KR20170123422A - Monitoring target classification and selective people counting system using infrared marker and video image analysis technology - Google Patents

Abstract

The present invention relates to a system for counting the number of people and, more specifically, to a classification and visitor counting system to define classification of a monitoring target, which is detected by a video analysis technology, by using an infrared marker and to perform selective counting by the defined classification when the monitoring target passes a specific region. According to the present invention, a monitoring target classification and selective visitor counting system using an infrared marker and a video analysis technology counts a monitoring target through a camera, acquires an image of an infrared marker through an infrared receiving unit, and selectively counts the monitoring target in accordance with the detection of the infrared marker. According to the present invention, it is possible to automatically classify the monitoring target by combining the infrared marker and the video monitoring technology, and to count the number of people for each classified target to collect data in accordance with the situation.

Description

Field of the Invention < RTI ID = 0.0 > [0001] < / RTI > A monitoring target classification and selective visitor counting system using infrared markers and image analysis techniques,

The present invention relates to a system for counting the number of people, and more particularly, to a system for counting a number of people, more specifically, to a system for counting a number of people, And a classification and personnel counting system for selectively counting by classification.

Korean Patent Laid-Open Publication No. 10-2014-0047714 discloses a method and system for counting personnel using passive infrared detectors, in which a plurality of infrared sensors are installed at a door to analyze patterns of infrared rays detected, Was used.

The method using infrared sensor is advantageous in that it has a good performance for detecting the passing, but it has a disadvantage that it is difficult to intuitively check the situation occurring at the door because there are many restrictions on the installation of the sensor.

In recent years, the development of techniques for using images such as CCTV has been actively developed in comparison with a method using an infrared sensor.

CCTV, etc., it is possible to automatically detect the person moving to the image analysis technology and to count the number of people passing through a specific area, and to be able to intuitively observe the situation occurring at the place where the CCTV is installed .

In Korean Patent Registration No. 10-0885418, a method for detecting and tracking a person in an overhead camera image is disclosed. In Korean Patent Registration No. 10-1377394, a stereo camera is used to enhance the performance of image analysis. Techniques for systems and methods for automatically counting numbers are disclosed.

Personnel counting technology using image analysis is advantageous in that it is convenient to install and can observe the situation intuitively, but there is a disadvantage in that it provides only the function of counting past passages because the passengers can not be classified by the image alone.

For example, in the case of a store, it is difficult to effectively divide the measured personnel count data into the store operation, because it is difficult to distinguish the employee who is passing through the door or the visiting customer.

In order to overcome such disadvantages, many studies have recently been made on a method of using a wireless device as disclosed in Korean Patent Registration No. 10-1171281.

Many researches have been conducted on a method of counting the number of persons having wireless devices by detecting signals generated by devices having unique IDs such as BLE (Blue-tooth Low Energy), WiFi, and RFID.

However, due to the characteristics of wireless communication, especially in the indoor environment, it is difficult to accurately analyze the strength and characteristics of received radio waves in accordance with the arrangement of furniture, interference between radio devices, and the state of radio equipment.

Therefore, it is very difficult to find out the 1: 1 correspondence relationship between the ID given by the wireless device and the person who has the wireless device if the person possessing the wireless device exists in the same space. Because of this difficulty, there is a disadvantage in that the performance of the method using the wireless device is degraded in the counting task that can be detected only when the ID of the wireless device is exactly matched with the ID of the wireless device.

For example, if two people who have a wireless device and two people who do not have one are in the same space and move in the same direction, it is very difficult to define which person has a wireless device. it's difficult.

And most of all, it has many difficulties to combine the measured data from the wireless device with the surveillance object detected by the image analysis technology, so it has a disadvantage that there is a lot of synergy effect with the image analysis technology.

Patent Document 1: Korean Patent Publication No. 10-2014-0047714 Patent Document 2: Korean Patent Publication No. 10-0885418 Patent Document 3: Korean Patent Publication No. 10-1377394 Patent Document 4: Korean Patent Publication No. 10-1171281

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems occurring in the prior art, and it is an object of the present invention to provide a method of defining a classification of a monitoring target using an infrared marker, .

