Remote identification and identification platform based on alien invasive organisms

Abstract

The invention discloses a remote identification and identification platform based on alien invasive organisms. To-be-detected image data is obtained through a client terminal and preprocessed; the system screens alien species big data from a historical image library, extracts multi-dimensional biological image features, and constructs a biological feature knowledge graph. The method comprises the following steps: performing biological region positioning and image division at different view angles on an image to be detected, and extracting features by using PCA and Canny operators to form an image feature set; and importing the feature set into a knowledge graph, analyzing relevance, and performing classification to obtain feature validity of the main image, the auxiliary image and different images. And based on the feature validity, introducing an attention mechanism to fuse the main and auxiliary image features, and generating a fusion feature set. And finally, the client terminal uploads data such as the fusion feature set to the cloud platform for remote identification, and a result is fed back to the client terminal. According to the invention, the identification efficiency and the identification effect of the identification platform are effectively improved, and the purpose of optimizing the identification process is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of biological image detection and identification, and more particularly to a remote identification and identification platform based on alien invasive organisms. BACKGROUND

[0002] Alien invasive organisms refer to non-native natural occurrence or introduction that escapes to the natural environment and causes negative effects on the ecological system, species diversity and human social and economic activities. Alien invasive organisms generally pose a great threat to the local ecological system, not only destroying the ecological balance, but also leading to a decrease in species diversity, and even having adverse effects on agricultural production, breeding production, etc. Therefore, timely and accurate identification and identification of alien invasive organisms is the premise of effective management and control.

[0003] Traditional alien invasive organism identification methods mainly rely on manual identification, which not only requires rich professional knowledge and experience, but also consumes time and effort, making it difficult to meet the needs of large-scale and high-frequency monitoring. Moreover, traditional identification platforms often directly identify based on raw image data, which is inefficient and puts a lot of pressure on cloud platforms to process data, resulting in unsatisfactory identification efficiency and effectiveness. Furthermore, the original image features and associated information are not fully exploited, and the process of analyzing biological image features and corresponding correlations from different view angles is lacking, making it difficult to improve the efficiency of the identification platform.

[0004] Therefore, there is an urgent need for an efficient, accurate and real-time alien invasive organism identification and identification platform to achieve rapid response and effective management of invasive organisms. SUMMARY

[0005] The present application overcomes the defects of the prior art and provides a remote identification and identification platform based on alien invasive organisms.

[0006] The present application provides a remote identification and identification method based on alien invasive organisms, comprising:

[0007] Through the client terminal, the image data of the alien invasive organism to be tested is obtained, and the image data to be tested is preprocessed;

[0008] Alien species big data is filtered from the historical image database, multi-dimensional biological image feature extraction is performed on the alien species big data, and an entity data is set based on each biological image. The feature extraction process includes the characteristics of each part of the species, analyzes and stores the correlation between the characteristics, forms relationship data and attribute data, and constructs a biological feature knowledge graph;

[0009] The image data to be measured is subjected to biological region positioning and biological view angle analysis, a plurality of unit images are divided based on the view angle, and multi-dimensional feature extraction is performed on the plurality of unit images based on the PCA analysis method and the Canny operator to form an image feature set;

[0010] The image feature set is imported as original knowledge data into the biological feature knowledge graph, and the unit image is taken as the research object, and the feature knowledge analysis and correlation analysis are performed in combination with the original knowledge data, the feature correlation between the unit images is calculated, and the feature validity is generated through the correlation strength, the unit images are classified according to the feature validity, and a main image and a plurality of sub-images are obtained;

[0011] Through the feature validity index, the image features corresponding to the main image are respectively fused with the image features corresponding to the plurality of sub-images, the attention mechanism is introduced in the fusion process, the feature validity is taken as the attention weight for fusion, and a fusion feature set is generated;

[0012] Through the customer terminal, the main image, the sub-image and the fusion feature set are sent to the cloud platform for remote identification, and the identification result is sent to the customer terminal.

[0013] In the scheme, the customer terminal is used to obtain the image data to be measured of the alien invasive organism, and the image data to be measured is preprocessed, specifically:

[0014] The customer terminal is used to obtain the real-time uploaded data of the user, the uploaded data is subjected to data analysis, the image data is extracted, and the image data to be measured of the alien invasive organism is obtained;

[0015] The image data to be measured is subjected to noise reduction, enhancement and standardization preprocessing.

[0016] In the scheme, the alien species big data is screened out from the historical image database, multi-dimensional biological image feature extraction is performed on the alien species big data, and an entity data is set based on each biological image, the feature extraction process includes the features of each part of the species, the correlation between the features is analyzed and stored, the relationship data and the attribute data are formed, and the biological feature knowledge graph is constructed, specifically:

[0017] Based on the system database, the alien species image data is retrieved from the historical image database, and the alien species big data is formed;

[0018] For the alien species big data, an entity data is set based on each biological image record, and the entity data includes species identification information and biological name;

[0019] The biological image set is acquired based on a biological image record, feature extraction is performed on the biological image set by a PCA analysis method, main feature data is formed, edge detection and contour feature analysis are performed based on a Canny operator, and texture features and contour features are extracted;

[0020] The main feature data, the texture features and the contour features are used as multi-dimensional features of each entity data;

[0021] According to the biological image record, correlation information between parts of a biological body in the biological image set is acquired, the correlation information is associated with the multi-dimensional features, and internal correlation information is obtained, information storage is performed based on a recognition relationship and a biological classification relationship between the biological image records, and external correlation information is formed;

[0022] Relationship data of the entity data is generated based on the internal correlation information and the external correlation information;

[0023] The entity data is used as a node, the multi-dimensional features are used as node data, the relationship data is used as node edge information, and biological recognition information of each entity data is used as attribute data, and a biological feature knowledge graph is constructed.

