KR102614062B1 - Method and apparatus for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification - Google Patents

Abstract

The present invention relates to a method and device for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification.
The device for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification according to the present invention receives data from the user terminal and the user terminal possessed by the quarantine officer, integrates ornamental fish information in different formats, and is based on a deep neural network. It determines fish species and disease and includes a processing server that automatically corrects errors contained in text received from the user terminal.

Description

Method and apparatus for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification

The present invention relates to a method and device for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification.

Recently, as we have entered the era of artificial intelligence (AI), artificial intelligence technology that can solve cognitive problems linked to human intelligence, such as learning ability, reasoning ability, and perception ability, has been applied to various application fields such as manufacturing, medical care, and financial services. It is becoming.

In particular, artificial intelligence image recognition technology recognizes objects through detection and classification processing methods from photos or videos, and innovative recognition techniques such as deep learning continue to emerge.

Using this, commercialized technologies such as Naver's Smart Lens and POSCO ICT's Pace emerged domestically, Tesla's Full Self Driving (FSD) overseas, and image-based autonomous operation in Korea, Europe, Japan, Finland, etc. As research on ships (MASS: Marine Autonomous Surface Ships) progresses, artificial intelligence image recognition technology is becoming a part of the lives of people around the world, while also attracting attention in various fields where it has not been applied.

Currently, ornamental fish imported from overseas are brought in by air or ship, and are packed in boxes, with each ornamental fish individually/group packed in plastic.

Ornamental fish imported in this way must necessarily go through a quarantine process. Quarantine experts carry out the quarantine manually using the ornamental fish information packing list received from the export company or the quarantine documents received from the National Fisheries Quality Control Center. However, there are many problems with this quarantine process. exist.

The form of quarantine documents for ornamental fish is different for different exporters, countries, and seafood quality control agencies. Because quarantine officers conduct quarantine using a list of all ornamental fish information, the time required for quarantine is very long. And because quarantine officers identify with the naked eye, confidence in accuracy is low.

In fact, during the visual identification process at the quarantine site, it is difficult to distinguish ornamental fish only from the list of ornamental fish information received from export companies and the Fishery Products Quality Control Center, so it is often necessary to distinguish ornamental fish through individual Internet searches.

In addition, since the number of imported ornamental fish is very large and most of them have English names, typos are frequent during the search process, which slows down the quarantine process.

According to the prior art, image processing and artificial intelligence technology are used as a method of extracting fish information.

However, it is difficult to classify fish species using a fish information extraction method using image processing technology, and it is usually used as a technology to measure the body length of fish. By taking an image photo of a fish, you can measure the body length of the fish using image preprocessing techniques such as binarization and edge detection. However, measuring fish using only 2D images requires limited conditions such as marking. Additionally, the performance of image processing techniques varies greatly depending on the intensity of light, the color of the fish, and the environment, making it difficult to extract information about ornamental fish.

Types of objects can be classified using artificial intelligence technology. An image dataset of the fish species to be classified is created, and the input fish images are analyzed and classified using learning parameters extracted by learning from CNN, a deep neural network. As such, there are difficulties in recognizing fish and extracting good performance using currently open general-purpose artificial intelligence models. For this reason, since it is a technology that analyzes images/videos, it has limitations that make it realistically difficult to distinguish similar shapes and similar colors.

The present invention was proposed to solve the above-mentioned problems, and integrates a list of ornamental fish in different formats during the quarantine process for imported ornamental fish, thereby creating an integrated ornamental fish that can be processed more quickly than the current manual quarantine process. We propose a quarantine aquarium fish app SW linked to the database, perform character error correction to quickly and accurately search for ornamental fish species with English names, and improve the low accuracy in the process of visual inspection of aquarium fish. The purpose is to provide a device and method for recognizing ornamental fish species and analyzing diseases using images based on a specialized deep neural network.

