KR102580715B1 - Apparatus and method for counting number of visitors based on deep learning model - Google Patents

Abstract

A method for counting the number of visitors based on an artificial neural network includes the steps of sequentially inputting a plurality of frames of images captured by a detection unit into a learning model, and the learning model applies learned weights between a plurality of layers to the plurality of frames. Calculating an output value including the center coordinates, width and height of an area box including an object, reliability of the area box, and probability that the object in the area box is a human object through a plurality of operations, wherein the detection unit selecting an area box in which the reliability is greater than a predetermined value and the probability of being the human object is more than a predetermined value from the plurality of frames, and detecting one or more human objects included in the plurality of frames through the area boxes; a count unit setting a counter line dividing an entrance area and an exit area in an image including the plurality of frames, wherein the count unit is positioned at a position corresponding to the counter line of a human object detected through the area box in the plurality of frames; It includes counting the number of visitors according to.

Description

Apparatus and method for counting number of visitors based on deep learning model}

The present invention relates to a technology for counting the number of visitors, and more specifically, to a device and method for counting the number of visitors based on a learning model.

There are various exhibition halls and theme parks in Korea, and analysis of visitors' level of interest and interest in the actual exhibits and contents is conducted through surveys.

Korean Patent Publication No. 2015-0018121 (published on February 23, 2015)

The purpose of the present invention is to provide a device and method for counting the number of visitors based on a learning model.

A method for counting the number of visitors based on an artificial neural network according to a preferred embodiment of the present invention to achieve the above-described purpose includes sequentially inputting a plurality of frames of images captured by a detector into a learning model, and the learning step. The model calculates the center coordinates, width and height of an area box containing an object, the reliability of the area box, and the object within the area box through a plurality of operations in which weights learned between a plurality of layers are applied to the plurality of frames. calculating an output value including a probability of being a human object, wherein the detection unit selects an area box in which the reliability is greater than a predetermined value and the probability of being a human object is greater than a predetermined value from the plurality of frames and includes the region box in the plurality of frames. Detecting one or more human objects through an area box, a count unit setting a counter line dividing an entrance area and an exit area in an image including the plurality of frames, and the count unit setting a counter line in the image including the plurality of frames. and counting the number of visitors according to the position of the human object detected through the area box corresponding to the counter line.

The step of counting the number of visitors is to determine the number of visitors according to the area where the first detected human object is located among the entrance area and exit area divided by the counter line (CL) in the previous frame that the counter is relatively preceding in time. assigning an identifier and an initial flag value, wherein the count unit is determined according to an area where the human object to which the identifier is assigned is located among the entrance area and the exit area divided by the counter line in a subsequent frame that is relatively subsequent in time; assigning a change flag value to the corresponding human object, wherein the counting unit compares the initial flag value and the change flag value and counting the number of human objects that passed the counter line from the exit area to the entrance area as the number of visitors. Includes.

In the step of counting the number of visitors, the counting unit compares the coordinates of the center of the area box with the coordinates of two or more points on the counter line to derive the area where the human object is located.

In order to achieve the above-described object, a device for counting the number of visitors based on an artificial neural network according to a preferred embodiment of the present invention is a device in which learned weights between a plurality of layers are applied to a plurality of frames of an image in which a learning model is captured. By calculating an output value including the center coordinates, width and height of an area box containing an object, the reliability of the area box, and the probability that the object in the area box is a human object through a plurality of operations, the a detection unit that selects an area box whose reliability is a predetermined value or more and a probability of being the human object is a predetermined value or more and detects one or more human objects included in the plurality of frames through the area box, and an image including the plurality of frames A counter line is set to separate the entrance area and the exit area, and a count unit is included to count the number of visitors according to the position of the human object detected through the area box in the plurality of frames corresponding to the counter line.

The count unit assigns an identifier and an initial flag value to the corresponding human object according to the area where the first detected human object is located among the entrance area and exit area divided according to the counter line (CL) in the preceding frame that is relatively preceding in time. , in a subsequent frame that is relatively subsequent in time, a variable flag value is assigned to the corresponding human object according to the area in which the human object to which the identifier is assigned is located among the entrance area and the exit area divided according to the counter line, and the count Additionally, the initial flag value is compared with the change flag value, and the number of human objects passing the counter line from the exit area to the entrance area is counted as the number of visitors.

The count unit compares the coordinates of the center of the area box with the coordinates of two or more points on the counter line to derive the area where the human object is located.

