KR102513803B1 - Method for counting object using artificial neural network, and computing apparatus for performing the same - Google Patents

Abstract

An method for counting an object using artificial neural network comprises: a step of extracting, by the computing device, image frames at regular time intervals from a captured image; a step of recognizing, by the computing device, objects for each image frame; a step of detecting, by the computing device, the center point of objects recognized for each image frame; a step of identifying, by the computing device, center points corresponding to the same object among center points detected in different image frames based on a result of the computing device comparing the positions of center points between consecutive image frames; and a step of performing, by the computing device, object counting based on the number of detected center points and a result of identifying center points corresponding to the same object. Accordingly, the present invention can make the system lighter.

Description

Object counting method using artificial neural network and computing device for performing the same

Embodiments disclosed herein relate to a method for counting objects using an artificial neural network, and a computing device for performing the same.

The development of technology has freed humans from simple labor, and this trend has especially accelerated with the recent development of artificial intelligence technology.

For example, the task of counting the number of objects takes a lot of time and can increase the fatigue of the task performer as the size of the objects is small and the number of objects increases. technology is being used.

However, the counting technology using the existing artificial neural network requires a lot of resources due to complex calculations, so it can be executed only on high-performance computers. Therefore, it is necessary to develop a counting technology that can maintain a high level of accuracy while being lighter.

On the other hand, the above-mentioned background art is technical information that the inventor possessed for derivation of the present invention or acquired in the process of derivation of the present invention, and cannot necessarily be said to be known art disclosed to the general public prior to filing the present invention. .

Embodiments disclosed in this specification are intended to provide a lightweight object counting method and apparatus using an artificial neural network that can be executed even in a limited computing environment.

As a technical means for achieving the above-described technical problem, according to an embodiment, an object counting method using an artificial neural network includes the steps of extracting, by a computing device, image frames at predetermined time intervals from a photographed image, the computing device using the image Recognizing objects frame by frame, detecting, by the computing device, the center points of the objects recognized by the image frame, based on a result of comparing the locations of the center points between successive image frames by the computing device, different image frames Identifying center points corresponding to the same object among the center points detected in and performing, by the computing device, object counting based on the number of the detected center points and a result of determining the center points corresponding to the same object. can do.

According to another embodiment, as a computer program for performing an object counting method using an artificial neural network, the object counting method using an artificial neural network includes extracting, by a computing device, image frames at predetermined time intervals from a photographed image, the computing device Recognizing objects for each image frame, detecting, by the computing device, center points of objects recognized for each image frame, based on a result of the computing device comparing locations of center points between successive image frames, Identifying center points corresponding to the same object among center points detected in different image frames, and the computing device performing object counting based on the number of the detected center points and a result of determining the center points corresponding to the same object steps may be included.

According to another embodiment, a computer-readable recording medium on which a program for performing an object counting method using an artificial neural network is recorded, and an object counting method using an artificial neural network, in which a computing device extracts image frames at predetermined time intervals from a photographed image. The computing device recognizing objects for each image frame, the computing device detecting the center points of the recognized objects for each image frame, the computing device comparing the locations of the center points between successive image frames Based on the result, determining center points corresponding to the same object among center points detected in different image frames, and based on the result of the computing device determining the number of the detected center points and the center points corresponding to the same object and performing object counting.

According to another embodiment, a computing device for performing an object counting method using an artificial neural network includes an input/output unit for receiving a captured image of objects and outputting an object counting result, and counting objects from the captured image. and a control unit including a storage unit for storing a program for storing a program and at least one processor, wherein the control unit extracts image frames at predetermined time intervals from the photographed image, recognizes objects for each image frame, and recognizes the image frames for each image frame. Detect the center points of objects recognized by each star, determine the center points corresponding to the same object among the center points detected in different image frames based on the result of comparing the positions of the center points between successive image frames, and determine the center points of the detected center points. Object counting may be performed based on the number of objects and the result of identifying center points corresponding to the same object.