The device for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification according to the present invention receives data from the user terminal and the user terminal possessed by the quarantine officer, integrates ornamental fish information in different formats, and is based on a deep neural network. It determines fish species and disease and includes a processing server that automatically corrects errors contained in text received from the user terminal.

The processing server includes a data integration management unit, and the data integration management unit receives the ornamental fish packing list and the ornamental fish information quarantine document received during the ornamental fish import process, analyzes format similarity for the lists in different formats, and creates a unified format. Integrate data into a format.

The processing server includes an image recognition unit, and the image recognition unit recognizes fish species and diseases of the fish using pre-learned parameters for the ornamental fish photographed image received from the user terminal.

The image recognition unit uses multiple deep neural networks to recognize the fish species and multiple deep neural network learning parameters to recognize diseases of the fish.

The processing server maps pre-stored fish information with the fish species and fish disease recognition results and transmits them to the user terminal.

When the image recognition unit receives new data rather than existing data from the user terminal, it retrains the artificial intelligence model using the new image data.

The processing server includes a typo character correction unit, wherein the typo character correction unit uses pre-stored information to automatically correct typos included in the ornamental fish search text to be searched, and provides a list of fish species names with a standard similarity level or higher. do.

When the typo character correction unit receives new data rather than existing data from the user terminal, it re-trains the artificial intelligence model using the new fish species text data.

The processing server requests reading from another expert's device and provides an alarm to the user terminal upon completion of reading.

The method for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification according to the present invention includes the steps of (a) receiving data from a user terminal, and (b) using the data to obtain ornamental fish information in different formats. It includes the steps of integrating and (c) using the data to determine fish species and diseases based on a deep neural network and correcting errors included in the text.

Step (b) receives the ornamental fish packing list and the ornamental fish information quarantine document and integrates the different formats into a unified format.

In step (c), fish species and fish diseases are recognized using parameters already learned for images of ornamental fish, and retraining the artificial intelligence model is performed when new images are received.

In step (c), typos included in the ornamental fish search text are corrected, a list of fish species names higher than the standard similarity level is provided, and artificial intelligence model retraining is performed when new fish species text data is received.

According to the present invention, as a quarantine app SW, it can be used on a variety of devices rather than a designated device, which has the effect of improving the convenience of quarantine.

By integrating the different types of ornamental fish information required for quarantine and building them on a DB server, it is possible to check with quarantine app SW linked to the DB server that allows integration of information in different formats without the need to refer to multiple documents, allowing quarantine officers to It has the effect of speeding up quarantine processing.

By applying deep neural network technology, it can serve as an assistant to quarantine officers, replacing human eyes to identify species and diseases of ornamental fish to be quarantined, and automatically correct errors in input characters for the fish species to be found through text search. This has the effect of improving the convenience and accuracy of quarantine.

The advantage is that newly introduced ornamental fish can be quickly applied to the learning engine and quarantined through periodic deep learning engine updates.

If recognition is not possible with the deep neural network and the corresponding quarantine officer's naked eyes, images and other information of the ornamental fish in question are sent to various experts through the expert request service, which has the effect of increasing the accuracy of ornamental fish recognition by allowing them to conduct quarantine instead.

We propose an ornamental fish information DB server and quarantine app SW for accurate and quick quarantine by quarantine officers, and can further be applied as a service to help the public obtain ornamental fish information.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

Figure 1 shows an imported ornamental fish quarantine device through data preprocessing and deep neural network-based image detection and classification according to an embodiment of the present invention.
Figure 2 shows a processing process using a deep neural network according to an embodiment of the present invention.
Figure 3 shows the result of integrating ornamental fish data of different formats in a DB (Server) according to an embodiment of the present invention.
Figure 4 shows a deep neural network-based automatic search character error correction process and a fish species recommendation screen according to an embodiment of the present invention.
Figure 5 shows deep neural network-based image recognition (fish species recognition/disease recognition) and ornamental fish DB data mapping according to the results according to an embodiment of the present invention.
Figure 6 shows fish species recognition and disease recognition of a multi-deep neural network structure according to an embodiment of the present invention.
Figure 7 shows an ornamental fish recognition structure based on a multiple deep neural network according to a fish classification method according to an embodiment of the present invention.
Figure 8 shows an ornamental fish disease recognition structure based on a multiple deep neural network according to region division according to an embodiment of the present invention.
Figure 9 shows the SW UI of an app dedicated to ornamental fish data mapping and quarantine according to the results of a deep neural network according to an embodiment of the present invention.

