction model to obtain the first convolutional feature;
[0073] The feature extraction subunit is used to input the first convolutional feature into the feature extraction network for feature extraction processing to obtain target feature data;
[0074] The second operation subunit is used to input the target feature data into the second convolutional network to perform convolution operations and obtain the second convolutional feature;
[0075] The fusion processing subunit is used to input the second convolutional features into the fully connected network for set fusion processing and output an initial result set.
[0076] Optionally, the feature extraction subunit is specifically used for:
[0077] The first convolutional feature is input into the feature extraction network, wherein the feature extraction network includes: a first feature extraction layer and a second feature extraction layer; the first feature extraction layer performs convolutional feature operations on the first convolutional feature to obtain initial feature data; the second feature extraction layer performs feature normalization processing on the initial feature data to obtain target feature data.
[0078] Optionally, the feature extraction subunit is further configured to:
[0079] The first convolutional feature is input into the first feature extraction layer, wherein the first feature extraction layer includes a first feature processing block and a second feature processing block; the first convolutional feature is processed by the first feature processing block and the second feature processing block to obtain initial feature data.
[0080] Optionally, the comparison module 303 is specifically used for:
[0081] Probability calculations are performed on the plankton target detection results and the calibration result set to obtain the first element probability and the second element probability; the first element probability and the second element probability are compared according to a preset calibration function to obtain the abnormal objects in the plankton target detection results; the abnormal objects in the plankton target detection results are standardized to obtain the target result set.
[0082] In this embodiment of the invention, the total number of biological categories to be calibrated is obtained, and a detailed calibration category data set is constructed based on the total number of biological categories; the planktonic target detection results are obtained, and a calibration result set is generated based on the planktonic target detection results and the detailed calibration category data set; the planktonic target detection results and the calibration result set are compared for differences to obtain a target result set. This invention reduces the interference of non-biological factors on the recognition results by calibrating and correcting the target detection results, removes the non-biological result part, ensures the accuracy of planktonic identification, and solves the problem of high false positive rate in planktonic identification.
[0083] above Figure 3 and Figure 4 The planktonic calibration device in this embodiment of the invention is described in detail from the perspective of modular functional entities. The planktonic calibration equipment in this embodiment of the invention is described in detail below from the perspective of hardware processing.
[0084] Figure 5 This is a schematic diagram of a plankton calibration device 500 provided in an embodiment of the present invention. The plankton calibration device 500 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 510 (e.g., one or more processors) and a memory 520, and one or more storage media 530 (e.g., one or more mass storage devices) for storing application programs 533 or data 532. The memory 520 and storage media 530 can be temporary or persistent storage. The program stored in the storage media 530 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the plankton calibration device 500. Furthermore, the processor 510 may be configured to communicate with the storage media 530 and execute the series of instruction operations in the storage media 530 on the plankton calibration device 500.
[0085] The plankton calibration device 500 may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input / output interfaces 560, and / or one or more operating systems 531, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 5 The illustrated planktonic calibration device structure does not constitute a limitation on the planktonic calibration device, and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.
[0086] The present invention also provides a plankton calibration device, which includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor performs the steps of the plankton calibration method described in the above embodiments.
[0087] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the plankton calibration method.
[0088] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0089] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0090] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for calibrating plankton, characterized in that, The planktonic calibration method includes: Obtain the total number of biological categories to be calibrated, and construct a detailed calibration category data set based on the total number of biological categories; Obtaining planktonic target detection results and generating a set of correction results based on the planktonic target detection results and the detailed correction category data set; specifically including: obtaining planktonic target detection results; inputting the planktonic target detection results into a preset detailed correction model, wherein the detailed correction model includes: a first convolutional network, a feature extraction network, a second convolutional network, and a fully connected network; performing convolution operations on the planktonic target detection results through the first convolutional network in the detailed correction model to obtain first convolutional features; inputting the first convolutional features into the feature extraction network, wherein the feature extraction network includes: a first feature extraction layer and a second feature extraction layer; and inputting the planktonic target detection results into the feature extraction network. The first convolutional feature is input into the first feature extraction layer, wherein the first feature extraction layer includes: a first feature processing block and a second feature processing block; the first convolutional feature is processed by the first feature processing block and the second feature processing block to obtain initial feature data; the initial feature data is normalized by the second feature extraction layer to obtain target feature data; the target feature data is input into the second convolutional network for convolution operation to obtain second convolutional features; the second convolutional features are input into the fully connected network for set fusion processing to output an initial result set; a correction result set is generated based on the detailed correction category data set and the initial result set; The discrepancy between the plankton target detection results and the calibration result set is compared to obtain a target result set. Specifically, this includes: performing probability calculations on the plankton target detection results and the calibration result set to obtain a first element probability and a second element probability; comparing the first element probability and the second element probability according to a preset calibration function to obtain abnormal objects in the plankton target detection results; and standardizing the abnormal objects in the plankton target detection results to obtain the target result set.
2. A planktonic calibration device, characterized in that, For performing the plankton calibration method as described in claim 1, the plankton calibration device comprises: The acquisition module is used to acquire the total number of biological categories to be calibrated, and to construct a detailed calibration category data set based on the total number of biological categories; The processing module is used to acquire the planktonic target detection results and generate a set of correction results based on the planktonic target detection results and the detailed correction category data set; The comparison module is used to compare the differences between the planktonic target detection results and the calibration result set to obtain the target result set.
3. A plankton calibration device, characterized in that, The planktonic calibration device includes: a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the plankton calibration device to perform the plankton calibration method as described in claim 1.
4. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the planktonic calibration method as described in claim 1.