According to any one of the above-described problem solving means, based on the result of comparing the positions of the center points between successive image frames, center points corresponding to the same object are identified among the center points detected in different image frames, and the result is By reflecting and performing object counting, the effect of improving accuracy while efficiently using resources can be expected.

In addition, by setting some areas of the image frame as the detection area and counting area, rather than the entire image frame, and performing object recognition and center point detection only within the corresponding areas, the necessary calculation can be reduced and the effect of lightening the system can be expected. .

Effects obtainable from the disclosed embodiments are not limited to those mentioned above, and other effects not mentioned are clear to those skilled in the art from the description below to which the disclosed embodiments belong. will be understandable.

1 is a diagram illustrating a system for performing an object counting method using an artificial neural network according to an embodiment.
FIG. 2 is a diagram illustrating a detailed configuration of a computing device included in the system shown in FIG. 1 .
3 to 6 are diagrams for explaining an object counting method using an artificial neural network according to embodiments.
7 is a diagram for explaining a method of setting a detection area and a counting area in an image frame, and performing object recognition and object center point detection.
8 is a diagram for explaining in detail a method of counting the number of objects through location comparison of center points between successive image frames.
9 and 10 are diagrams for explaining why the moving distance of an object during two consecutive image frames should be limited to a predetermined reference value or less.

Hereinafter, various embodiments will be described in detail with reference to the accompanying drawings. Embodiments described below may be modified and implemented in various different forms. In order to more clearly describe the characteristics of the embodiments, detailed descriptions of matters widely known to those skilled in the art to which the following embodiments belong are omitted. And, in the drawings, parts irrelevant to the description of the embodiments are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a component is said to be “connected” to another component, this includes not only the case of being “directly connected” but also the case of being “connected with another component intervening therebetween”. In addition, when a certain component "includes" a certain component, this means that other components may be further included without excluding other components unless otherwise specified.

Embodiments described herein relate to a method of counting objects using an artificial neural network.

In the present specification, an embodiment of 'counting the number of eggs (objects) moved through a conveyor belt' is described in detail to enhance understanding of the invention through specific examples, and then to explain expandable embodiments. do it with

1 is a diagram illustrating a system for performing an object counting method using an artificial neural network according to an embodiment. Referring to FIG. 1 , a system for performing an object counting method using an artificial neural network according to an embodiment may include a photographing device 10, a computing device 100, and a conveyor belt 20.

Eggs are placed on the conveyor belt 20, and it is assumed that the conveyor belt 20 moves in a constant direction and at a constant speed.

The photographing device 10 photographs the eggs moving through the conveyor belt 20 and transmits the photographed image to the computing device 100 . To this end, the location of the photographing device 10 may be set to photograph a certain area of the conveyor belt 20 .

The computing device 100 may count the number of eggs moving through the conveyor belt 20 by analyzing the captured image received from the photographing device 10 . The computing device 100 may include an artificial neural network for recognizing an object from an image, and may store a program for counting the number of objects (a program for executing an object counting algorithm) according to the object recognition result. . Object recognition using an artificial neural network and an object counting algorithm will be described in detail below.

FIG. 2 is a diagram illustrating a detailed configuration of a computing device included in the system shown in FIG. 1 . Referring to FIG. 2 , the computing device 100 may include an input/output unit 110 , a control unit 120 and a storage unit 130 .

The input/output unit 110 is a component for transmitting and receiving data with an external device, receiving a command from a user, and outputting a result of performing a task. The computing device 100 receives a captured image from the photographing device 10 through the input/output unit 110, and displays a result of counting the number of objects through the captured image analysis on the screen through the input/output unit 110, or It can be transmitted to other external devices.

The controller 120 is a component including at least one processor such as a CPU and controls the overall operation of the computing device 100 . The controller 120 may perform an object counting method using an artificial neural network by executing a program stored in the storage unit 130 to be described later. A detailed method of performing object counting by the controller 120 using an artificial neural network will be described below with reference to other drawings.