The present invention designates exhibitions, contents, etc. that are subject to investigation through a counting line that can specify the entrance and exit direction within the video, and presents quantitative data on the movement path of visitors and the number of visitors. It is possible to understand the space, and through this, data can be provided on which exhibits should be replaced when exhibits need to be replaced.

1 is a diagram illustrating the configuration of a system for counting the number of visitors based on a learning model according to an embodiment of the present invention.
Figure 2 is a diagram for explaining the configuration of an edge device for counting the number of visitors based on a learning model according to an embodiment of the present invention.
Figure 3 is a diagram for explaining the detailed configuration of an edge device for counting the number of visitors based on a learning model according to an embodiment of the present invention.
Figure 4 is a diagram for explaining a learning model for counting the number of visitors according to an embodiment of the present invention.
Figure 5 is an example screen to explain how a learning model detects a human object according to an embodiment of the present invention.
Figure 6 is a flowchart illustrating a learning method for a learning model according to an embodiment of the present invention.
Figure 7 is a flow chart to explain a method for counting the number of visitors based on a learning model.
Figures 8 and 9 are example screens to explain a method for counting the number of visitors based on a learning model.
Figure 10 is a diagram showing a computing device according to an embodiment of the present invention.

Prior to the detailed description of the present invention, the terms and words used in the specification and claims described below should not be construed as limited to their ordinary or dictionary meanings, and the inventor should use his/her invention in the best possible manner. In order to explain, it must be interpreted as meaning and concept consistent with the technical idea of the present invention based on the principle that the term can be appropriately defined as a concept. Therefore, the embodiments described in this specification and the configurations shown in the drawings are only the most preferred embodiments of the present invention, and do not represent the entire technical idea of the present invention, and therefore, various equivalents that can replace them at the time of filing the present application may be used. It should be understood that there may be variations and examples.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. At this time, it should be noted that in the attached drawings, identical components are indicated by identical symbols whenever possible. Additionally, detailed descriptions of well-known functions and configurations that may obscure the gist of the present invention will be omitted. For the same reason, in the accompanying drawings, some components are exaggerated, omitted, or schematically shown, and the size of each component does not entirely reflect the actual size.

First, the configuration of a system for counting the number of visitors based on a learning model according to an embodiment of the present invention will be described. 1 is a diagram illustrating the configuration of a system for counting the number of visitors based on a learning model according to an embodiment of the present invention. Referring to FIG. 1, a system for counting the number of visitors based on a learning model (hereinafter abbreviated as 'counting system') includes a plurality of edge devices 10 and a management server 20.

A plurality of edge devices 10 are installed in different locations, count the number of visitors based on a learning model at the installed location, and transmit the counting result to the management server 20. Then, the management server 20 can provide statistical data on the number of visitors based on the counting results.

Next, an edge device for counting the number of visitors based on a learning model according to an embodiment of the present invention will be described in more detail. Figure 2 is a diagram for explaining the configuration of an edge device for counting the number of visitors based on a learning model according to an embodiment of the present invention. Figure 3 is a diagram for explaining the detailed configuration of an edge device for counting the number of visitors based on a learning model according to an embodiment of the present invention. Figure 4 is a diagram for explaining a learning model for counting the number of visitors according to an embodiment of the present invention. Figure 5 is an example screen to explain how a learning model detects a human object according to an embodiment of the present invention.

Referring to FIG. 2, the edge device 10 includes a communication unit 11, a camera unit 12, a storage unit 13, and a control unit 14.

The communication unit 11 is for communication with the management server 20 and other devices, for example, a remote device (not shown) connected to the edge device 10. The communication unit 11 may include an RF (Radio Frequency) transmitter (Tx) that up-converts and amplifies the frequency of a transmitted signal and an RF receiver (Rx) that amplifies the received signal with low noise and down-converts the frequency. And the communication unit 11 may include a modem that modulates the transmitted signal and demodulates the received signal. The communication unit 11 may receive an input for setting a counting line according to an embodiment of the present invention from the management server 20 or a remote device (not shown). The communication unit 11 can transmit the counting result according to an embodiment of the present invention to the management server 20 under the control of the control unit 14.

The camera unit 12 is used to capture images including a plurality of frames, that is, moving images. The camera unit 12 may include a lens and an image sensor. Each image sensor receives light reflected from the subject and converts it into an electrical signal. Image sensors can be implemented based on Charged Coupled Device (CCD), Complementary Metal-Oxide Semiconductor (CMOS), etc. In addition, the camera unit 12 may further include one or more analog to digital converters, and may convert the electrical signal output from the image sensor into a digital sequence and output it to the control unit 14.