[0024] In the scheme, the biological region positioning and the biological view angle analysis are performed on the to-be-tested image data, a plurality of unit images are divided based on the view angle, multi-dimensional feature extraction is performed on the plurality of unit images based on the PCA analysis method and the Canny operator, and an image feature set is formed, and the specific process is as follows:

[0025] The biological region positioning is performed on the to-be-tested image data, a biological image region is obtained, a view angle of the biological image region is determined, and a plurality of unit images are divided based on different view angles;

[0026] Multi-dimensional feature extraction is performed on the plurality of unit images based on the PCA analysis method and the Canny operator, and multi-dimensional features of different view angles are integrated, and an image feature set is obtained.

[0027] In the scheme, the image feature set is used as original knowledge data and is imported into the biological feature knowledge graph, the unit images are used as research objects, feature knowledge analysis and correlation analysis are performed in combination with the original knowledge data, feature correlation between the unit images is calculated, feature effectiveness is generated through correlation strength, the unit images are classified according to the feature effectiveness, a main image and a plurality of sub-images are obtained, and the specific process is as follows:

[0028] The image feature set is used as original knowledge data and is imported into the biological feature knowledge graph, and N entity node data is generated based on features of unit images corresponding to different view angles in the image feature set;

[0029] Take the entity node data as the analysis node, in the biometric knowledge graph, the similarity of the analysis node and the existing node is evaluated, and the most similar existing node to the analysis node is marked to obtain N first nodes;

[0030] Map the N first nodes to the corresponding N entity node data, take the N first nodes as the analysis object, select one first node to calculate its relevance, calculate the average node distance between the one first node and the remaining first nodes in the knowledge graph based on the edge distance, analyze the feature relevance intensity through the average node distance, and calculate the feature effective degree index based on the average node distance, the feature effective degree is inversely proportional to the average node distance;

[0031] Through the calculation result, set the corresponding feature effective degree for each unit image, sort the feature effective degrees based on the size, and classify the unit images to obtain a main image and multiple sub-images.

[0032] In the scheme, the main image corresponding image features are respectively fused with the multiple sub-image corresponding image features through the feature effective degree index, the attention mechanism is introduced in the fusion process, the feature effective degree is taken as the attention weight for fusion, and a fusion feature set is generated, specifically:

[0033] Based on the feature effective degree index, the attention weight is given to the main image and the sub-image;

[0034] Introduce the attention mechanism, convert the multi-dimensional features corresponding to the main image and the multi-dimensional features corresponding to one sub-image into feature maps to form a first feature map and a second feature map, weight the first feature map and the second feature map through the corresponding attention weight, and generate a fusion feature;

[0035] Fuse the main image corresponding image features with the multiple sub-image corresponding image features respectively, and integrate the fusion features to obtain a fusion feature set.

[0036] In the scheme, the main image, the sub-image and the fusion feature set are sent to the cloud platform for remote identification through the customer terminal, and the identification result is sent to the customer terminal, specifically:

[0037] In the cloud platform, collect the main image, the sub-image and the fusion feature set sent by the customer terminal;

[0038] Based on the recognition model in the cloud platform, perform a recognition based on the main image and the sub-image, perform a secondary recognition based on the fusion feature set, comprehensively evaluate the recognition result and generate an identification report;

[0039] The identification report is sent to the customer terminal in real time.

[0040] In the scheme, the cloud platform adopts a distributed architecture, deploys services in multiple server nodes, and is connected with the client terminal through Ethernet or a wireless network.

[0041] The second aspect of the application also provides a remote identification and identification platform based on alien invasive organisms, which comprises a memory and a processor, the memory comprises a remote identification and identification program based on alien invasive organisms, and the remote identification and identification program based on alien invasive organisms realizes the following steps when executed by the processor.

[0042] Through the client terminal, obtain the to-be-tested image data of the alien invasive organisms, and pre-process the to-be-tested image data;

[0043] Filter out alien species big data from the historical image database, perform multi-dimensional biological image feature extraction on the alien species big data, and set an entity data based on each biological image, the feature extraction process includes the characteristics of each part of the species, analyzes and stores the correlation between the characteristics, forms relationship data and attribute data, and constructs a biological feature knowledge graph;

[0044] Perform biological region positioning and biological view angle analysis on the to-be-tested image data, divide multiple unit images based on the view angle, perform multi-dimensional feature extraction on the multiple unit images based on the PCA analysis method and the Canny operator, and form an image feature set;

[0045] Import the image feature set as original knowledge data into the biological feature knowledge graph, take the unit image as the research object, perform feature knowledge analysis and correlation analysis on the original knowledge data, calculate the feature correlation between the unit images, generate feature effectiveness through the correlation strength, classify the unit images according to the feature effectiveness, and obtain a main image and multiple sub-images;

[0046] Through the feature effectiveness index, perform feature fusion on the image features corresponding to the main image and the image features corresponding to the multiple sub-images respectively, introduce an attention mechanism in the fusion process, take the feature effectiveness as the attention weight for fusion, and generate a fusion feature set;

[0047] Through the client terminal, send the main image, the sub-image and the fusion feature set to the cloud platform for remote identification, and send the identification result to the client terminal.

[0048] The third aspect of the application also provides a computer readable storage medium, which comprises a remote identification and identification program based on alien invasive organisms, and the remote identification and identification program based on alien invasive organisms realizes the steps of the remote identification and identification method based on alien invasive organisms as described in any one of the above aspects when executed by the processor.

