KR102411209B1 - System and Method for Image Classification Based on Object Detection Events by Edge Device - Google Patents

Abstract

The present invention provides an image classification system and method based on an object detection event generated by an edge device, a photographing module for generating image data by photographing an image, A first analysis module that analyzes the shape and behavior to determine the type of the subject as any one of a person, an animal, a plant, and an object to generate a large classification code corresponding to the type of the subject and metadata about the position of the subject; An image photographing apparatus comprising: an image processing module for generating post-processing data by merging metadata; a first communication module for transmitting the post-processing data generated through the image processing module; and a second communication module for receiving post-processing data from the image capturing device, an extraction module for extracting large classification codes and location information among metadata included in the post-processing data received by the second communication module, and the extraction module. A second analysis module that sub-classifies according to the classification criteria stored in the storage module for each major classification code extracted through the sub-classification, adds the sub-classification code corresponding to the sub-classification to metadata, and then merges it with the image data to regenerate post-processing data; It relates to an object detection event-based image classification system and method generated by an edge device including a main server. According to the present invention, inference is performed on both the edge (image photographing apparatus in the present invention) device and the main server to detect an object in the edge device and classify the object in the main server to tag, store, or photograph the image so that the user can It is possible to provide an image classification system and method based on an object detection event generated by an edge device that can perform image search under various conditions using tags and tag combinations.

Description

Image classification system based on object detection event generated by edge device {System and Method for Image Classification Based on Object Detection Events by Edge Device}

The present invention relates to an image classification system and method based on an object detection event generated by an edge device, and more particularly, to a photographing module that captures an image to generate image data, and a subject is detected in an image input through the photographing module a first analysis module that analyzes the shape and behavior of the subject, determines the type of the subject as any one of a person, an animal, a plant, and an object, and generates a large classification code corresponding to the type of the subject and metadata about the position of the subject; an image photographing apparatus comprising: an image processing module configured to generate post-processing data by merging the metadata with the image data; and a first communication module configured to transmit the post-processed data generated through the image processing module; and a second communication module for receiving post-processing data from the image capturing device, an extraction module for extracting large classification codes and location information among metadata included in the post-processing data received by the second communication module, and the extraction module. A second analysis module that sub-classifies according to the classification criteria stored in the storage module for each major classification code extracted through the sub-classification, adds the sub-classification code corresponding to the sub-classification to metadata, and then merges it with the image data to regenerate post-processing data; It relates to a system and method for classifying an image based on an object detection event generated by an edge device including a main server.

The present invention relates to a cloud edge system for transmitting and receiving image data.

The existing video-based CCTV system transmits video information acquired from the video collection device (camera) of the endpoint of the system using a medium such as a wired/wireless network or cable.

Depending on the interface between the camera and the control device, it is possible to encode the video at the endpoint or to perform it in a separate device at a remote location.

The artificial intelligence inference system configured with a structure similar to the above is also configured in such a way that an inference process is performed at the endpoint using the image information acquired from the camera or the inference is performed by transmitting the acquired image information to a remote location.

The configuration of the existing artificial intelligence system (in various applicable fields such as object detection, recognition, tracking, classification, etc.) is a server-based system using a deep and wide neural network-based learning model and an edge-based system using a shallower lightweight learning model. is divided into a system of

However, it is impossible to execute various AI-based inferences such as object detection, identification, classification, and tracking in one device due to limitations in performance due to restrictions such as installation location and instrument shape of each edge device (image collection device). .

In the case where each node receives the video signal that has not been subjected to the inference process and executes all inference operations on the central server, the number of controllable nodes per server is significantly reduced, thereby reducing scalability and efficiency.

Although the technology that is the background of the present invention is disclosed in Korean Patent Publication No. 10-2152237, etc., a fundamental solution to the above-described problem is not presented.

Republic of Korea Patent Publication No. 10-2152237

The purpose of the present invention, which was devised to solve the above-described problems, is to detect an object in the edge device and detect the object in the main server by performing inference in both the edge (image capturing apparatus in the present invention) device and the main server. To provide an image classification system and method based on an object detection event generated by an edge device that allows users to perform image search under various conditions using tags and tag combinations in the future by classifying, tagging, storing, or shooting images. .

In addition, another object of the present invention is that, when one or more objects are detected and the moving directions of the objects are the same, a code for a plant or object is attributed to a code for a person or an animal to generate one or more codes. This is to provide an image classification system and method based on an object detection event generated by an edge device.