According to another aspect of the present invention, there is provided a method for classifying and selecting a visitor object using an infrared marker and an image analysis technique, comprising the steps of: counting an object to be monitored through a camera; acquiring an image of an infrared marker through an infrared receiver; And the monitoring target is selectively counted according to the detection of the infrared marker.

Here, the infrared marker may be a one-dimensional pattern or a two-dimensional pattern.

Here, the object to be monitored can be re-selected and counted through a combination of at least one of the pattern of the infrared marker, the color, and the divergence period of the infrared LED.

Here, the light source of the infrared marker may be performed according to a passive method of emitting an infrared ray from an infrared ray LED and composed of an infrared ray reflective material, or an active method of emitting infrared rays independently including an infrared ray LED.

Here, the monitored object is detected, classified and counted through an image analysis module. The infrared marker is detected by a marker detection unit, and transmitted to the image analysis module. The camera image and the infrared image are matched with each other, Can be counted.

Here, it is preferable that the infrared marker is assigned a classification attribute such that the infrared marker corresponds to one group of the monitoring target at 1: 1.

Here, when a new infrared marker is detected in the course of real-time tracking of the monitored object, the classification attribute for the new infrared marker can be updated by comparing the attribute with the pre-stored infrared marker and the classification criteria.

Here, the selective visitor counting system may be classified into a real-time classification by the real-time classification unit for real-time classification according to the classification attribute, and a post-classification of the personnel coefficient for correcting the number of persons counted by the real-time classification.

According to the configuration of the present invention described above, the monitoring target can be automatically classified by combining the infrared marker and the video surveillance technology, and the number of persons can be counted according to the classified target. Therefore, There is an advantage that it can be.

Specifically, it is possible to measure the number of pure visitors excluding the number of past employees, and it is also advantageous that the pattern of entrance and exit can be collected separately for employees and visitors over time.

In addition, the monitoring function for the controlled area can be implemented efficiently by detecting the entrance of a person who does not wear the marker.

In addition, by using the classification method for the monitoring object according to the configuration of the present invention, it is possible to easily and accurately classify the monitoring object by easily combining with the image analysis technology. Accordingly, not only the selected personnel coefficient but also the functions It can be implemented easily and effectively through combination with image analysis technology.

1 is a schematic diagram of a selective visitor counting system using an infrared marker and an image analysis technique according to the present invention.
2 is a schematic diagram of a method of classifying a surveillance object using an infrared marker and an image analysis technique according to the present invention.
Figure 3 is an illustration of an infrared marker pattern applied to the selective visitor counting system according to the present invention.
FIG. 4 is a view illustrating a passive infrared marker and detection configuration of the selective visitor counting system according to the present invention, and an infrared marker and detection configuration of the active method.
FIG. 5 is a flowchart of a method of detecting an infrared marker in the selective visitor counting system according to the present invention.
6 is a flowchart of a method of classifying a monitored object in an image analysis module using an infrared marker in the selective visitor counting system according to the present invention.
FIG. 7 is a conceptual view for assigning a classification attribute to a monitored object using an infrared marker in the selective visitor counting system according to the present invention.
FIG. 8 is a diagram showing a classification attribute granting and updating sequence of a monitoring object in the selective visitor counting system according to the present invention.
9 is a flowchart of a selective personnel counting process using the classification attribute in the selective visitor counting system according to the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a monitoring target classification method, a selective visitor counting system, and an action effect using an infrared marker and an image analysis technique according to the present invention will be described in detail with reference to the accompanying drawings.

The detailed description of the specific embodiments shown in the accompanying drawings is read in conjunction with the accompanying drawings, which are considered a part of the description of the entire invention. The reference to direction or orientation is for convenience of description only and is not intended to limit the scope of the invention in any way.

Terms indicating the position or mutual coupling relationship between configurations should be understood with reference to both the drawings and the related descriptions.

FIG. 1 is a schematic view of a selective visitor counting system using an infrared marker and an image analysis technique according to the present invention, and FIG. 2 is a schematic diagram of a method of classifying a monitored object using an infrared marker and an image analysis technique according to the present invention.

The present invention provides a method of classifying an object to be monitored using an infrared marker. The infrared marker may be classified into a passive infrared marker and an active infrared marker according to an operation method, and a detector for detecting an infrared marker may be configured differently according to the method.