[0049] The application discloses a remote identification and identification platform based on alien invasive organisms. Image data to be measured is obtained and preprocessed through a client terminal. The system screens alien species big data from a historical image library, extracts multi-dimensional biological image features, and constructs a biological feature knowledge graph. The biological region positioning and different view angle image division are performed on the image to be measured, the features are extracted using PCA and Canny operator, and the image feature set is formed. The feature set is imported into the knowledge graph, the correlation is analyzed, and the main image and the auxiliary image and the feature effectiveness of different images are classified. Based on the feature effectiveness, the attention mechanism is introduced to fuse the main and auxiliary image features, and the fusion feature set is generated. Finally, the fusion feature set and other data are uploaded to the cloud platform through the client terminal for remote identification, and the result is fed back to the client terminal. Through the application, the identification efficiency and effect of the identification platform are effectively improved, and the purpose of optimizing the identification process is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 A flowchart of a remote identification and identification method based on alien invasive organisms is shown.

[0051] Figure 2 A block diagram of a remote identification and identification platform based on alien invasive organisms is shown. DETAILED DESCRIPTION

[0052] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the present application will be further described in detail below in combination with the drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0053] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, however, the present application can also be implemented in other ways different from those described herein, therefore, the protection scope of the present application is not limited by the specific embodiments disclosed below.

[0054] Figure 1 A flowchart of a remote identification and identification method based on alien invasive organisms is shown.

[0055] As Figure 1 shown, the first aspect of the present application provides a remote identification and identification method based on alien invasive organisms, comprising:

[0056] S102, through a client terminal, obtaining image data to be measured of alien invasive organisms, and preprocessing the image data to be measured;

[0057] S104, foreign species big data is screened out from a historical image database, multi-dimensional biological image feature extraction is performed on the foreign species big data, an entity data is set based on each biological image, the feature extraction process includes part features of the species, the correlation between the features is analyzed and stored, relationship data and attribute data are formed, and a biological feature knowledge graph is constructed;

[0058] S106, biological region positioning and biological view angle analysis are performed on the to-be-tested image data, a plurality of unit images are divided based on the view angle, multi-dimensional feature extraction is performed on the plurality of unit images based on a PCA analysis method and a Canny operator, and an image feature set is formed;

[0059] S108, the image feature set is imported into the biological feature knowledge graph as original knowledge data, the unit images are taken as research objects, feature knowledge analysis and correlation analysis are performed in combination with the original knowledge data, feature correlation between the unit images is calculated, feature effectiveness is generated through the correlation strength, the unit images are classified according to the feature effectiveness, and a main image and a plurality of sub-images are obtained;

[0060] S110, through the feature effectiveness index, the image features corresponding to the main image are respectively fused with the image features corresponding to the plurality of sub-images, the fusion process introduces an attention mechanism, the fusion is performed by taking the feature effectiveness as an attention weight, and a fused feature set is generated;

[0061] S112, the main image, the sub-images and the fused feature set are sent to a cloud platform for remote identification through a client terminal, and an identification result is sent to the client terminal.

[0062] According to the embodiment of the present application, the to-be-tested image data of the invasive alien organism is obtained through the client terminal, and the to-be-tested image data is preprocessed, specifically:

[0063] The to-be-tested image data of the invasive alien organism is obtained through the client terminal, and the to-be-tested image data is preprocessed, specifically:

[0064] The to-be-tested image data is preprocessed by denoising, enhancing and standardizing.

[0065] According to the embodiment of the present application, the foreign species big data is screened out from a historical image database, multi-dimensional biological image feature extraction is performed on the foreign species big data, an entity data is set based on each biological image, the feature extraction process includes part features of the species, the correlation between the features is analyzed and stored, relationship data and attribute data are formed, and a biological feature knowledge graph is constructed, specifically:

[0066] Retrieving alien species image data from the historical image database based on the system database, and forming alien species big data;

[0067] For the alien species big data, an entity data is set based on each biological image record, and the entity data includes species identification information and biological name;

[0068] Based on a biological image record, a biological image set is obtained, feature extraction is performed on the biological image set by PCA analysis method to form main feature data, edge detection and contour feature analysis are performed based on Canny operator to extract texture features and contour features;

[0069] The main feature data, texture features and contour features are used as multi-dimensional features of each entity data;

[0070] According to the biological image record, the correlation information between each part of the biological body in the biological image set is obtained, and the correlation information is associated with the multi-dimensional features to obtain internal correlation information, and the recognition relationship and biological classification relationship between the biological image records are stored to form external correlation information;

[0071] The relationship data of the entity data is generated based on the internal correlation information and the external correlation information;

[0072] The entity data is used as a node, the multi-dimensional features are used as node data, the relationship data is used as node edge information, and the biological identification information of each entity data is used as attribute data to construct a biological feature knowledge graph.

[0073] It should be noted that the alien species big data is identified species data or preset image data, which is used as an existing database to construct a knowledge graph. The main feature data can be used for rapid identification, and the texture features and contour features can be used for precise identification. Each biological image sets an entity data corresponding to each biological image record. The correlation information between each part of the biological body in the biological image set includes the biological body information corresponding to different images in a biological image set and the relationship information between different images, for example, in fishery invasive organisms, including head, trunk and tail parts, and each part has positional correlation and correlation in the identification process. The recognition relationship and biological classification relationship are the record relationship between different records and the correlation relationship between biological species, such as the same fish, insect, identification time, identification terminal number and other related relationships. In addition to the biological identification information as attribute data, based on research needs, attribute data can also include user's biological information.

[0074] The biological feature knowledge graph stores knowledge data of various invasion biological records, including correlation data (internal feature correlation and external feature correlation), and through the knowledge graph, feature correlation mining of real-time biological features can be effectively performed, corresponding feature knowledge information is analyzed, and feature effectiveness is calculated through the graph subsequently.