The above-mentioned object and other objects, advantages and features of the present invention, and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings.

However, the present invention is not limited to the embodiments disclosed below and can be implemented in various different forms. The following embodiments are merely intended to convey to those skilled in the art the purpose of the invention, It is only provided to easily inform the configuration and effect, and the scope of rights of the present invention is defined by the description of the claims.

Meanwhile, the terms used in this specification are for describing embodiments and are not intended to limit the present invention. As used herein, singular forms also include plural forms, unless specifically stated otherwise in the context. As used in the specification, “comprises” and/or “comprising” means that the mentioned element, step, operation and/or element precludes the presence of one or more other elements, steps, operations and/or elements. Or it is not excluded that it is added.

According to an embodiment of the present invention, in the method of quarantining imported ornamental fish brought into the country, a deep neural network CNN (Convolution Neural Network) model is used to enable rapid quarantine by eliminating unnecessary quarantine processes, which is based on a fish classification method. It recognizes fish species and diseases using multiple deep neural networks, and automatically corrects error letters written incorrectly for the fish species you are searching for.

According to an embodiment of the present invention, we propose a dedicated ornamental fish quarantine app SW that can be used on various types of devices to perform the above-described functions.

According to an embodiment of the present invention, a dedicated quarantine app SW is provided, automatic integration of ornamental fish information in different formats is proposed, and deep neural network-based image and text technology is proposed to enable quarantine officers to be more accurate about imported ornamental fish. We provide support for rapid quarantine processing.

According to an embodiment of the present invention, ornamental fish data in different formats are automatically integrated in the DB (Server) so that quarantine officers can easily check the information, and quarantine can be quickly implemented through a quarantine app SW that can be processed on a general-purpose device. It has technical features that enable accurate and fast search through deep neural network-based automatic error character correction, and fish species recognition and disease identification based on multiple deep neural networks specialized for ornamental fish.

Figure 1 shows an imported ornamental fish quarantine device through data preprocessing and deep neural network-based image detection and classification according to an embodiment of the present invention.

Quarantine app SW can be installed on all terminals with a built-in camera and WIFI communication module.

The user terminal 100 includes a transmitter 110 that connects to the quarantine app SW and transmits ornamental fish packet data and captured images to the processing server, and a receiver that receives a list of integrated ornamental fish data and processing results from the processing server 200 ( 120).

The ornamental fish packet data and captured images transmitted by the transmitting unit 110 of the user terminal 100 are received by the communication unit 210 of the processing server 200, and a plurality of functions are performed through the processing unit 220.

The processing server 200 includes a data integration management unit 230, an image recognition unit 240, and a typo character correction unit 250. The image recognition unit 240 and a typo character correction unit 250 are based on artificial intelligence. performs its function.

They process their respective functions according to data transmitted by the quarantine app SW of the user terminal 100.

The data integration management unit 230 uses the quarantine app SW accessed on the user terminal 100 to collect ornamental fish packing lists in different formats received from importers in the process of importing ornamental fish for quarantine or ornamental fish information received from the National Fisheries Quality Management Institute. Quarantine documents are received, format similarity is analyzed for different lists, and the data is integrated into a single format.

The image recognition unit 240 receives an image of an ornamental fish to be quarantined through the quarantine app SW accessed from the user terminal 100 through the communication unit 210, and uses a deep neural network in advance through artificial intelligence image recognition of the processing unit 220. Recognize fish species and fish diseases using learning parameters (Weight/Bias) trained with data from the ornamental fish DB. Fish species and disease recognition using deep neural networks is structured with multiple deep neural networks specialized for fish, and will be described later with reference to FIGS. 8 and 9.