1, a selective visitor counting system using an infrared marker and an image analysis technique according to the present invention includes a camera 111 and an infrared ray receiving unit 113 for photographing a shop visitor 115a and a staff 115b And the captured image is analyzed and the image analysis result 117 and the infrared marker detection 119 detected by the infrared ray receiver 113 are matched to determine whether the visitor is a shop visitor or an employee, And stores them in the visitor DB 123a and the employee DB 123b, respectively.

In the present invention, the surveillance target can be basically divided into a shop visitor 115a having a general object and an employee 115b having an infrared marker.

In addition, the authorized personnel 115b can be defined in different patterns so as to have various categories according to the application of the personnel coefficient.

For example, in the case of the counting system for the shop visitor (115a) and the employee (115b), the monitoring target that does not have an infrared marker is defined as a shop visitor, and the infrared marker , And payment staff, it is possible to define the target of monitoring the employees.

2, in the present invention, a camera 111 and an infrared ray receiving unit 113 are used to distinguish the entering and exiting personnel, and the camera 111 can use a color camera.

The image obtained by the camera 111 is detected and tracked by the object detection unit 221 of the image analysis module 211.

When passing through a specific area (meaning a detectable area), the object passing detection unit 225 counts the number of persons.

The marker detecting unit 213 detects an infrared marker held by an employee from the image acquired by the infrared ray receiving unit 113 and analyzes the pattern of the marker and then transmits the classification and position information of the marker to the object classification and tracking unit 223 do.

In the object classification and tracking unit 223 of the image analysis module 211 receiving the classification and position of the infrared marker, the personnel are counted selectively by referring to the classification of the object to be monitored, .

The object passage sensing unit 225 refers to the classification of objects to be monitored and stores the infrared marker in the visitor DB 123a as a shop visitor in accordance with the presence or absence of the infrared marker in accordance with the pattern type of the infrared marker, The first employee DB 123b, the second employee DB 123c, ..., the nth employee DB in accordance with the pattern pattern of the infrared marker. The first employee is defined as a sales employee, the second employee as a management employee, the third employee as a settlement employee, and the fourth employee as an administrator, thereby enabling a selective personnel count.

FIG. 3 is an example of an infrared marker pattern applied to the selective visitor counting system according to the present invention. FIG. 4 is a block diagram illustrating the passive infrared marker and detection structure of the selective visitor counting system according to the present invention, Fig.

According to the present invention, the infrared marker detection method is employed in order to selectively count the number of employees.

The infrared marker pattern may have one-dimensional patterns 311 and 313 or two-dimensional patterns 315 and 317 to suit various situations, and various combinations of colors are possible.

Infrared markers according to the combination of patterns and colors not only select employees, but also can determine which employees are employees.

3, when the left one-dimensional pattern 311 is defined as a salesperson and the right first pattern 313 is defined as a management employee, in accordance with the pattern and color classification of the infrared marker, It can be identified and counted whether the person being counted is a sales person or a management person.

Likewise, when the two-dimensional pattern 315 on the left side is defined as a payment staff and the two-dimensional pattern 317 on the right side is defined as an administrator, counting is possible by selecting the counted personnel.

In addition to pattern definition and color definition for infrared markers, you can also adjust the divergence period of infrared LEDs.

Fig. 4 shows such an example. As a method of adjusting the divergence period of the infrared LED in order to enhance the performance distinguishing it from other light bars, Fig. 4A shows a passive infrared marker and a detector, 4 (B) shows an active infrared marker and detector.

4A, infrared rays are emitted from the infrared LED 415 of the detection unit 413 and the infrared marker pattern 417 composed of infrared reflection material is analyzed by the marker detection unit 413 And obtains the marker image 411, and detects the classification of the monitored object.

Each pattern may be made of a reflective material so as to facilitate detection of a pattern to classify the object to be monitored. The pattern may be formed by arranging the rotated pattern in a two-dimensional array for easy detection, or in a barcode format .

The colors of the patterns may be used in various ways so that the markers can be distinguished from each other while the markers are being moved over each other.

In the marker detecting unit 413, a method of diverging an infrared LED and diverging a divergence cycle and an on / off pattern may be used in order to enhance the detection performance of the pattern.

The detecting unit 413 may include an infrared ray passing filter and an infrared ray detecting sensor.