[0075] According to the embodiment of the present application, the biological region positioning and biological view angle analysis of the to-be-tested image data are performed, a plurality of unit images are divided based on the view angle, and multi-dimensional feature extraction is performed on the plurality of unit images based on the PCA analysis method and the Canny operator to form an image feature set, specifically as follows:

[0076] The biological region positioning of the to-be-tested image data is performed to obtain a biological image region, the view angle of the biological image region is judged, and a plurality of unit images are divided based on different view angles;

[0077] Multi-dimensional feature extraction is performed on the plurality of unit images based on the PCA analysis method and the Canny operator, and multi-dimensional features of different view angles are integrated to obtain an image feature set.

[0078] It should be noted that the image feature set includes multiple data, and each data corresponds to a unit image and corresponding multi-dimensional features. The multi-dimensional feature extraction process of the plurality of unit images is consistent with the above biological image record analysis process.

[0079] According to the embodiment of the present application, the image feature set is imported into the biological feature knowledge graph as original knowledge data, the unit image is taken as a research object, the feature knowledge analysis and correlation analysis are performed in combination with the original knowledge data, the feature correlation between the unit images is calculated, the feature effectiveness is generated through the correlation strength, the unit images are classified according to the feature effectiveness, and a main image and a plurality of sub-images are obtained, specifically as follows:

[0080] The image feature set is imported into the biological feature knowledge graph as original knowledge data, and N entity node data are generated based on the features of the unit images corresponding to different view angles in the image feature set;

[0081] Taking the entity node data as an analysis node, similarity evaluation is performed on the analysis node and the existing nodes in the biological feature knowledge graph, and the existing node most similar to the analysis node is marked to obtain N first nodes;

[0082] The N first nodes are mapped and associated with the corresponding N entity node data, the N first nodes are taken as analysis objects, one first node is selected to calculate its relevance, the average node distance of the one first node and the remaining first nodes is calculated based on edge distance in the knowledge graph, the feature relevance intensity is analyzed through the average node distance, and the feature validity index is calculated based on the average node distance, and the feature validity is inversely proportional to the average node distance.

[0083] Through the calculation result, the corresponding feature validity of each unit image is set, the feature validity is sorted in size based on the feature validity, and the unit images are classified to obtain one main image and multiple sub-images.

[0084] It should be noted that the features of the unit images corresponding to the multiple view angles are the corresponding multi-dimensional features. The similarity between the feature data is analyzed, that is, the similarity analysis between the multi-dimensional feature data, which can calculate the distance between the features based on the Euclidean distance algorithm to evaluate the similarity. The N first nodes correspond to N unit images. The main image corresponds to the highest feature validity. The feature validity is inversely proportional to the average node distance. The N first nodes and the N entity node data are in one-to-one mapping relationship.

[0085] The feature validity calculation formula is as follows:

[0086]

[0087] Wherein, P y is the feature validity, K is a preset correction coefficient, and D is the average node distance.

[0088] The smaller the average node distance, the stronger the relevance, the greater the correlation between the unit images, the greater the feature validity, and the higher the prediction recognition rate.

[0089] It is worth mentioning here that the traditional identification platform often directly identifies based on the original collected images, which is low in efficiency, and has great pressure on cloud platform processing data, and the identification efficiency and effect are not ideal, and the feature information and associated information of the original image are not fully mined, and the process of analyzing biological image features and corresponding relevance from different view angles is lacking, which makes it difficult to improve the efficiency of the identification platform.

[0090] Based on this, the application constructs a corresponding knowledge graph based on the existing biological big data, performs multi-angle feature analysis on the real-time collected to-be-measured image, performs corresponding feature correlation analysis in combination with the knowledge graph, calculates the feature correlation degree of the images in different view angles based on the correlation analysis, analyzes the feature effectiveness based on this, performs image classification and feature fusion through the feature effectiveness, forms fused feature data with recognition significance, and classifies the corresponding unit images, and the process is completed based on the customer terminal, the cloud platform can perform efficient recognition through the acquisition and transmission of data, reduces the processing pressure of the cloud platform, and improves the efficiency of the information identification platform.

[0091] According to the embodiment of the application, the feature of the main image is fused with the feature of the plurality of sub-images through the feature effectiveness index, the attention mechanism is introduced in the fusion process, the feature effectiveness is taken as the attention weight for fusion, and a fused feature set is generated, specifically:

[0092] The feature effectiveness index is used to assign attention weights to the main image and the sub-images;

[0093] The attention mechanism is introduced to convert the multi-dimensional feature corresponding to the main image and the multi-dimensional feature corresponding to one sub-image into a first feature map and a second feature map, the first feature map and the second feature map are weighted and fused through the corresponding attention weights, and a fused feature is generated;

[0094] The feature of the main image is fused with the feature of the plurality of sub-images, and the fused feature set is obtained by integrating the fused features.

[0095] It should be noted that the fused feature set is transmitted to the cloud platform, which can greatly improve the recognition efficiency, and the process is completed based on the client, thereby reducing the processing pressure of the cloud platform and improving the recognition efficiency of the cloud platform. The fused feature includes effective information and potential feature information of each part of the feature.

[0096] According to the embodiment of the application, the main image, the sub-image and the fused feature set are sent to the cloud platform for remote identification through the customer terminal, and the identification result is sent to the customer terminal, specifically:

[0097] In the cloud platform, the main image, the sub-image and the fused feature set sent by the customer terminal are collected;

[0098] Based on the recognition model in the cloud platform, the main image and the sub-image are identified once, the fused feature set is identified twice, the identification result is comprehensively evaluated, and an identification report is generated;

[0099] The identification report is sent to the customer terminal in real time.

[0100] It should be noted that the cloud platform deploys a corresponding identification module, and the identification module can perform biological identification analysis based on a CNN model.

[0101] According to an embodiment of the present application, the cloud platform adopts a distributed architecture, deploys services in multiple server nodes, and is connected with the client terminal through an Ethernet or a wireless network.