According to a feature of the present invention for achieving the above object, the object detection event-based image classification system generated by the edge device of the present invention includes a photographing module for generating image data by photographing an image, and the photographing module When a subject is detected in an image input through a first analysis module generating is generated, it generates a PTP (Precision Time Protocol) synchronization signal and transmits it to one or more adjacent imaging devices, and periodically synchronizes the time of one or more imaging devices through PTP to create metadata an image photographing device including a first communication module for matching reaming signals; and a second communication module for receiving post-processing data from the image capturing device, an extraction module for extracting large classification codes and location information from metadata included in the post-processing data received by the second communication module, and the extraction A second analysis module for sub-classifying according to the classification criteria stored in the storage module for each major classification code extracted through the module, adding the sub-classification code corresponding to the sub-classification to metadata, and merging it with the image data to regenerate post-processing data; By monitoring the load generated according to data transmission and reception and data processing, when the load is higher than the preset load, only the image data is received from the image photographing device, the metadata is generated through the second analysis module, and the metadata is added to the image data. A control module for generating post-processing data by merging, and parsing metadata of one or more post-processing data generated through a second analysis module according to a major classification code and a sub-classification code of metadata included in the post-processing data a main server including a storage module for classifying and storing processed data; and a user terminal that generates search information including any one or more of a major classification code or a sub-category code and time range information and transmits the generated search information to the main server.
At this time, the first analysis module determines that the subject determined to be a person or a plant belongs to the subject determined to be a person or an animal if the subject determined to be an object or a plant moves in the same direction as the movement direction of the subject determined to be a person or an animal. or a large classification code for a subject determined to be an animal, a large classification code attributed to the large classification code, and metadata about a location, wherein the second analysis module includes any one or more of a large classification code or a small classification code and a location According to the information, any one or more additional processing of preset masking processing, mosaic processing, and personal identification processing is performed and added to the metadata, and one or more major classification codes are subdivided and each sub-classification code corresponding to the sub-classification is added to the metadata After merging with the corresponding image data, the post-processing data is regenerated, and the image quality is changed by regenerating the post-processing data.
And the second analysis module further includes a facial analysis module that analyzes the image data received from the image photographing device and analyzes the race, gender, age group or emotion of each subject by using the statistical data of the subject's face shape, , the main server detects pre-stored post-processing data within a time range corresponding to the search information and detects at least one of the detected post-processing data when search information including any one or more of a major classification code and a sub-category code and time range information is received from the user terminal The post-processing data is streamed to the user terminal, but the post-processed data is sequentially streamed according to the movement of the subject corresponding to the search information, or one or more post-processed data are streamed at the same time.

delete

delete

delete

delete

delete

delete

delete

According to the present invention as described above, inference is performed on both the edge (image capturing apparatus in the present invention) device and the main server to detect an object in the edge device and classify the object in the main server to tag, store, or shoot an image. By doing so, it is possible to provide an image classification system and method based on an object detection event generated by an edge device that allows a user to perform image search under various conditions using tags and tag combinations in the future.

In addition, according to the present invention, when one or more objects are detected and the moving directions of the objects are the same, a code for a plant or object is attributed to a code for a person or an animal to generate one or more codes, which makes it easier to search for an edge device It is possible to provide an image classification system and method based on the object detection event generated in .

1 is an exemplary diagram illustrating the configuration of an image classification system based on an object detection event generated by an edge device according to an embodiment of the present invention;
2 is a block diagram showing the configuration of an image classification system based on an object detection event generated by an edge device according to an embodiment of the present invention;
3 is a flowchart illustrating a method of providing an image classification method based on an object detection event generated by an edge device according to an embodiment of the present invention.

Advantages and features of the present invention and methods of achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings.

However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only these embodiments allow the disclosure of the present invention to be complete, and common knowledge in the art to which the present invention pertains It is provided to fully inform those who have the scope of the invention, and the present invention is only defined by the scope of the claims. Like reference numerals refer to like elements throughout.

Hereinafter, the present invention will be described with reference to the drawings for explaining an image classification system and method based on an object detection event generated by an edge device according to embodiments of the present invention.

The object detection event-based image classification system created by the present invention edge device is for efficient channel management and performance improvement of a system that monitors, controls, and records multi-channel images (CCTV control system or comprehensive image management system such as VMS). will be.

Integrated post-processing data (images) including metadata acquired through an inference process in the image data by analyzing and inferring images acquired through the image processing module 130 installed inside or outside the image capturing apparatus 100 data packet) and transmit, it is possible to increase transmission efficiency and the number of simultaneous channels that can be processed.