The storage unit 13 serves to store programs and data necessary for the operation of the user device 10. In particular, the storage unit 13 can store temporary data and counting result data necessary for counting human objects according to an embodiment of the present invention. Additionally, various types of data stored in the storage unit 16 can be deleted, changed, or added according to the operation of the user of the user device 10.

The control unit 14 controls the overall operation of the user device 10 and signal flow between internal blocks of the user device 10, and may perform a data processing function to process data. Additionally, the control unit 14 basically controls various functions of the user device 10. The control unit 14 may include, for example, a Central Processing Unit (CPU), a baseband processor (BP), an application processor (AP), a Graphic Processing Unit (GPU), and a Digital Signal Processor (DSP). Referring to FIG. 3, the control unit 14 includes a learning unit 100, a detection unit 200, and a counting unit 300. The operation of the control unit 14 including the learning unit 100, the detection unit 200, and the counting unit 300 will be described in more detail below.

As shown in Figure 4, a learning model (LM: machine learning model or deep learning model) includes a plurality of layers, and each of the plurality of layers performs a plurality of operations. Each calculation result of a plurality of calculation modules in one layer is weighted and transmitted to the next layer. This means that weights are applied to the calculation results of the current layer and input into the calculations of the next layer. In other words, the learning model (LM) performs multiple operations to which weights of multiple layers are applied. The plurality of layers includes a convolution layer (CVL) that performs a convolution operation, a pooling layer (PLL: Pooling Layer) that performs a down sampling operation or an up sampling operation, It includes a fully connected layer (FCL) that performs calculations using activation functions. Convolution, downsampling, and upsampling operations each use a kernel composed of a predetermined matrix, and the values of the elements of the matrix forming this kernel become weights (w). Here, examples of activation functions include sigmoid, hyperbolic tangent (tanh), Exponential Linear Unit (ELU), Rectified Linear Unit (ReLU), Leakly ReLU, Maxout, Minout, Softmax, etc. . This learning model (LM) may be an example of a model to which algorithms such as YOLO (You Only Look Once), YOLOv2, YOLO9000, and YOLOv3 are applied. When a frame of an image is input, the learning model (LM) performs a plurality of operations to which weights of a plurality of layers are applied, divides the frame into a plurality of cells as shown in Figure 5, and then assigns a center point to each of the plurality of cells. One or more bounding boxes (BB) with coordinates (x, y), width (w), and height (h). An object is contained within the bounding box (BB), and the area of the bounding box (BB) is a human object. Confidence, which represents the probability of matching the area occupied by (obj), and the probability that the object in the area box (BB) is a human object (obj) (human=0.9161) can be calculated and output as the output value (out).

The present invention counts the number of visitors based on a learning model. For this purpose, learning about the learning model is required. The learning unit 100 is for learning a learning model (LM). The learning unit 100 trains the learning model LM to detect and specify the human object obj as shown in FIG. 5 through the area box BB. We will explain these methods. Figure 6 is a flowchart for explaining a learning method for a learning model according to an embodiment of the present invention.

Referring to FIG. 6, the learning unit 100 prepares learning data in step S110. Learning data includes a plurality of learning frames. The learning frame includes a ground-truth box that specifies the area occupied by the human object (obj). Once such training data is prepared, the learning unit 100 inputs a learning frame to the learning model (LM) initialized in step S120. Then, the learning model (LM) will calculate and output an output value through a plurality of operations to which weights of a plurality of layers are applied in step S130. The output value is a plurality of area boxes (BB) with center coordinates (x, y), width (w), and height (h) in each of the plurality of cells of the image, and the area of the area box (B) represents the human object (obj). Confidence (0~1) indicating the degree to which it matches the ground-truth box containing 100%, and the probability that the object in the area box (BB) is a human object (obj) (e.g., human = 0, 8711).

Based on the output value of the learning model (LM), the learning unit 100 may derive a loss value according to the loss function in step S140. For example, the loss function is expressed as Equation 1 below.

는 영역상자(BB)의 변수에 대한 값을 더 반영하기 위한 것으로, 영역상자의 중심 좌표, 폭 및 높이(x, y, w, h)에 대한 손실과 다른 손실들과의 균형을 위한 파라미터이다.

는 영역상자(BB)의 변수에 대한 값을 더 반영하고, 물체가 없는 영역에 대한 값을 덜 반영하기 위한 것이다. 즉,

는 객체가 있는 영역상자(BB)와 객체가 없는 영역상자(BB) 간의 균형을 위한 파라미터이다. 여기서,

=5이고,

=0.5가 될 수 있다.