[0102] According to an embodiment of the present application, the method further comprises:

[0103] In a preset time period, the number P1 of client terminals connected and the amount P2 of real-time transmission data are counted.

[0104] If both P1 and P2 are greater than a preset data range, a preset compression scheme is obtained.

[0105] The client terminal obtains unit images to be transmitted, which are marked as to-be-transmitted images.

[0106] The compression rate is set based on the feature effectiveness of the to-be-transmitted images, and the compression rate is in a positive correlation with the feature effectiveness.

[0107] According to the compression rate range setting and the compression rate, the multiple to-be-transmitted images are grouped to form multiple groups of transmission data.

[0108] The multiple groups of transmission data are set with corresponding compression algorithms, the preset compression scheme is dynamically adjusted, and a real-time compression scheme based on the transmission data is formed.

[0109] It should be noted that the compression rate range setting includes multiple compression rate ranges, different ranges correspond to different network compression algorithms, and multiple groups of data are divided based on different ranges. The real-time compression scheme includes the setting of multiple compression algorithms, and each group of transmission data corresponds to one compression algorithm.

[0110] It is worth mentioning here that in the process of simultaneously identifying alien species by multiple client terminals, when the data amount is greater than a preset range, the identification efficiency of the server is often reduced. Based on this, an embodiment of the present application performs feature effectiveness analysis on the classified unit images, sets a corresponding compression rate, groups the image data, and dynamically adjusts the compression scheme, thereby effectively reducing the data amount while ensuring the feature information and the identification rate, and achieving the purpose of dynamically adjusting the platform processing and identification efficiency.

[0111] Figure 2 A block diagram of a remote identification and identification platform based on alien invasive organisms is shown.

[0112] The second aspect of the application also provides a remote identification and identification platform based on alien invasive organisms, which comprises a memory 21 and a processor 22, the memory 21 comprises a remote identification and identification program based on alien invasive organisms, and the remote identification and identification program based on alien invasive organisms is executed by the processor 22 to realize the following steps:

[0113] Through the client terminal, obtain the to-be-tested image data of the alien invasive organisms, and pre-process the to-be-tested image data;

[0114] Filter out alien species big data from the historical image database, perform multi-dimensional biological image feature extraction on the alien species big data, and set an entity data based on each biological image, the feature extraction process includes the characteristics of each part of the species, analyzes and stores the correlation between the characteristics, forms relationship data and attribute data, and constructs a biological feature knowledge graph;

[0115] Perform biological region positioning and biological view angle analysis on the to-be-tested image data, divide multiple unit images based on the view angle, perform multi-dimensional feature extraction on the multiple unit images based on the PCA analysis method and the Canny operator, and form an image feature set;

[0116] Import the image feature set as original knowledge data into the biological feature knowledge graph, take the unit image as the research object, perform feature knowledge analysis and correlation analysis on the original knowledge data, calculate the feature correlation between the unit images, generate feature effectiveness through the correlation strength, classify the unit images according to the feature effectiveness, and obtain a main image and multiple sub-images;

[0117] Through the feature effectiveness index, perform feature fusion on the image features corresponding to the main image and the image features corresponding to the multiple sub-images respectively, introduce an attention mechanism in the fusion process, take the feature effectiveness as the attention weight for fusion, and generate a fusion feature set;

[0118] Through the client terminal, send the main image, the sub-image and the fusion feature set to the cloud platform for remote identification, and send the identification result to the client terminal.

[0119] According to the embodiment of the application, the to-be-tested image data of the alien invasive organisms is obtained through the client terminal, and the to-be-tested image data is pre-processed, specifically:

[0120] Through the client terminal, obtain the user real-time upload data, perform data analysis on the uploaded data, extract the image data, and obtain the to-be-tested image data of the alien invasive organisms;

[0121] The to-be-tested image data is pre-processed by noise reduction, enhancement and standardization.

[0122] According to the embodiment of the present application, the alien species big data is screened out from the historical image database, multi-dimensional biological image feature extraction is performed on the alien species big data, and an entity data is set based on each biological image. In the feature extraction process, the features of each part of the species are included, the correlation between the features is analyzed and stored, the relationship data and attribute data are formed, and the biological feature knowledge graph is constructed. Specifically, the following steps are included:

[0123] Based on the system database, the alien species image data is retrieved from the historical image database, and the alien species big data is formed;

[0124] For the alien species big data, an entity data is set based on each biological image record, and the entity data includes species identification information and biological name;

[0125] Based on a biological image record, a biological image set is obtained, PCA analysis is performed on the biological image set to extract features and form main feature data, edge detection and contour feature analysis are performed based on Canny operator to extract texture features and contour features;

[0126] The main feature data, texture features and contour features are used as multi-dimensional features of each entity data;

[0127] According to the biological image record, the correlation information between each part of the biological body in the biological image set is obtained, and the correlation information is associated with the multi-dimensional features to obtain internal correlation information. The recognition relationship and biological classification relationship between the biological image records are stored to form external correlation information;

[0128] The relationship data of the entity data is generated based on the internal correlation information and the external correlation information;

[0129] The entity data is used as a node, the multi-dimensional features are used as node data, the relationship data is used as node edge information, and the biological identification information of each entity data is used as attribute data to construct a biological feature knowledge graph.

[0130] It should be noted that the alien species big data is identified species data or preset image data, which is used as an existing database to build a knowledge graph. The main feature data can be used for rapid identification, and the texture feature and the contour feature can be combined to realize accurate identification. Each biological image is set as an entity data, that is, corresponding to each biological image record. The correlation information between the parts of the biological body includes the biological body information corresponding to different images in a biological image set and the relationship information between different images. For example, in the fishery invasive organisms, the parts include the head, the trunk and the tail, and each part has positional correlation and correlation in the identification process. The identification relationship and the biological classification relationship are the record relationship between different records and the correlation relationship of biological species, such as the same kind of fish, insect, identification time, identification terminal number and other correlation relationship. In addition to the biological identification information as attribute data, based on research needs, the attribute data can also include biological information noted by the user.