Here, the post-processing data is preferably encoded in a form that satisfies the ONVIF (Open Network Video Interface Forum; hereinafter ONVIF) protocol, but is not limited thereto. If image data can be decoded and output by the user terminal 300, It is desirable to be encoded in various ways.

The imaging apparatus 100 described throughout the present invention is a device for acquiring an image, and generally preferably a camera, but receiving an image signal such as an optical sensor, thermal imaging camera, 3D-Depth camera, lidar, etc. If possible, it can be used as the image photographing apparatus 100 .

It is preferable that the user terminal 300 described throughout the present invention outputs post-processing data received from the image capturing apparatus 100 and the main server 200 in the form of a personal computer, a notebook computer, etc. in general. It is not limited. That is, if it is possible to transmit/receive data to and from the image photographing apparatus 100 and the main server 200 such as the mobile phone terminal 300, the smart phone terminal 300, the PDA terminal 300, etc. by wire or wirelessly, the user terminal 300 available as

1 is an exemplary diagram showing the configuration of an image classification system based on an object detection event generated by an edge device according to an embodiment of the present invention, and FIG. 2 is based on an object detection event generated by an edge device according to an embodiment of the present invention. It is a block diagram showing the configuration of the image classification system of

1 and 2 , the object detection event-based image classification system and method generated by the present invention edge device includes an image capturing apparatus 100 , a main server 200 , and a user terminal 300 .

Here, the image photographing apparatus 100 includes a photographing module 110 , a first analysis module 120 , an image processing module 130 , and a first communication module 140 .

The photographing module 110 generates image data by photographing an image.

Here, the photographing module 110 further includes a voice recognition module 112 .

The voice recognition module 112 recognizes a voice.

The first analysis module 120, when a subject is detected in the image input through the photographing module 110, analyzes the shape and behavior of the subject to determine the type of the subject create data

That is, the first analysis module 120, when a subject is detected, analyzes the shape and behavior of the subject, determines the type of the subject as any one of a person, an animal, a plant, and a large classification code corresponding to the type of the subject and Generate metadata about the subject's location. Here, it is preferable that the first analysis module 120 performs analysis and reasoning operations such as object detection/identification/tracking/classification of the image data acquired through the photographing module 110 .

In addition, the first analysis module 120 determines that the subject determined to be an object or a plant belongs to the subject determined to be a person or an animal when the subject determined to be a human or an animal moves in the same time period as the moving direction of the subject. Thus, a large classification code for a subject determined to be a human or an animal, a large classification code attributed to the large classification code, and metadata about a location are generated.

That is, when a person or an animal moves together through various methods such as gripping an object or a plant, it is desirable to determine that it is attributed and generate metadata for each major classification code.

If an object or plant moves by itself without action of a person or an animal, it is preferable to generate an abnormal signal and transmit it to the user terminal.

The image processing module 130 generates post-processing data by merging the metadata with the image data. Here, the metadata is preferably a result generated through analysis and inference of image data. In addition, it is preferable to use a method in which the metadata is included in a Realtime Transport Control Protocol (RTCP) packet transmitted in real time. Here, it is preferable to designate the format of the RTCP packet as 204 (RFC1899, Application specific RTCP).

RTCP packets are PAYLOAD-TYPE, OBJECT-COUNT (OBJECT-ID, OBJECT-BBOX-LEFT-TOP coordinates, OBJECT-BBOX-RIGHT-BOTTOM coordinates, etc. It is preferable to include, etc. And, the image processing module 130 generates the post-processing data by including the generated voice data recognized through the voice recognition module in the post-processing data when generating the post-processing data.

The first communication module 140 transmits the post-processing data generated through the image processing module 130 . Here, the first communication module 140 generates a PTP (Precision Time Protocol; hereafter referred to as PTP) synchronization signal when the metadata is generated by the first analysis module 120 to be transmitted to one or more adjacent image photographing devices 100 . It is characterized by transmitting.

That is, it is preferable that the first communication module 140 periodically synchronizes the time of one or more image photographing apparatuses 100 through PTP to match the image streaming signals of the connected image photographing apparatuses 100 when generating metadata.

When N image photographing apparatuses 100 are installed, the time of the N image photographing apparatuses 100 is synchronized, and the synchronized image data is received by the main server 200 or the user terminal 300 to display the entire picture. It is preferable to configure

The main server 200 includes a second communication module 210 , an extraction module 220 , a second analysis module 230 , a control module 240 , and a storage module 250 .