4B, the infrared LED mounted on the marker unit 401 diverges infrared rays using an independent power source 419, and the marker detecting unit 413 detects the marker unit 401, And detects the pattern.

Each of the patterns may be composed of a combination of LEDs 415 of various colors so as to facilitate detection of a pattern for sorting the monitoring object, and may be arranged in a two-dimensional arrangement or in a barcode format so as to increase the detection performance of the pattern .

In addition, it is possible to produce a pattern for classifying the monitored object by diversifying the period of the divergence of the infrared LED and the ON / OFF pattern so that the LED can be distinguished from other LEDs.

FIG. 5 is a flowchart of a method of detecting an infrared marker in the selective visitor counting system according to the present invention.

As shown in FIG. 5, in the infrared marker detection method, an infrared marker image is acquired first (S511). As described above, the method of acquiring the infrared marker image can be received by using the infrared receiver 113 by using various methods such as the pattern of the marker, the color, and the divergence period of the infrared LED.

The acquired infrared marker first binarizes the image to classify the marker (S513).

Image binarization is performed to find a candidate group of the marker in the infrared image acquired using the following Equation (1).

It goes without saying that the threshold value used in the equation (1) can be defined by various methods.

IR (x, y) is the image of the infrared marker acquired by the infrared receiver, IR (x, y) is the output image of the infrared marker output by binarization, And Thresh _bin means a threshold value.

Herein, in the infrared image binarization process (S513), a noise region is removed from the infrared image acquired by the detection unit, and a region having a value less than the threshold value is removed from the infrared image to detect only the infrared marker region.

After the candidate group of the marker is detected through the binarization process of the infrared image, a color camera image matching process is performed to add color information to the candidate region of the marker (S515).

Since the camera 111 and the infrared receiver 113 have different characteristics of acquiring images, two images must be matched to each other in order to use the color information obtained by the camera 111 in an infrared image.

Generally, the process of matching two images can be performed by estimating a homography transformation between two images.

Specifically, it is divided into Extrinsic conversion, which indicates the relationship between the camera 111 and the image of the infrared receiver 113, and Intrinsic conversion, which indicates the relationship between the focal length and distortion of the lens.

Equation (2) represents a process of converting an infrared image to match a camera image.

Here, IR is an infrared image input to the infrared receiver, IR _trans is an image matched with a camera image, A _intrinsic is an Intrinsic transformation, A _extrinsic is an Extrinsic transformation, A _intrinsic A _extrinsic is a homography matrix it means.

In the image matching process with the camera, the process of adding the color information to the infrared marker detected by binarization is performed by matching the marker candidate region detected in the process of binarization of the infrared image with the image of the camera.

Since the infrared image is a monochrome image having one kind of channel, the performance of distinguishing the detected markers when they are overlapped with each other or located in an adjacent area may be deteriorated. Therefore, it is possible to enhance the tracking and classification performance of the marker by adding the color information obtained from the color camera to the region detected as the candidate region of the marker in the infrared image.

In order to add the color information of the camera to the infrared image, a process of matching the color image and the infrared image is required. The matching process includes extrinsic alignment to match the physical positions of the camera 111 and the infrared ray receiver 113, And an intrinsic alignment process for matching the difference between the distortion of the lens used in the infrared receiver 113 and the focal distance.

After matching with the camera image, a process of matching the color information of the candidate region of the marker detected in the infrared image and the pattern of the image with the pattern of the marker to classify the predefined monitored object is performed (S517) (S519).

Pattern matching can be performed in various ways. A function for comparing the similarity between the predefined pattern and the detected marker candidate region can be defined and calculated and displayed in the pattern most similar to the detected region.

In the pattern matching process (S517), the process of checking the correspondence relationship between the pattern and the pattern of the color and shape of the infrared marker unit is performed. The position and classification of the marker can be finally determined in the infrared image using the morphological pattern (color and positional relationship) of the defined pattern and the temporal pattern of on / off.

FIG. 6 is a flow chart of a method of classifying a monitored object in an image analysis module using an infrared marker in a selective visitor coefficient system according to the present invention. FIG. 7 is a flowchart illustrating a method of classifying a monitored object using an infrared marker in the selective visitor coefficient system according to the present invention FIG. 8 is a diagram showing a classification attribute granting and updating sequence of a monitored object in the selective visitor counting system according to the present invention.