When fish species and fish diseases are recognized through a multi-deep neural network, the recognition results are used to perform mapping with the fish information stored in the data integration management unit 230, and various information about the recognized fish required for quarantine is obtained through the quarantine app SW. provides.

The typo correction unit 250 provides a search function through the quarantine app SW accessed from the user terminal 100, and uses a deep neural network to search for ornamental words to be found using information stored in the data integration management unit 230. Performs automatic correction of typos included in .

Most ornamental fish have English names rather than Korean names, and because they indicate short and long fish depending on the species, typos often occur, which causes the problem of slow quarantine processing.

The typo correction unit 250 not only automatically corrects typos in the search text using a primary deep neural network, but also quarantines several fish species most similar to the input text so that the user can secondary check the list of similar fish species through the quarantine app SW. Provided through.

Figure 2 shows a processing process using a deep neural network according to an embodiment of the present invention.

The user terminal 100 includes a camera and a WIFI communication module, and runs the ornamental fish quarantine SW app.

The data integration management unit 230 receives quarantine packet data in different formats, analyzes similar formats of the data, integrates them, and connects to the server from the ornamental fish quarantine SW app to allow the user to check.

When the data integration management unit 230 receives new data rather than existing data, the image recognition unit 240 and the typo character correction unit 250 perform re-processing in each deep neural network based on comprehensive data including the new data. Learning progresses.

The image recognition unit 240 retrains the artificial intelligence model using new image data, and the typo correction unit 250 retrains the artificial intelligence model using new fish type text data.

When new data is not received from the user terminal 100, the image recognition unit 240 and the typo character correction unit 250 use previously learned parameters (Weight/Bias).

When an ornamental fish species is entered as searched by the data integration management unit 230 through the ornamental fish quarantine SW app used by the user terminal 100, the input text is transmitted to the typo character correction unit 250 (i.e., real-time input text data is transmitted).

The typo character correction unit 250 determines whether an error has occurred in the received real-time input text using parameters previously learned from an artificial intelligence model, and performs correction when an error occurs. At this time, through recognition in a deep neural network, a list of the most likely preset number of fish species (e.g. 5) is transmitted to the ornamental fish quarantine SW app so that the user can check it.

The user terminal 100 transmits images taken of ornamental fish required for quarantine in the field to the image recognition unit 240 through the ornamental fish quarantine SW app.

The image recognition unit 240 recognizes the fish species and disease in the received image using parameters previously extracted through learning, and the fish species and disease are processed by each multiple deep neural network (fish species deep neural network, disease deep neural network). do.

The recognition result information is transmitted to the aquarium fish quarantine SW app, and the user checks the image recognition result and goes through a secondary confirmation process. If the result is correct, quarantine is carried out. Otherwise, if it is not possible to determine with the quarantine officer's naked eye, an expert interpretation is performed. Proceed.

Expert reading sends the image and information of the unreadable object to another expert to request reading, and when the reading is completed by another expert, the reading result is sent to the on-site quarantine officer. When the reading result is sent, it is sent to the on-site text service. Notify quarantine officers.

Figure 3 shows the result of integrating ornamental fish data of different formats in a processing server according to an embodiment of the present invention.

As shown at the top of Figure 3, quarantine documents received from different import companies or the National Fisheries Quality Institute contain ornamental fish information in different formats.

The data integration management unit 230 analyzes the similarity of characters in the table of contents list of different ornamental fish information and integrates them into the table of contents for objects with high similarity.

Through this, as shown at the bottom of FIG. 3, documents of different formats are constructed into a database that integrates the ornamental fish information list of the same format.

Figure 4 shows a deep neural network-based automatic search character error correction process and a fish species recommendation screen according to an embodiment of the present invention.