[0131] The biological feature knowledge graph stores knowledge data of various invasive biological records, including correlation data (internal feature correlation and external feature correlation). Through the knowledge graph, the feature correlation of real-time biological features can be effectively mined, the corresponding feature knowledge information can be analyzed, and the feature effectiveness can be calculated through the graph subsequently.

[0132] According to the embodiment of the present application, the biological region positioning and biological view angle analysis of the to-be-tested image data are performed, a plurality of unit images are divided based on the view angle, and multi-dimensional feature extraction is performed on the plurality of unit images based on the PCA analysis method and the Canny operator to form an image feature set. Specifically,

[0133] The biological region positioning of the to-be-tested image data is performed to obtain a biological image region, the view angle of the biological image region is judged, and a plurality of unit images are divided based on different view angles.

[0134] Multi-dimensional feature extraction is performed on the plurality of unit images based on the PCA analysis method and the Canny operator, and multi-dimensional features of different view angles are integrated to obtain an image feature set.

[0135] It should be noted that the image feature set includes multiple data, and each data corresponds to a unit image and corresponding multi-dimensional features. The multi-dimensional feature extraction process of the plurality of unit images is consistent with the above biological image record analysis process.

[0136] According to the embodiment of the present application, the image feature set is introduced into the biometric feature knowledge graph as original knowledge data, a unit image is taken as a research object, feature knowledge analysis and correlation analysis are carried out in combination with the original knowledge data, feature correlation between unit images is calculated, and feature effectiveness is generated through correlation intensity; the unit images are classified according to the feature effectiveness, and a main image and multiple sub-images are obtained, specifically as follows:

[0137] The image feature set is introduced into the biometric feature knowledge graph as original knowledge data, and N entity node data is generated based on the features of the unit images corresponding to different view angles in the image feature set;

[0138] Taking the entity node data as an analysis node, similarity of the analysis node and the existing node in the biometric feature knowledge graph is evaluated, and the existing node most similar to the analysis node is marked to obtain N first nodes;

[0139] The N first nodes are mapped and correlated with the corresponding N entity node data, one first node is selected as an analysis object to calculate its correlation, the average node distance between the one first node and the remaining first nodes in the knowledge graph is calculated based on the edge distance, the feature correlation intensity is analyzed through the average node distance, and the feature effectiveness index is calculated based on the average node distance; the feature effectiveness is inversely proportional to the average node distance;

[0140] Through the calculation result, the corresponding feature effectiveness of each unit image is set, the unit images are sorted in size based on the feature effectiveness, and the unit images are classified to obtain a main image and multiple sub-images.

[0141] It should be noted that the features of the unit images corresponding to the multiple view angles are corresponding multi-dimensional features. The similarity between the feature data is analyzed, that is, the similarity analysis between multi-dimensional feature data, which can calculate the distance between features based on algorithms such as Euclidean distance to evaluate the similarity. The N first nodes correspond to N unit images. The main image corresponds to the highest feature effectiveness. The feature effectiveness is inversely proportional to the average node distance. The N first nodes and the N entity node data are in one-to-one mapping relationship.

[0142] The feature effectiveness calculation formula is as follows:

[0143]

[0144] Wherein, P y is the feature effectiveness, K is a preset correction coefficient, and D is the average node distance.

[0145] The smaller the average node distance, the stronger the correlation, the greater the correlation between the unit images, and the greater the feature effectiveness, and at the same time, the higher the prediction recognition rate.

[0146] It is worth mentioning here that the traditional identification platform is often based on the original collected image for direct identification, which is low in efficiency, and the cloud platform is under great pressure to process data, and the identification efficiency and effect are not ideal, and the characteristic information and associated information of the original image are not fully mined, and the process of analyzing the biological image features and the corresponding correlation from different view angles is lacking, so that the identification platform efficiency is difficult to improve.

[0147] Based on this, the present application constructs a corresponding knowledge graph based on the existing biological big data, analyzes the multi-angle features through the real-time collected images to be measured, combines the knowledge graph to analyze the correlation of the corresponding features, calculates the feature correlation degree of the images from different view angles based on the correlation analysis, analyzes the feature effectiveness based on this, classifies the images and fuses the features through the feature effectiveness, forms the fused feature data with the identification significance, and classifies the corresponding unit images, and the process is completed based on the customer terminal, and the cloud platform can perform efficient identification by acquiring and transmitting data, thereby reducing the processing pressure of the cloud platform and improving the efficiency of the information identification platform.

[0148] According to the embodiment of the present application, the feature fusion is performed between the image features corresponding to the main image and the image features corresponding to the plurality of sub-images through the feature effectiveness index, the attention mechanism is introduced in the fusion process, the feature fusion is performed by taking the feature effectiveness as the attention weight, and the fused feature set is generated, specifically:

[0149] Based on the feature effectiveness index, the attention weight is given to the main image and the sub-image;

[0150] The attention mechanism is introduced, the multi-dimensional features corresponding to the main image and the multi-dimensional features corresponding to one sub-image are converted into feature maps to form a first feature map and a second feature map, the first feature map and the second feature map are weighted and fused through the corresponding attention weight, and the fused feature is generated;

[0151] The feature fusion is performed between the image features corresponding to the main image and the image features corresponding to the plurality of sub-images, and the fused feature set is obtained by integrating the fused features.

[0152] It should be noted that the fused feature set is transmitted to the cloud platform, which can greatly improve the identification efficiency, and the process is completed based on the client, thereby reducing the processing pressure of the cloud platform and improving the identification efficiency of the cloud platform. The fused features include the effective information and potential feature information of each part of the features.