The image analysis module 211 of the present invention performs a process of assigning a unique classification to a target to be monitored and tracked by the image analysis technique using the target classification and position detected by the marker detection unit 213, The technology is configured so that the personnel can be counted selectively.

Referring to FIG. 6, the marker detected through the marker detection unit 213 performs validation before combining with the image analysis technique (S611).

In the validity check, whether or not the detected marker is consistently determined to be the same classification for a predetermined time is checked (S611). If the marker and the monitored object correspond to each other at 1: 1 (S613), the classification information of the monitored object is updated (S615).

In the validation of the marker, the marker detecting unit 213 examines the persistence of the detected result, thereby preventing the erroneously detected marker from being utilized.

In the process of searching for 1: 1 correspondence between markers and surveillance objects, the process of searching for 1: 1 correspondence by checking the relevance of the surveillance objects detected by the image analysis based on the positions of the validated markers. In order to check the 1: 1 correspondence relationship, the positional adjacency and time-continuous matching are checked by using the matching information of the infrared image and the camera image which are performed to detect the marker.

In the process of updating the classification information of the object to be monitored, the classification information of the marker having a 1: 1 correspondence relationship with the object to be monitored detected by the image analysis technique may be allocated.

Referring to FIG. 7, the concept of assigning a classification attribute to the monitored object 713 of the image analysis technique using the information of the detected infrared marker 711 is shown.

The object to be monitored detected by the image analysis module 211 is given a classification attribute to the object to be monitored using the information of the marker.

Then, the classification attribute of the monitored object is updated in real time using the information of the marker updated in real time.

The classification attribute is assigned to the employee 713a so as to correspond to the detected infrared marker 711. Since the infrared marker is not attached to the general visitor 713b other than the employee, the classification attribute need not be given.

With reference to FIG. 8, setting and updating of the classification attribute for the monitoring target performed in the image analysis module will be described.

First, the monitored object is tracked in real time (S811).

Through the tracking of the object to be monitored, it is confirmed whether or not a new marker is detected (S813).

If no new marker is detected, the monitoring target is continuously tracked. If a new marker is detected, it is checked whether there is an existing attribute (S815).

If the detected new marker is not included in the existing attribute, the attribute is set as the classification of the new marker, and the monitored object is tracked again after the setting (S817).

If the detected new marker is included in the attribute previously assigned, it is checked whether or not the new marker is the same as the new marker (S819).

If the new marker is included in the same classification, the classification attribute is maintained and the object to be monitored is continuously tracked (S821). If the new marker is not included in the same classification, it is checked whether or not the existing classification is valid, The monitoring object is continuously tracked (S823). If the existing classification is not valid, the attribute is set to the classification of the new marker, and the monitored object is continuously tracked (S825).

The marker detecting unit 213 sets or updates the attribute classification using the information of the infrared marker while classifying the markers in real time. The set or updated classification attribute is used in the process of selecting and managing by classification attribute in the personnel counting process.

9 is a flowchart of a selective personnel counting process using the classification attribute in the selective visitor counting system according to the present invention.

A method of selectively counting the number of persons using the classification attributes of the markers in FIGS. 6 to 8 will be described.

As shown in FIG. 6 to FIG. 8, the classification attribute can be assigned to the object to be monitored in the image analysis technology through the use of the infrared marker.

This classification attribute is used to track the monitored objects and to observe the passage event for the personnel coefficient, and to reflect the coefficients by classification.

The selected personnel counting method can be composed of the personnel counting process by the real time classification and the postprocessing process of the personnel counting.

The real-time person counting process is performed by the image analysis module to update the person count according to the classification attribute each time a pass event is generated while tracking the monitored object in real time.

For example, in the case of a store monitoring application, it means a process of counting general visitors when there is no category to be monitored, and updating the count information of a sales person or manager according to classification in case of a monitoring subject having a marker.

In the process of posting the personnel coefficient, when the tracking of the monitored object ends, the change history of the classification attribute is detected, and the previous count history which is counted as the visitor before the marker is detected is corrected using the final classification attribute .