The storage unit 130 may store various types of programs and data. In particular, a program for performing an object counting method using an artificial neural network may be stored in the storage unit 130 .

3 to 6 are diagrams for explaining an object counting method using an artificial neural network according to embodiments. Hereinafter, a method of performing object counting from an image using the systems and apparatuses of FIGS. 1 and 2 will be described in detail.

Referring to FIG. 3 , in step 301 , the controller 120 of the computing device 100 extracts image frames at regular time intervals from a captured image received from the photographing device 10 . For example, if a captured image is 60 fps, and it is desired to extract image frames such that the time interval between image frames is 0.033 seconds (corresponding to 30 fps), the controller 120 may extract the image frames included in the captured image while skipping one by one. can It may be necessary to adjust the time interval between image frames according to the moving speed of objects or the distance between different objects, which will be described in detail below.

In step 302, the controller 120 recognizes objects for each extracted image frame. Although the neural network used for object recognition is not particularly limited, according to an embodiment, the controller 120 may recognize objects using a convolutional neural network (CNN) widely used in the field of image recognition. In order to increase the recognition accuracy, the artificial neural network can be trained in advance through images taken of objects in various states (e.g. normal, defective) under various conditions (e.g. lighting).

The controller 120 may perform object recognition on the entire image frame, but in order to increase efficiency, a part of the image frame may be set as a region of detection and object recognition may be performed only on the detection region. may be The detection area may be set in advance or may be set in real time according to the result of analyzing the image frame. The setting of the detection area will be described in detail below with reference to FIG. 7 .

In step 303, the controller 120 detects the center points of recognized objects for each image frame. The controller 120 may detect the center points of all objects included in the image frame, but in order to increase efficiency, a part of the image frame is set as a region of counting and the center points of objects are detected only within the counting region. You may. The counting area may be set in advance or set in real time, and the setting of the counting area will be described in detail with reference to FIG. 7 below.

Hereinafter, referring to FIG. 7 , a method of setting a detection area and a counting area for each image frame and performing object recognition and object center point detection will be described in detail.

Referring to FIG. 7 , a detecting area 710 and a counting area 720 are set on one image frame 700 . As described above, if object recognition and center point detection are performed on the entire image frame 700, the amount of calculation increases and resources are consumed. and a counting area 720, and object recognition and central point detection can be performed only within the set area.

As the detection area 710 and the counting area 720 are set narrower, the number of objects to be searched by the control unit 120 decreases and efficiency may increase. However, if they are set too narrowly, detection accuracy decreases. Therefore, hereinafter, a method of setting the detection area 710 and the counting area 720 to maintain detection accuracy at a certain level or higher will be described. Hereinafter, it is assumed that the movement direction of the objects is the 'length' direction of the detection area 710 and the counting area 720, and the direction perpendicular thereto is the 'width' direction of the detection area 710 and the counting area 720. do.

First, the detecting area 710 will be described. The controller 120 may be implemented to recognize an object only within the detecting area 710 . In order for all objects to be recognized as complete in at least one image frame, the length (length_ROD) of the detecting area 710 is the maximum value of the object's width (max_R_object) and the object's movement during consecutive image frames. It should be set greater than the sum of the maximum distances (max_d_move). That is, a relationship such as Equation 1 below is established.

Next, the counting area 720 will be described. The controller 120 may be implemented to detect center points of objects only within the counting area 720 . Referring to FIG. 7 , the center points of objects 'where the center points are located' in the counting area 720 are detected. In order for the center point of all objects to be detected in the counting area of at least one image frame, the length (length_ROC) of the counting area 720 is greater than the maximum value (max_d_move) of the distance moved by the object during consecutive image frames. should be set. However, the length (length_ROC) of the counting region 720 should be set smaller than the length (length_ROD) of the detecting region. Therefore, a relationship such as Equation 2 below is established.