The second communication module 210 receives post-processing data from the image capturing apparatus 100 . Here, since the second communication module 210 receives post-processing data in which metadata is encoded in the image data, it is unnecessary to execute a separate data parser, and the process of sorting and processing according to the time sequence is also unnecessary. In addition, since the inference result of the post-processed data in which the media stream is in-band can be received in an arranged state in the order of execution time, an image including metadata is directly provided to the user terminal 300 without a separate encoding or sorting process It is possible.

The extraction module 220 extracts the type of object, that is, the large classification code and location information, from among the metadata included in the post-processing data received by the second communication module 210 .

The second analysis module 230 sub-classifies according to the classification criteria stored in the storage module for each major classification code extracted through the extraction module 220, adds the sub-classification code corresponding to the sub-classification to the metadata, and then merges it with the image data. Regenerate post-processing data.

The criteria for subdividing classification in the second analysis module 230 are as follows.

People - race, age, gender, etc.

Animals - classification by species, classification of breeds, etc.

Plants - classification, etc.

Object - Car, fire, smoke, road surface, climate (rainfall, snowfall, etc.)

After the sub-classification as above, it is desirable to classify by adding more subdivided classification criteria according to each sub-classification item. For example, in the case of a car, it can be divided into sub-categories such as car model, color, and model and license plate in case of the same car model.

In the case of a person, identification through the facial analysis module, body parts, postures, and clothes may be classified into subdivisions. That is, it is preferable that the large classification divides people, animals, plants, and things, and the sub-classification divides the characteristics of the subject divided into the large classifications. In addition, the second analysis module performs additional processing of any one or more of masking processing, mosaic processing, and personal identification identification processing preset according to any one or more of the large classification code or the small classification code and location information, and adding it to the metadata.

Through this, when detection information for different types of objects is transmitted from N different image capturing apparatuses A-Z, it is also possible to perform different additional processing according to the major classification code or the small classification code. In addition, the second analysis module 230 re-analyses the image data included in the post-processing data when the accuracy value included in the metadata received by the second communication module 210 is less than or equal to a preset accuracy value, and the Post-processing data is regenerated by merging the reanalyzed metadata with the image data.

Here, the second analysis module 230 may change the image quality by reprocessing the quality of the received image data. And, the second analysis module 230 further includes a face analysis module (232). The facial analysis module 232 analyzes the post-processing data to analyze the shape of the subject's face, compares the analyzed facial shape with one or more pre-stored face data of the subject, and generates an intrusion signal if the match is less than or equal to a preset level of agreement to generate an intrusion signal and transmits the post-processing data to the user terminal 300 . That is, if a new person who is not already registered enters and exits, it is preferable to transmit a signal indicating that the new person has entered the user terminal 300 so that the user can confirm that the new person has entered. Then, when the sub-classification code is generated, the second analysis module 230 detects voice data previously matched to the sub-classification code, compares and analyzes the detected voice data with the voice data stored in the post-processing data, so that the degree of matching is less than a preset matching level. An abnormal signal is generated and transmitted to the user terminal, and if the matching level is greater than the preset level, personal identification processing is performed. That is, if the subject is a person, personal identification processing is performed to detect the stored voice data that is matched with the subclassification code generated in advance, compares it with the voice data included in the post-processing data to determine whether it is the same person, and then identifies the same person When the processing is completed and it is determined that it is not the same person, it is preferable that the user judges the abnormality by simulating the individual corresponding to the subclassification code and infiltrating, generating an abnormal signal and sending it to the user terminal 300 .

Here, the facial analysis module 232 analyzes the image data received from the image capturing device 100 and analyzes the race, gender, age, or emotion of each subject by using the statistical data of the subject's face shape. desirable. That is, it is preferable to analyze the race, gender, age group, and emotion of the subject through the facial analysis module 232 using pre-stored statistical data. In addition, the facial analysis module 232 analyzes the facial shape of the subject when the large classification code included in the metadata is a person, compares the analyzed facial shape with one or more pre-stored facial data of the subject, and matches a preset degree or more In this case, personal identity identification processing is performed, and additional processing set in advance with the subject's facial data is performed. That is, personal identification processing is performed through the facial analysis module 232, and when a pre-stored individual is detected, additional matching processing is performed. ) is preferably transmitted to

Here, the additional processing is set by matching in advance with the large classification code or the sub-category code.