When you enter the text of the fish species you want to search in the quarantine app SW of the user terminal 100, the text is input in real time into the deep neural network model of the typo character correction unit 250 of the processing server 200, and creates a list of the most similar candidates. Select and extract.

As shown at the bottom of FIG. 4, the extracted character error correction list is provided through the quarantine app SW, and displays a list of fish species names similar to the characters searched by the user in order of accuracy.

The user can click on the listed search list, and when clicking, information about the selected fish species is retrieved from the data integration management unit 230 and provided to the user.

This function can be selected as enabled or disabled so that users can use the quarantine app SW flexibly.

Figure 5 shows deep neural network-based image recognition (fish species recognition/disease recognition) and ornamental fish DB data mapping according to the results according to an embodiment of the present invention.

When the user terminal 100 transmits ornamental fish data (including images) to the ornamental fish integrated management DB through the quarantine app SW, it is determined whether it is a newly introduced ornamental fish.

When there are no newly introduced ornamental fish, parameters extracted by learning from existing data in a deep neural network are used, as described above.

When there is a newly introduced ornamental fish, the existing data and the newly introduced data are integrated to perform re-learning in the deep neural network, and the newly learned learning parameters are used in the image recognition process.

When the user terminal 100 accesses the quarantine app SW and takes a picture of the ornamental fish to be quarantined while performing quarantine, the captured image is transmitted to the image recognition unit 240.

The image recognition unit 240 recognizes fish species and diseases in a multiple deep neural network using previously learned learning parameters from the received captured images.

The recognition results of fish species and diseases and the information of the data integration management unit 230 are mapped, and information about the corresponding ornamental fish is provided to the user.

Users refer to various information to quarantine ornamental fish in the quarantine app SW.

Figure 6 shows fish species recognition and disease recognition of a multi-deep neural network structure according to an embodiment of the present invention.

The multiple deep neural network is composed of multiple deep neural network structures rather than one deep neural network structure. As shown in Figure 6, fish species recognition and disease recognition are composed of multiple deep neural network structures, and separate recognition is performed for each. .

Figure 7 shows an ornamental fish recognition structure based on a multiple deep neural network according to a fish classification method according to an embodiment of the present invention.

Fish species recognition utilizes a methodology for classifying actual fish and performs learning and recognition by inputting body-shape, body-body color, caudal fin-shape, dorsal fin-shape, and snout-shape into each deep neural network.

The images needed for learning are collected in a format according to the fish classification method for fish species recognition, as shown on the left of Figure 7, and a dataset is constructed.

According to the input image, the results of each classification method are synthesized and analyzed to recognize the fish species.

At this time, the weights extracted from the deep neural network according to each classification method are flexibly assigned.

The standard for flexible weights is that a higher weight is assigned to an object in the deep neural network according to each classification method as the recognition accuracy is higher.

For example, when the input fish is red, the body-body color deep neural network uses the ratio when the red color has a high recognition rate and other blue, white, brown, and black colors have a low recognition rate in the body-body color deep neural network. A high weight is assigned to the extracted parameters.

Figure 8 shows an ornamental fish disease recognition structure based on a multi-deep neural network according to region division according to an embodiment of the present invention.

The user terminal 100 captures an image for quarantine through the quarantine app SW, and when the captured image is transmitted to the image recognition unit 240, the disease is recognized through a multiple deep neural network.

Disease recognition divides the input image into head, fin, body, and tail regions.

The divided areas are input into a deep neural network for each area to recognize the disease, and the disease recognition results from each area are integrated and sent to the user.

For example, it shows that the ornamental fish in question has parasites on its head, that there are wounds on its fins, and that there are no abnormalities in its body and tail.

Figure 9 shows the SW UI of an app dedicated to ornamental fish data mapping and quarantine according to the results of a deep neural network according to an embodiment of the present invention.

Quarantine is processed by accessing the quarantine app SW from the user terminal that includes the camera and WIFI communication module.