In this case, 'distance (d_move) moved by the object during consecutive image frames' is a value obtained by multiplying the moving speed of the object by the time interval between image frames. In the case of counting eggs moving through a conveyor belt as in the embodiment shown in FIG. 7 , there is no need to take the 'maximum value' because the moving speed of all objects is the same. However, in expandable embodiments to be described later, since moving speeds may be different for each object, a 'maximum value' should be taken.

The controller 120 may set the detection area and the counting area in real time. For example, the control unit 120 measures the maximum value of the width of an object (max_R_object) or the maximum value of the distance (max_d_move) of a distance moved by an object during consecutive image frames by analyzing the image frame, and calculates the measurement result using Equation 1 and 2, the length of the detecting area and counting area may be set in real time.

Returning to FIG. 3 , in step 304, the controller 120 determines the center point corresponding to the same object among the center points detected in different image frames based on the result of comparing the locations of the center points between successive image frames. In detail, the controller 120 calculates the separation distances of center points between successive image frames, and if the calculated separation distance is smaller than a preset reference value, it is determined that center points corresponding to the separation distances are center points of the same object. The setting of the reference value will be described in detail below.

Hereinafter, referring to FIG. 8 , a method of identifying a center point corresponding to the same object among center points detected in different image frames by comparing the separation distance between the center points between successive image frames with a reference value will be described in detail.

8 shows counting areas 810, 820, and 830 of three consecutive image frames. For convenience of explanation, the counting area 810 of the first image frame is referred to as a first counting area, the counting area 820 of the second image frame is referred to as a second counting area, and the counting area of the third image frame ( 830) is referred to as a third counting area. In each of the counting areas 810, 820, and 830, there are dots filled with black and dots marked with dotted lines. The dots filled with black are center points detected in the corresponding image frame, and the dots marked with dotted lines are detected in the immediately previous image frame. are the central points. In FIG. 8, objects (eggs) are omitted and only the center points of the objects are shown.

Referring to FIG. 8 , the controller 120 compares the positions of the center points P21 and P23 detected in the second counting area 820 and the center points P11 and P12 detected in the first counting area 810. to calculate the separation distance between the center points. At this time, the controller 120 controls all center points detected in the first counting area 810 of the previous image frame for each of the center points P21 and P23 detected in the second counting area 820 of the current image frame. Calculate the separation distance with the fields (P11, P12). That is, the control unit 120 calculates the separation distances from P11 and P12 for P21, respectively, and calculates the separation distances from P11 and P12 for P23, respectively. The controller 120 determines whether there is a distance smaller than a preset reference value among the calculated distances between center points, and determines that center points corresponding to distances smaller than the reference value are center points of the same object.

It is assumed that the separation distance between P21 and P11 in the second counting area 820 of FIG. 8 is smaller than the reference value, and the separation distances between the other center points are all greater than the reference value. In this case, the controller 120 determines that P21 and P11 are central points corresponding to the same object.

Referring to the third counting area 830 of FIG. 8 , the controller 120 determines the center point P33 detected in the third counting area 830 and the center points P21 and P23 detected in the second counting area 820. ) to calculate the separation distance between the center points. It is assumed that the separation distance between P33 and P23 in the third counting area 830 is smaller than the reference value, and the separation distances between the other center points are all greater than the reference value. In this case, the controller 120 determines that P33 and P23 are central points corresponding to the same object.

The reference value compared with the distance between the center points of successive image frames may be set to an appropriate value according to the shape or characteristics of the object, the level of accuracy required, etc. Specific examples of setting the reference value will be described below. .

According to an embodiment, the reference value may be set to a value obtained by multiplying a minimum distance between center points of different objects (hereinafter referred to as 'minimum distance between objects') by 0.5.

In the case where objects to be counted have almost the same size and shape, such as eggs, and the objects can be placed in contact with each other, the minimum separation distance between objects may be 'the minimum value of the object width (min_R_object)'. Accordingly, in an embodiment of counting eggs moving through a conveyor belt, the reference value may be 'a value obtained by multiplying the minimum value (min_R_object) of the width of an egg (object) by 0.5'. At this time, the minimum value (min_R_object) of the width of the object may be detected by the controller 120 analyzing the image frame.