In addition, it is also possible for the user to set up additional processing according to the large classification code or the sub-category code.

The control module 240 monitors a load generated according to data transmission/reception and data processing, and when the load is higher than a preset load, receives only image data from the image capturing device 100 and transmits the image data through the second analysis module 230 . Post-processing data is generated by generating data and merging the metadata with the image data.

Here, it is preferable that the control module 240 monitors and controls the resources, processes, etc. of the main server 200 . In addition, it is preferable that the control module 240 monitors and controls the communication-related data (network speed, network connection state and record) of the image output device and the main server 200 . That is, it is preferable that the control module 240 monitors the state of the image output device and the main server 200, and monitors the load state and network connection state of each edge (image output device) to switch the analysis method.

The storage module 250 parses the metadata of one or more post-processing data generated through the second analysis module 230, and the post-processing data according to the major classification code and the sub-classification code of the metadata included in the post-processing data. Classify and save.

Here, when the main server 200 receives search information including any one or more of the major classification code and the sub-category code and time range information from the user terminal 300, pre-stored post-processing data within the time range corresponding to the search information Stream the detected and detected one or more post-processed data to the user terminal 300, but sequentially stream the post-processed data according to the movement of the subject corresponding to the search information or stream one or more post-processed data at the same time ,

That is, when the user terminal 300 searches for a specific subject (subdivision code) between 10:00 and 12:00 on January 1 in a specific time period, the main server 200 sends the information to the user terminal 300 on January 1 It is checked whether there is a specific subject detected by the one or more image capturing apparatuses 100 between 10 o'clock and 12 o'clock. It is possible to stream and output as , and it is also possible to simultaneously output one or more post-processed data in which the subject is observed to the user terminal 300 .

If a specific subject is not detected, it is also possible to search by gradually expanding the search range from 10:00 to 12:00 on January 1st. Here, the main server 200 makes it possible to configure the entire screen according to the search information, and the screen is output to the user terminal 300 and recording and storage are possible at the same time.

When the user terminal 300 receives post-processing data from one or more image capturing apparatuses 100 or the main server 200, the user terminal 300 synchronizes and outputs the one or more received post-processed data through the PTP synchronization signal. In addition, the user terminal 300 generates search information including at least one of specific subject information and behavior information, and transmits the generated search information to the main server 200 . In addition, when the intrusion signal is received from the main server 200, the user terminal 300 outputs post-processing data corresponding to the intrusion signal on the screen.

3 is a flowchart illustrating a method of providing an image classification method based on an object detection event generated by an edge device according to an embodiment of the present invention.

Referring to FIG. 3 , first, an image photographing apparatus captures an image to generate image data ( S110 ).

Then, when a subject is detected in the image input through the step (S110), the shape and behavior of the subject are analyzed to determine the type of the subject as any one of a person, an animal, a plant, and an object, which corresponds to the type of the subject. A large classification code and metadata about the location of the subject are generated (S120). Here, in the step (S120), if the subject determined to be an object or a plant moves in the same direction as the movement direction of the subject determined to be a person or an animal, it is determined as belonging to the subject determined as the person or animal, and a person Alternatively, a large classification code for a subject determined as an animal, a large classification code attributed to the large classification code, and metadata about a location are generated. Then, post-processing data is generated by merging the metadata with the image data (S130). Then, the image photographing apparatus transmits the post-processing data generated in step (c) to the main server (S140). Then, the main server receives the post-processing data from the image photographing apparatus (S150). Next, a large classification code and location information are extracted from the metadata included in the post-processing data received by the second communication module (S160). Then, the main server sub-classifies according to the classification criteria stored in the storage module for each major classification code extracted through the step (S160), adds the sub-classification code corresponding to the sub-classification to the metadata, and then merges it with the image data for post-processing The data is regenerated (S170). Here, in the step (S170), any one or more of masking processing, mosaic processing, and personal identification identification processing set in advance according to any one or more of a large classification code or a small classification code and location information are additionally processed and added to the metadata.

In addition, the main server parses the metadata of one or more post-processing data generated through the step (S170) and classifies the post-processing data according to the major classification code and the sub-classification code of the metadata included in the post-processing data. Save.

In addition, in the step (S170), when the large classification code included in the metadata is a person, the facial shape of the subject is analyzed, and the analyzed facial shape is compared with one or more pre-stored facial data of the subject. It generates an intrusion signal and transmits the intrusion signal and the corresponding post-processing data to the user terminal.