[0153] According to the embodiment of the present application, the main image, the sub-image and the fused feature set are sent to the cloud platform for remote identification through the customer terminal, and the identification result is sent to the customer terminal, specifically:

[0154] In the cloud platform, the main image, the secondary image and the fusion feature set sent by the client terminal are collected;

[0155] Based on the recognition model in the cloud platform, the main image and the secondary image are subjected to primary recognition, the fusion feature set is subjected to secondary recognition, the recognition results are comprehensively evaluated, and an identification report is generated;

[0156] The identification report is sent to the client terminal in real time.

[0157] It should be noted that the cloud platform deploys a corresponding recognition module, and the recognition module can perform biological recognition analysis based on a CNN model.

[0158] According to an embodiment of the present application, the cloud platform adopts a distributed architecture, deploys services in multiple server nodes, and is connected with the client terminal through an Ethernet or a wireless network.

[0159] The third aspect of the present application also provides a computer readable storage medium, which comprises a remote identification and identification program based on alien invasive organisms, and the remote identification and identification program based on alien invasive organisms, when executed by a processor, realizes the steps of the remote identification and identification method based on alien invasive organisms according to any one of the above.

[0160] The present application discloses a remote identification and identification platform based on alien invasive organisms. The client terminal obtains the image data to be measured and pre-processes it. The system selects alien species big data from the historical image library, extracts multi-dimensional biological image features, and constructs a biological feature knowledge graph. The biological region positioning and different view angle image division are performed on the image to be measured, the features are extracted using PCA and Canny operator, and the image feature set is formed. The feature set is imported into the knowledge graph, the correlation is analyzed, and the main image, the secondary image and the feature effective degree of different images are obtained. Based on the feature effective degree, the attention mechanism is introduced to fuse the main and secondary image features, and the fusion feature set is generated. Finally, the data such as the fusion feature set is uploaded to the cloud platform through the client terminal for remote identification, and the result is fed back to the client terminal. Through the present application, the identification efficiency and effect of the identification platform are effectively improved, and the purpose of optimizing the identification process is achieved.

[0161] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other manners. The described device embodiments are merely illustrative. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling, or direct coupling or communication connection between the components can be indirect coupling or communication connection through some interfaces, devices, or units, and can be electrical, mechanical, or in other forms.

[0162] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units; they can be located in one place, or distributed on multiple network units; and some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0163] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional units.

[0164] Those of ordinary skill in the art can understand that all or part of the steps of the above-described method embodiments can be completed by a program instructing related hardware, and the foregoing program can be stored in a computer readable storage medium, and when the program is executed, the steps of the method embodiments are executed; and the foregoing storage medium includes: mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks or optical disks, and various media that can store program codes.

[0165] Alternatively, the integrated units of the present application, if implemented in the form of software functional modules and sold or used as independent products, can also be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the embodiments of the present application. The foregoing storage medium includes: mobile storage devices, ROMs, RAMs, magnetic disks or optical disks, and various media that can store program codes.

[0166] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A remote identification and identification method for invasive alien organisms, characterized in that: include: Obtaining image data of the invasive alien organisms to be detected through the client terminal and pre-processing the image data to be detected; Screening alien species big data from historical image databases, performing multi-dimensional biological image feature extraction on the alien species big data, and setting an entity data based on each biological image. The feature extraction process includes the characteristics of each part of the species, analyzing and storing the correlation between the features, forming relationship data and attribute data, and constructing a biological feature knowledge map; The image data to be tested is subjected to biological region positioning and biological view angle analysis, and multiple unit images are divided based on the view angle. Based on the PCA analysis method and Canny operator, multi-dimensional features of the multiple unit images are extracted to form an image feature set; Import the image feature set as raw knowledge data into the biometric knowledge graph, take the unit image as the research object, combine the raw knowledge data to perform feature knowledge analysis and correlation analysis, calculate the feature correlation between unit images, and generate feature validity based on the correlation strength. Classify the unit images according to the feature validity to obtain a main image and multiple secondary images; By using the feature validity index, the image features corresponding to the main image are fused with the image features corresponding to multiple secondary images. The attention mechanism is introduced into the fusion process, and the feature validity is used as the attention weight for fusion to generate a fused feature set. The main image, the secondary image and the fusion feature set are sent to the cloud platform through the client terminal for remote identification, and the identification result is sent to the client terminal.

2. The remote identification and identification method for invasive alien organisms according to claim 1, characterized in that: The method of obtaining the image data of the invasive alien organisms to be detected through the client terminal and preprocessing the image data to be detected is as follows: Through the client terminal, the real-time uploaded data of the user is obtained, the uploaded data is analyzed, the image data is extracted, and the image data of the invasive alien organisms to be tested is obtained; The image data to be tested is subjected to noise reduction, enhancement and standardization preprocessing.

3. The remote identification and identification method based on alien invasive organisms according to claim 1 is characterized in that: The method involves filtering out alien species big data from the historical image database, performing multi-dimensional biological image feature extraction on the alien species big data, and setting an entity data based on each biological image. The feature extraction process includes the characteristics of each part of the species, analyzing and storing the correlation between the features, forming relationship data and attribute data, and constructing a biological feature knowledge graph, specifically: Based on the system database, image data of alien species are retrieved from the historical image database to form big data of alien species; For alien species big data, an entity data is set based on each biological image record. The entity data includes species identification information and biological name. Based on a biological image record, a biological image set is obtained, and the features of the biological image set are extracted by PCA analysis to form main feature data. The edge detection and contour feature analysis are performed based on the Canny operator to extract texture features and contour features. The main feature data, texture features and contour features are used as multi-dimensional features of each entity data; According to the biological image records, the correlation information between the parts of the biological body in the biological image set is obtained, and the correlation information is mapped with the multi-dimensional features to obtain the internal correlation information. The information is stored based on the identification relationship between the biological image records and the biological classification relationship to form the external correlation information; Generate relational data of entity data based on internal and external relational information; A biometric knowledge graph is constructed using entity data as nodes, multi-dimensional features as node data, relationship data as node edge information, and biometric information of each entity data as attribute data.