On the other hand, when the object to be counted is a person (e.g. counting the number of people by analyzing an image taken from above in a space where people exist), not only are the objects slightly different in size and shape, but also there is a general difference between different people. Since the distance is maintained to a certain extent, the controller 120 cannot determine the reference value using the minimum value of the object width as in the case of counting eggs. In this case, when the user inputs a value considered to be the minimum distance between people to the computing device 100 as the minimum distance between objects, the controller 120 may determine the reference value accordingly.

In order to compare the separation distances of center points between successive image frames with a reference value to determine whether they are the same object, the maximum value (max_d_move) of the distance moved by an object during successive image frames must be controlled to be smaller than the reference value. The reason for this will be described with reference to FIGS. 9 and 10 below.

9 and 10 are diagrams for explaining why the maximum value (max_d_move) of a distance moved by an object during two consecutive image frames must be limited to a predetermined reference value or less. 9 and 10 both show a situation in which two eggs (objects) are moved adjacent to each other, but only the moving speed of the eggs is different.

Referring to FIGS. 9 and 10 , the minimum separation distance between objects 910a and 920a is d, which is the distance between P1a and P2a, and this value is the width of the object assuming that the eggs have the same size and shape. Same as the minimum value (min_R_object). Therefore, the reference value may be set to (d * 0.5).

9 illustrates objects 910a and 920a recognized in the nth image frame and objects 910b and 920b recognized in the (n+1)th image frame. As can be seen in the drawing, the distance (d_move) that the object moves during two consecutive image frames is greater than the reference value (d * 0.5). As a result, looking at the distance between the center points, d91, which is the distance between the center points P1a and P1b corresponding to the same object 910a and 910b, is the center point P1b and P1b corresponding to different objects 910b and 920a. P2a) became larger than d92, the separation distance between them. That is, since the separation distance between the center points of different objects may be smaller than the separation distance between the center points of the same object, it is impossible to determine whether the object is the same based on the separation distance between the center points.

10 illustrates objects 1010a and 1020a recognized in the nth image frame and objects 1010b and 1020b recognized in the (n+1)th image frame. As can be seen in the figure, the distance (d_move) that the object moves during two consecutive image frames is smaller than the reference value (d * 0.5). As a result, d101, which is the distance between the center points P1a and P1b corresponding to the same object 1010a and 1010b, is d102, which is the distance between the center points P1b and P2a corresponding to different objects 1010b and 1020a. smaller than As such, since the distance between center points of the same object is always kept smaller than the distance between center points of different objects, the controller 120 can determine whether the object is the same based on the distance between center points.

A method of limiting a maximum value (max_d_move) of a distance moved by an object during two consecutive image frames to a predetermined reference value or less may be implemented in various ways. For example, (max_d_move) can be reduced by lowering the object's movement speed. If the object's movement speed cannot be changed or the minimum separation distance between objects is too small, changing the object's movement speed is not sufficient. 301 (max_d_move) may be reduced by shortening the time interval between image frames extracted in the step.

Returning to FIG. 3 again, in step 305, the controller 120 performs object counting using the number of detected center points and the result of identifying center points corresponding to the same object.

Basically, the controller 120 may count the number of objects appearing in a captured image by summing the number of center points detected in image frames. However, the same object may appear over a plurality of image frames, and therefore, if the number of center points detected in the image frames is simply summed up, some identical objects may be counted redundantly. According to the counting algorithm according to an embodiment, in order to prevent the above problem, the control unit 120 determines the same object among the center points detected in different image frames based on the result of comparing the locations of the center points between successive image frames. The corresponding center point may be identified, and the counting value may be corrected by reflecting the result.

Hereinafter, with reference to FIGS. 4 and 8 , an algorithm for counting objects by detecting a center point within a counting area will be described in detail.

FIG. 4 is a flowchart illustrating detailed steps included in step 305 of FIG. 3 .