At this time, when the intrusion signal is received from the main server, the user terminal outputs post-processing data corresponding to the intrusion signal on the screen.

Then, when the large classification code included in the metadata is a human, the facial shape of the subject is analyzed, and the analyzed facial shape is compared with one or more pre-stored face data of the subject. and additional processing set in advance with the subject's facial data.

Here, the additional processing is set by matching in advance with the large classification code or the sub-category code.

Finally, when search information including any one or more of a major classification code and a sub-category code and time range information is received from the user terminal, the main server detects pre-stored post-processing data within the time range corresponding to the search information, and detects the detected one The above-mentioned post-processed data is streamed to the user terminal, but the post-processed data is sequentially streamed according to the movement of the subject corresponding to the search information, or one or more post-processed data are streamed at the same time (S180). At this time, the user terminal generates search information including any one or more of a major classification code or a sub-category code and time range information and transmits the generated search information to the main server.

Those of ordinary skill in the art to which the present invention pertains will understand that the present invention may be embodied in other specific forms without changing the technical spirit or essential characteristics thereof. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. The scope of the present invention is indicated by the following claims rather than the above detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts are included in the scope of the present invention. should be interpreted

100: image photographing device 110: photographing module
112: voice recognition module 120: first analysis module
130: image processing module 140: first communication module
200: main server 210: second communication module
220: extraction module 230: second analysis module
232: facial analysis module 240: control module
250: storage module 300: user terminal

Claims (8)

A photographing module that generates image data by photographing an image, and when a subject is detected in the image input through the photographing module, analyzes the shape and behavior of the subject to determine the type of the subject as any one of a person, an animal, a plant, and an object A first analysis module that determines and generates a large classification code corresponding to the type of object and metadata on the location of the object, and post-processing that can be communicated with the ONVIF (Open Network Video Interface FORUM) protocol by merging the metadata with the image data When the metadata is generated by the image processing module for generating data and the first analysis module, a PTP (Precision Time Protocol) synchronization signal is generated and transmitted to one or more adjacent image photographing apparatuses, and the one or more image photographing apparatuses through PTP an image photographing apparatus including a first communication module that synchronizes time periodically to match image streaming signals of a connected image photographing apparatus when generating metadata; and
a second communication module for receiving post-processing data from the image capturing device; an extraction module for extracting large classification codes and location information from among metadata included in the post-processing data received by the second communication module; and the extraction module A second analysis module for sub-classifying according to the classification criteria stored in the storage module for each major classification code extracted through adding the sub-classification code corresponding to the sub-classification to metadata and merging it with the image data to regenerate post-processing data;
By monitoring the load generated according to data transmission and reception and data processing, when the load is higher than the preset load, only the image data is received from the image photographing device, the metadata is generated through the second analysis module, and the metadata is added to the image data. A control module for generating post-processing data by merging;
Parsing the metadata of one or more post-processing data generated through the second analysis module, and a storage module for classifying and storing the post-processing data according to the major classification code and the sub-classification code of the metadata included in the post-processing data. main server;
A user terminal that generates search information including any one or more of a major classification code or a sub-category code and time range information and transmits it to the main server,
The first analysis module,
If the subject determined to be an object or plant moves in the same direction as the movement direction of the subject determined to be a person or an animal, it is determined that the subject determined to be a person or an animal belongs to the subject determined to be a person or an animal. It is characterized in that the large classification code, the large classification code attributed to the large classification code, and metadata about the location are generated,
The second analysis module,
Add to metadata by performing additional processing of any one or more of masking processing, mosaic processing, and personal identification processing set in advance according to any one or more of the large classification code or the small classification code and location information;
After sub-classifying one or more major classification codes, adding each sub-classification code corresponding to the sub-classification to metadata, merging with the corresponding image data to regenerate post-processing data, and regenerating post-processing data to change the image quality,
The second analysis module,
It further comprises a facial analysis module that analyzes the image data received from the image recording device and uses the statistical data of the subject's face shape to analyze the race, gender, age group or emotion of each subject,
The main server is
When search information including any one or more of a major classification code and a sub-category code and time range information is received from the user terminal, pre-stored post-processing data within a time range corresponding to the search information is detected, and the detected one or more post-processed data are stored in the An object detection event-based image classification system created by an edge device that streams to a user terminal but streams post-processing data sequentially or one or more post-processing data at the same time according to the movement of the subject corresponding to the search information .

delete delete delete delete delete delete delete

Images