The quarantine app SW loads data on searched fish species from the data integration management unit 230 through search, or recognizes fish species and diseases in ornamental fish images taken from the user terminal through the image recognition unit 240, according to the recognition results. The information and mapping of the data integration management unit 230 are mapped and the ornamental fish data for quarantine is displayed in the quarantine app SW.

The quarantine list for quarantine of ornamental fish in the quarantine app SW includes, for example, 9 types, including ornamental fish name, classification, similar species, classification key, shape, ecology, disease, quarantine information, and photos.

If you find an ornamental fish through search or image recognition, you can check whether the ornamental fish is being quarantined in the quarantine app SW.

The quarantine officer carries out the quarantine by referring to the ornamental fish data from the quarantine app SW, and displays a screen where you can write about the stocking/non-stocking status of the ornamental fish currently undergoing quarantine, diseases of the ornamental fish, or other matters, and the completion of the current quarantine of the ornamental fish. Through the perfect function that tells the number, it is possible to understand the current situation at a glance.

Additionally, when you click on a photo of an ornamental fish in the quarantine process, you can see quarantine progress information for that ornamental fish at a glance.

If the quarantine officer is unable to identify the ornamental fish to be quarantined with the naked eye, the image and reference information of the ornamental fish in question are sent to another quarantine officer upon request from an expert.

When another quarantine officer completes the quarantine after confirmation and transmits the completed quarantine information to the existing quarantine officer, a sound or vibration notification is given in the quarantine app SW used by the existing quarantine officer so that the quarantine officer can immediately confirm.

According to the prior art, for imported ornamental fish, official ornamental fish information or packing boxes (ornamental fish information) for each company have different formats and contain slightly different information, so there are many documents to be referred to during quarantine, and these documents There is a problem of low reliability as people inspect ornamental fish with the naked eye. In addition, because the quarantine results are written by hand, the larger the number of ornamental fish, the more quarantine time is required, making it difficult to proceed smoothly.

On the other hand, according to an embodiment of the present invention, different formats of ornamental fish information can be integrated into a DB server that stores ornamental fish information for quarantine processing by quarantine officers, and in conjunction with this, quarantine information can be used on smartphones and smart pads. It has the advantage of enabling convenient quarantine processing by providing an app SW system.

According to an embodiment of the present invention, a learning-based basic structure is established to automatically correct error characters, recognize the type and disease of the ornamental fish, and map the ornamental fish information in the DB server. Learning parameters are extracted by training the learning engine in advance using a pre-collected ornamental fish image dataset, and the pre-trained learning parameters are loaded and used in the actual application device.

The deep learning engine specialized in ornamental fish recognition has a multi-deep neural network structure, and is divided based on fish classification (head, tail, fins, body, etc.), recognizes the shape in each deep neural network, and finally synthesizes the recognition results to identify fish species and diseases. It is possible.

According to an embodiment of the present invention, when new ornamental fish data is uploaded from the ornamental fish quarantine app SW to the ornamental fish DB (Server), the learning parameters of the deep learning engine are relearned and updated to recognize newly emerging fish species. There is.

In addition, the deep learning engine supports a service that requests quarantine from other experts even when aquarium fish are not detected, are misidentified, or the quarantine officer is unable to classify the type of aquarium fish, enabling rapid quarantine processing of aquarium fish. There is an advantage.

Meanwhile, the method for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification according to an embodiment of the present invention may be implemented in a computer system or recorded on a recording medium. A computer system may include at least one processor, memory, user input device, data communication bus, user output device, and storage. Each of the above-described components communicates data through a data communication bus.

The computer system may further include a network interface coupled to the network. A processor may be a central processing unit (CPU) or a semiconductor device that processes instructions stored in memory and/or storage.

Memory and storage may include various types of volatile or non-volatile storage media. For example, memory may include ROM and RAM.

Therefore, the method for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification according to an embodiment of the present invention can be implemented as a method executable on a computer. When the method for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification according to an embodiment of the present invention is performed on a computer device, computer-readable commands are transmitted to data preprocessing and deep neural network-based image detection and classification according to an embodiment of the present invention. Methods for quarantine of imported ornamental fish can be implemented through detection and classification.