In step 401, the controller 120 determines the number of center points detected for each image frame as a preliminary counting value. In FIG. 8 , since the number of center points P11 and P12 detected in the first counting area 810 is two, the preliminary counting value for the first counting area 810 is 2. According to the same method, preliminary counting values of the second counting area 820 and the third counting area 830 become 2 and 1, respectively.

In step 402, the controller 120 determines a counting correction value for each image frame based on a result of identifying center points corresponding to the same object. In detail, the controller 120 increases the counting correction value by -1 whenever it is found that among the center points detected in the current image frame, the object corresponding to the center point already detected in the previous image frame is found. See below for specific examples.

Looking at the second counting area 820 of FIG. 8 , as described above, among the center points P21 and P23 detected in the second counting area 820, P21 is the center point ( Since P11) corresponds to the detected object, the controller 120 sets the counting correction value for the second counting area 820 to -1.

Looking at the third counting area 830 of FIG. 8 , as described above, the center point P33 detected in the third counting area 830 is the center point P23 already detected in the second counting area 820. Since it corresponds to the object, the controller 120 sets the counting correction value for the third counting area 830 to -1.

In step 403, the controller 120 calculates a final counting value by summing the preliminary counting value and the counting correction value for all image frames.

Referring to FIG. 8 , since a counting correction value does not exist for the first counting area 810, the counting value for the first counting area 810 becomes 2, the same as the preliminary counting value. The counting value for the second counting area 820 becomes 1, which is the result of adding the preliminary counting value (2) and the counting correction value (-1). The counting value for the third counting area 830 becomes 0, which is the result of adding the preliminary counting value (1) and the counting correction value (-1). The controller 120 obtains a final counting value of 3 by summing the counting values for the first to third counting areas 810, 820, and 830.

In the embodiment described above, since the eggs are moved through the conveyor belt, the moving direction and moving speed of all objects are the same. However, it is possible to extend the embodiment to a case where the moving directions and moving speeds of the objects are different.

This is because if center points corresponding to the same object are determined in step 304 of FIG. 3 , the controller 120 can measure the moving direction and moving speed of each object based on the result.

As a specific embodiment, the number of vehicles moving in each direction may be counted by photographing vehicles moving in both directions from above and analyzing the captured image. FIG. 5 shows detailed steps of step 305 of FIG. 3 corresponding to this embodiment.

In step 501, the control unit 120 determines the movement direction of each center point based on the result of finding center points corresponding to the same object. In step 502, the controller 120 counts the number of objects for each identified movement direction. In this embodiment, the counting correction value is determined based on the result of identifying the center points corresponding to the same object, and the final counting value is derived by reflecting this, as described above with reference to FIG. 8 .

As another embodiment, the number of people present in a specific space may be determined by counting the number of people entering and exiting through a doorway. FIG. 6 shows detailed steps of step 305 of FIG. 3 corresponding to this embodiment.

In step 601, the control unit 120 determines the movement direction of each center point based on the result of finding center points corresponding to the same object. In step 602, the controller 120 counts the number of objects entering and exiting through the doorway based on the identified movement direction. In this embodiment, the control unit 120 may count the number of people exiting through the doorway and the number of people entering through the doorway, respectively, based on the moving direction of objects around the doorway, and the number of people entering through the doorway. We can calculate the number of people staying in the space by subtracting the number of people leaving through the doorway. In this embodiment, the counting correction value is determined based on the result of identifying the center points corresponding to the same object, and the final counting value is derived by reflecting this, as described above with reference to FIG. 8 .

The term '~unit' used in the above embodiments means software or a hardware component such as a field programmable gate array (FPGA) or ASIC, and '~unit' performs certain roles. However, '~ part' is not limited to software or hardware. '~bu' may be configured to be in an addressable storage medium and may be configured to reproduce one or more processors. Therefore, as an example, '~unit' refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, and procedures. , subroutines, segments of program patent code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables.