In other words, in the process of posting the personnel coefficient, the process of correcting the previously updated personnel coefficient information is performed in the case where the monitoring object is changed in classification information of the monitoring object after the personnel coefficient information is updated by the action passing through the specific region . For example, if an employee is passing through a marker screen, the image analysis technique counts as a visitor without a marker. However, if a marker is detected for the same person a few moments later, Lt; / RTI >

In the selective person counting order using the classification attribute, the monitored object is first tracked in real time (S911).

It is checked whether or not a passage event has occurred as the monitored object is traced (S913).

If a pass event has occurred, it is checked whether or not it is a store visitor based on the classification attribute (S915).

In the case of the store visitor, that is, when the infrared marker is not identified, the store visitor is counted and updated, and the history of the visitor coefficient is set and stored in the DB (S917).

If it is not a shop visitor, that is, if there is an infrared marker, the employee count is updated (S919).

The order from S911 to S919 corresponds to the above-mentioned personnel counting process by real-time classification.

Subsequently, the procedure of S921 to S929 represents the post process of the personnel coefficient.

If the tracing event is generated and it is confirmed whether the shop visitor or the employee is successfully traced (S921), the monitoring target tracking (S911) is continuously performed if the monitoring target is successfully traced .

If the tracking of the monitoring target fails, the process of checking the visitor count history is performed (S923).

If the visitor count history does not remain in the visitor count history, it is removed from the monitoring subject (S929). If the visitor count history remains in the visitor count history, it is determined whether the visitor is a visitor according to the classification attribute (S925).

If it is a store visitor, it is removed from the monitoring object (S929). If it is not a store visitor, the visitor coefficient is readjusted (S927) and removed from the monitoring object (S927).

While the present invention has been described with reference to exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, As will be understood by those skilled in the art. Therefore, it should be understood that the above-described embodiments are to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than the foregoing description, It is intended that all changes and modifications derived from the equivalent concept be included within the scope of the present invention.

111: camera 113: infrared receiver
115a: shop visitor 115b: employee
117: Image analysis result 119: Infrared marker detection
121: Division setting 123a, 123b: DB
211: image analysis module 213: marker detection unit
221: object detection unit 223: object classification and object tracking unit
225: object passage detection unit 311, 313: one-dimensional pattern
315, 317: two-dimensional pattern 401: marker portion
411: marker image 413: marker detection section
415: Infrared LED 417: Infrared marker
419: Power supply

Claims (8)

An infrared marker and an image analysis technique for counting an object to be monitored through a camera, acquiring an image of an infrared marker through an infrared receiver, and selectively counting the object to be monitored in accordance with the detection of the infrared marker, Visitor Counting System. The method according to claim 1,
The infrared marker is a one-dimensional pattern or a two-dimensional pattern, and is classified into a monitoring target and an optional visitor counting system using an infrared marker and an image analysis technique. The method according to claim 1,
Wherein the monitoring object is re-selected and counted through a combination of at least one of the pattern of the infrared marker, the color of the infrared marker, and the divergence period of the infrared LED, using the infrared marker and the image analysis technique. The method according to claim 1,
The light source of the infrared marker may be an infrared marker, an infrared marker, an infrared marker, an infrared marker, and an infrared marker. The infrared marker may be an infrared marker, an infrared marker, or an infrared marker. Supervised Classification and Selective Visitor Counting System. The method according to claim 1,
The monitored object is detected, classified and counted through an image analysis module,
Wherein the infrared marker is detected by a marker detection unit and transmitted to the image analysis module so that a camera image and an infrared image are matched with each other and the monitored object is selected and counted. Coefficient system. The method according to claim 1,
Wherein the infrared marker assigns classification attributes such that the infrared marker corresponds to one group of the objects to be monitored in a 1: 1 manner. The method according to claim 1,
An infrared marker and an image analysis technique for updating a classification attribute for the new infrared marker by comparing the attribute with a previously stored infrared marker and a classification standard when a new infrared marker is detected in the course of real- Classification and Selective Visitor Counting System. The method according to claim 1,
The selected visitor counting system includes a real-time classification-based person counting module for counting a real-time person according to a classification attribute,
The target classification and selective visitor counting system using infrared markers and image analysis technology, which are classified into the post - correction of the number of people who correct the number counted by the real - time classification system.

Images