Meanwhile, the method for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification according to the present invention described above can be implemented as a computer-readable code on a computer-readable recording medium. Computer-readable recording media include all types of recording media storing data that can be deciphered by a computer system. For example, there may be read only memory (ROM), random access memory (RAM), magnetic tape, magnetic disk, flash memory, and optical data storage devices. Additionally, the computer-readable recording medium can be distributed to a computer system connected through a computer communication network, and stored and executed as a code that can be read in a distributed manner.

A method for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification according to an embodiment of the present invention includes the steps of (a) receiving data from a user terminal, and (b) using the data to produce different formats. It includes the steps of integrating ornamental fish information and (c) using the data to determine fish species and disease based on a deep neural network and correcting errors included in the text.

Step (b) receives the ornamental fish packing list and the ornamental fish information quarantine document and integrates the different formats into a unified format.

In step (c), fish species and fish diseases are recognized using parameters already learned for images of ornamental fish, and retraining the artificial intelligence model is performed when new images are received.

In step (c), typos included in the ornamental fish search text are corrected, a list of fish species names higher than the standard similarity level is provided, and artificial intelligence model retraining is performed when new fish species text data is received.

Claims (13)

User terminals carried by quarantine officers; and
It includes a processing server that receives data from the user terminal, integrates ornamental fish information in different formats, performs fish species and disease determination based on a deep neural network, and automatically corrects errors contained in the text received from the user terminal; ,
The processing server includes a typo character correction unit, and the typo character correction unit uses pre-stored information to automatically correct typos included in ornamental word search characters to be searched, and recognizes similar characters through recognition in a deep neural network. A list of fish species of a set number is provided, but a list of fish species names above the standard similarity level is provided in order of accuracy, error-corrected results and a similar ornamental fish species name recommendation screen are provided, and similar ornamental fish species names included in the above similar ornamental fish species name recommendation screen are provided. The number of ornamental fish species names is less than the number of fish species lists of the preset number,
The processing server includes a data integration management unit, wherein the data integration management unit receives the ornamental fish packing list and the ornamental fish information quarantine document received during the ornamental fish import process, analyzes the format similarity of characters for the lists in different formats, and generates the Korean scientific name. , English scientific name, Korean product name, English product name, taxonomic species, taxonomic family, ecology, source code, morphology, genetic information, LMO possibility, classification key, ecosystem disturbing species, disease, CITES, other product names, photos, similar species, counting traits Integrate data in a unified format that includes,
The processing server includes an image recognition unit, wherein the image recognition unit recognizes fish species and diseases of the fish using pre-learned parameters for the ornamental fish photographed image received from the user terminal, and the image recognition unit recognizes the fish species. Using multiple deep neural networks for and multiple deep neural network learning parameters to recognize diseases in the fish,
The image recognition unit classifies fish species by learning and recognizing body-shape, body-color, caudal fin-shape, dorsal fin-shape, and snout-shape using each deep neural network, taking recognition accuracy into consideration. Dynamically assigning weights extracted from each deep neural network above
A device for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification.
delete delete delete According to paragraph 1,
The processing server maps pre-stored fish information and the fish species and disease recognition results of the fish, and transmits the data to the user terminal. A device for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification.
According to paragraph 1,
When the image recognition unit receives new data rather than existing data from the user terminal, data preprocessing and deep neural network-based image detection and classification are used to retrain the artificial intelligence model using new image data. A device for quarantine of ornamental fish.
delete According to paragraph 1,
When receiving new data rather than existing data from the user terminal, the typo character correction unit performs data preprocessing and deep neural network-based image detection and classification, characterized in that it retrains the artificial intelligence model using new fish species text data. A device for quarantine of imported ornamental fish.
According to paragraph 1,
The processing server requests a reading from another expert's device, and provides an alarm to the user terminal upon completion of the reading. A device for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification.
delete delete delete delete