Functions provided within components and '~units' may be combined into smaller numbers of components and '~units' or separated from additional components and '~units'.

In addition, components and '~units' may be implemented to play one or more CPUs in a device or a secure multimedia card.

The object counting method using an artificial neural network according to the embodiment described with reference to FIGS. 3 to 6 may be implemented in the form of a computer-readable medium storing instructions and data executable by a computer. In this case, instructions and data may be stored in the form of program codes, and when executed by a processor, a predetermined program module may be generated to perform a predetermined operation. Also, computer-readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. Also, a computer-readable medium may be a computer recording medium, which is a volatile and non-volatile memory implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or other data. It can include both volatile, removable and non-removable media. For example, the computer recording medium may be a magnetic storage medium such as HDD and SSD, an optical recording medium such as CD, DVD, and Blu-ray disc, or a memory included in a server accessible through a network.

In addition, the object counting method using an artificial neural network according to the embodiment described with reference to FIGS. 3 to 6 may be implemented as a computer program (or computer program product) including instructions executable by a computer. A computer program includes programmable machine instructions processed by a processor and may be implemented in a high-level programming language, object-oriented programming language, assembly language, or machine language. . Also, the computer program may be recorded on a tangible computer-readable recording medium (eg, a memory, a hard disk, a magnetic/optical medium, or a solid-state drive (SSD)).

Therefore, the object counting method using the artificial neural network according to the embodiment described with reference to FIGS. 3 to 6 can be implemented by executing the computer program as described above by a computing device. A computing device may include at least some of a processor, a memory, a storage device, a high-speed interface connected to the memory and a high-speed expansion port, and a low-speed interface connected to a low-speed bus and a storage device. Each of these components are connected to each other using various buses and may be mounted on a common motherboard or mounted in any other suitable manner.

Here, the processor may process commands within the computing device, for example, to display graphic information for providing a GUI (Graphic User Interface) on an external input/output device, such as a display connected to a high-speed interface. Examples include instructions stored in memory or storage devices. As another example, multiple processors and/or multiple buses may be used along with multiple memories and memory types as appropriate. Also, the processor may be implemented as a chipset comprising chips including a plurality of independent analog and/or digital processors.

Memory also stores information within the computing device. In one example, the memory may consist of a volatile memory unit or a collection thereof. As another example, the memory may be composed of a non-volatile memory unit or a collection thereof. Memory may also be another form of computer readable medium, such as, for example, a magnetic or optical disk.

Also, the storage device may provide a large amount of storage space to the computing device. A storage device may be a computer-readable medium or a component that includes such a medium, and may include, for example, devices in a storage area network (SAN) or other components, such as a floppy disk device, a hard disk device, an optical disk device, or a tape device, flash memory, or other semiconductor memory device or device array of the like.

The above-described embodiments are for illustrative purposes, and those skilled in the art to which the above-described embodiments belong can easily transform into other specific forms without changing the technical spirit or essential features of the above-described embodiments. You will understand. Therefore, it should be understood that the above-described embodiments are illustrative in all respects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope to be protected through this specification is indicated by the following claims rather than the detailed description above, and should be construed to include all changes or modifications derived from the meaning and scope of the claims and equivalent concepts thereof. .

10: photographing device 100: computing device
110: input/output unit 120: control unit
130: storage unit

Claims (20)