4. The remote identification and identification method based on alien invasive organisms according to claim 3 is characterized in that: The image data to be tested is subjected to biological region positioning and biological view angle analysis, and multiple unit images are divided based on the view angle. Based on the PCA analysis method and the Canny operator, multi-dimensional features are extracted from the multiple unit images to form an image feature set, specifically: Positioning the image data to be tested in a biological region to obtain a biological image region, determining a viewing angle of the biological image region, and dividing the image into a plurality of unit images based on different viewing angles; Based on the PCA analysis method and the Canny operator, multi-dimensional features of multiple unit images are extracted, and the multi-dimensional features of different viewing angles are integrated to obtain the image feature set.

5. The remote identification and identification method based on alien invasive organisms according to claim 4 is characterized in that: The image feature set is imported into the biometric knowledge graph as the original knowledge data, and the unit image is used as the research object. The feature knowledge analysis and correlation analysis are performed in combination with the original knowledge data. The feature correlation between the unit images is calculated, and the feature validity is generated by the correlation strength. The unit images are classified according to the feature validity to obtain a main image and multiple sub-images. Specifically, Import the image feature set as raw knowledge data into the biometric knowledge graph, and generate N entity node data based on the features of the unit image corresponding to different viewing angles in the image feature set; Taking the entity node data as the analysis node, the similarity between the analysis node and the existing nodes in the biometric knowledge graph is evaluated, and the existing nodes that are most similar to the analysis node are marked to obtain N first nodes; Map and associate N first nodes with corresponding N entity node data, take the N first nodes as analysis objects, select one first node and calculate its relevance, calculate the average node distance between the first node and the remaining first nodes in the knowledge graph based on the edge distance, analyze the feature relevance strength through the average node distance, and calculate the feature validity index based on the average node distance. The feature validity is inversely proportional to the average node distance. Based on the calculation results, a corresponding feature validity is set for each unit image, the size is sorted based on the feature validity, and the unit images are classified to obtain a main image and multiple sub-images.

6. The remote identification and identification method for invasive alien organisms according to claim 1, characterized in that: The feature validity index is used to fuse the image features corresponding to the main image with the image features corresponding to multiple secondary images. The attention mechanism is introduced into the fusion process, and the feature validity is used as the attention weight for fusion, and a fusion feature set is generated, specifically: Based on the feature validity index, attention weights are assigned to the main image and the secondary image; The attention mechanism is introduced to convert the multi-dimensional features corresponding to the main image and the multi-dimensional features corresponding to a secondary image into feature maps to form a first feature map and a second feature map. The first feature map and the second feature map are weightedly fused using the corresponding attention weights to generate a fused feature. The image features corresponding to the main image are fused with the image features corresponding to multiple secondary images respectively, and the fused features are integrated to obtain a fused feature set.

7. The remote identification and identification method for invasive alien organisms according to claim 1, characterized in that: The main image, the secondary image and the fusion feature set are sent to the cloud platform for remote identification through the client terminal, and the identification result is sent to the client terminal, specifically: In the cloud platform, the main image, secondary image and fusion feature set sent by the client terminal are collected; Based on the recognition model in the cloud platform, the primary recognition is performed based on the main image and the secondary image, and the secondary recognition is performed based on the fusion feature set. The recognition results are comprehensively evaluated and an identification report is generated; The appraisal report is sent to the customer terminal in real time.

8. The remote identification and identification method for invasive alien organisms according to claim 1, characterized in that: The cloud platform adopts a distributed architecture and deploys services in multiple server nodes. The cloud platform and client terminals are connected via Ethernet or wireless network.

9. A remote identification and authentication platform for invasive alien organisms, characterized by: The platform includes: a memory and a processor. The memory includes a remote identification and authentication program based on alien invasive organisms. When the remote identification and authentication program based on alien invasive organisms is executed by the processor, the following steps are implemented: Obtaining image data of the invasive alien organisms to be detected through the client terminal and pre-processing the image data to be detected; Screening alien species big data from historical image databases, performing multi-dimensional biological image feature extraction on the alien species big data, and setting an entity data based on each biological image. The feature extraction process includes the characteristics of each part of the species, analyzing and storing the correlation between the features, forming relationship data and attribute data, and constructing a biological feature knowledge map; The image data to be tested is subjected to biological region positioning and biological view angle analysis, and multiple unit images are divided based on the view angle. Based on the PCA analysis method and Canny operator, multi-dimensional features of the multiple unit images are extracted to form an image feature set; Import the image feature set as raw knowledge data into the biometric knowledge graph, take the unit image as the research object, combine the raw knowledge data to perform feature knowledge analysis and correlation analysis, calculate the feature correlation between unit images, and generate feature validity based on the correlation strength. Classify the unit images according to the feature validity to obtain a main image and multiple secondary images; By using the feature validity index, the image features corresponding to the main image are fused with the image features corresponding to multiple secondary images. The attention mechanism is introduced into the fusion process, and the feature validity is used as the attention weight for fusion to generate a fused feature set. The main image, the secondary image and the fusion feature set are sent to the cloud platform through the client terminal for remote identification, and the identification result is sent to the client terminal.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a remote identification and authentication program based on alien invasive organisms. When the remote identification and authentication program based on alien invasive organisms is executed by a processor, the steps of the remote identification and authentication method based on alien invasive organisms as described in any one of claims 1 to 8 are implemented.

Images