In the object counting method using an artificial neural network,
extracting, by a computing device, image frames at predetermined time intervals from a photographed image;
recognizing objects by the image frame by the computing device;
detecting, by the computing device, center points of objects recognized for each image frame;
determining, by the computing device, center points corresponding to the same object among center points detected in different image frames, based on a result of comparing locations of the center points between successive image frames; and
Performing, by the computing device, object counting based on the number of the detected center points and a result of identifying center points corresponding to the same object;
Recognizing the objects may include setting some areas of the image frames as a detecting area; And recognizing the objects only within the detecting area,
The length of the detection area is set to be greater than the sum of the maximum value of the widths of the objects and the maximum value of the distance moved by the object during consecutive image frames,
The detecting of the center points of the objects includes setting a partial area of the image frames as a counting area, and detecting the center points of the objects only within the counting area,
The length of the counting area is set to be larger than the maximum value of the distance moved by the object during consecutive image frames and smaller than the length of the detecting area. According to claim 1,
The step of identifying the center points corresponding to the same object,
Calculating a separation distance between center points between the consecutive image frames;
and determining that center points corresponding to the separation distance are center points of the same object if the calculated separation distance is smaller than a preset reference value. According to claim 2,
The reference value is set to a value obtained by multiplying 0.5 by a minimum value of a separation distance between center points of different objects within the same image frame. According to claim 2,
The reference value is set to a value obtained by multiplying the minimum value of the width of the objects by 0.5. delete delete According to claim 1,
The step of performing the object counting,
determining the number of center points detected for each image frame as a preliminary counting value;
determining a counting correction value for each image frame based on a result of identifying center points corresponding to the same object; and
and calculating a final counting value by summing the preliminary counting value and the counting correction value for all image frames. According to claim 1,
The step of performing the object counting,
determining the movement direction of each center point based on a result of identifying center points corresponding to the same object; and
and counting the number of objects for each identified movement direction. According to claim 1,
The step of performing the object counting,
determining the movement direction of each center point based on a result of identifying center points corresponding to the same object; and
and counting the number of objects entering and exiting the entrance based on the identified movement direction. A computer-readable recording medium on which a program for executing the method according to claim 1 is recorded on a computer. A computer program executed by a computing device and stored in a medium for performing the method according to claim 1 . A computing device for performing an object counting method using an artificial neural network,
an input/output unit for receiving a photographed image of objects and outputting an object counting result;
a storage unit storing a program for counting objects from the photographed image; and
A control unit including at least one processor,
The control unit,
Based on a result of extracting image frames at a predetermined time interval from the captured image, recognizing objects for each image frame, detecting the center points of the recognized objects for each image frame, and comparing the locations of the center points between consecutive image frames. So, among the center points detected in different image frames, center points corresponding to the same object are identified, and object counting is performed based on the number of the detected center points and the result of determining the center points corresponding to the same object,
In recognizing the objects, the control unit sets a partial area of the image frames as a detection area, recognizes the objects only within the detection area, and the length of the detection area is the maximum value of the width of the objects. It is set greater than the sum of the maximum values of the distances the object has moved during continuous image frames,
In detecting the center points of the objects, the control unit sets a partial area of the image frames as a counting area, detects the center points of the objects only within the counting area, and the length of the counting area is determined during consecutive image frames. The computing device, which is set to be greater than the maximum value of the distance moved by the object and smaller than the length of the detecting area. According to claim 12,
In determining the center points corresponding to the same object, the controller
Calculating the separation distance of the center points between the consecutive image frames, and determining that the center points corresponding to the separation distance are center points of the same object if the calculated separation distance is smaller than a preset reference value. . According to claim 13,
The reference value is set to a value obtained by multiplying 0.5 by a minimum value of a separation distance between center points of different objects within the same image frame. According to claim 13,
The reference value is set to a value obtained by multiplying a minimum value of the width of the objects by 0.5. delete delete According to claim 12,
In performing the object counting, the controller
The number of center points detected for each image frame is determined as a preliminary counting value, a counting correction value is determined for each image frame based on a result of identifying center points corresponding to the same object, and the preliminary counting Computing device, characterized in that for calculating a final counting value by summing the value and the counting correction value. According to claim 12,
In performing the object counting, the controller
Based on a result of identifying the center points corresponding to the same object, a moving direction of each of the center points is determined, and the number of objects is counted for each of the determined moving directions. According to claim 12,
In performing the object counting, the controller
Based on a result of identifying the center points corresponding to the same object, a movement direction of each center point is identified, and the number of objects that have entered and exited the entrance is counted based on the identified movement direction.

Images