는 셀 i에 객체가 있는 경우 1이고, 없는 경우 0을 나타낸다.

는 셀 i에 있는 영역상자 j에 객체가 있으면 1이고, 없으면 0을 나타낸다.

는 셀 i에 있는 영역상자 j에 객체가 없으면 1이고, 있으면 0을 나타낸다.S represents the number of cells. 5 shows 12 cells. C represents the trust score. B represents the number of area boxes (BB) in one cell. pi(c) represents the probability that the object in the i-th cell is the corresponding class (c), and in the present invention, only the probability of being a single class, that is, a human object, is used. Here, i is a parameter indicating the cell in which the object exists, and j is a parameter indicating the predicted area box (BB). Additionally, x and y represent the center coordinates of the area box (BB), and w and h represent the width and height of the area box, respectively.

is intended to further reflect the values of the variables of the area box (BB), and is a parameter for balancing the loss for the center coordinate, width, and height (x, y, w, h) of the area box with other losses. .

is intended to reflect more the values of the variables of the area box (BB) and less reflect the values of the area without objects. in other words,

is a parameter for balance between the area box (BB) with objects and the area box (BB) without objects. here,

=5,

=0.5.

represents 1 if there is an object in cell i, and 0 if there is no object.

represents 1 if there is an object in area box j in cell i, and 0 otherwise.

represents 1 if there is no object in area box j in cell i, and 0 if there is an object.

Looking at the loss function in Equation 1, the first and second terms of Equation 1 are as shown in Equation 2 below.

Equation 2 is used to calculate a coordinate loss that represents the difference between the center coordinates, width, and height (x, y, w, h) of the area box and the area occupied by the human object (obj).

Additionally, the third and fourth terms of Equation 1 are as shown in Equation 3 below.

Equation 4 is used to calculate the confidence loss that represents the difference between the area of the area box (BB) and the ground-truth box containing 100% of the area occupied by the human object (obj).

Finally, the last term of Equation 1 is equivalent to Equation 4 below.

Equation 4 is used to calculate a classification loss that represents the difference between the class of the object in the area box (BB), that is, the class of the human object (obj) and the actual object.

In this way, after calculating the loss values, that is, coordinate loss, reliability loss, and classification loss through the loss function in step S140, the learning unit 100 learns to minimize the coordinate loss, reliability loss, and classification loss in step S150. Optimize the weights of the model (LM).

Next, the learning unit 100 determines whether the learning completion condition is satisfied in step S160. The learning unit 100 performs calculations on the entire learning model (LM) through a learning data set for evaluation among a plurality of learning data, and then, if the loss value calculated according to the output value of the learning model (LM) is within a predetermined value, , it can be judged that the learning completion conditions are satisfied.

As a result of the determination in step S160, if the learning completion condition is not satisfied, the learning unit 100 repeats steps S120 and S160 described above. On the other hand, if the learning completion condition is satisfied as a result of the determination in step S160, learning is terminated in step S170. As a result, a learning model (LM) with learned parameters, that is, weights, is completed.

Next, a method for counting the number of visitors will be described based on a learning model that has been trained according to the method described above. Figure 7 is a flow chart to explain a method for counting the number of visitors based on a learning model. Figures 8 and 9 are example screens to explain a method for counting the number of visitors based on a learning model.

Referring to FIG. 7, the detection unit 200 of the edge device 10 captures an image through the camera unit 12 in step S210, and sequentially inputs a plurality of frames of the captured image into the learning model (LM). .

Then, the learning model (LM) calculates an output value through a plurality of operations in which the learned weights between the plurality of layers are applied to the plurality of frames input in step S220. These output values include the center coordinates (x, y), width (w) and height (h) of the area box (BB) containing the object, the reliability of the area box (BB), and whether the object in the area box (B) is a human object ( obj).

In step S230, the detection unit 220 selects an area box (BB) in which the reliability is higher than a predetermined value and the probability of being a human object is higher than a predetermined value in the plurality of frames, and selects one or more human objects (obj) included in the plurality of frames into the area. Detected through box (BB).

In step S240, the count unit 300 sets a counter line CL that separates the entrance area and the exit area in the image including a plurality of frames. For example, examples of frame screens according to embodiments of the present invention are shown in FIGS. 8 and 9 . The frame in FIG. 8 is a preceding frame in time that precedes the frame in FIG. 9. In other words, the frame in Figure 9 is a subsequent frame in time to the frame in Figure 8. As shown, the counter line CL is set at the same position in a plurality of frames included in the image. Additionally, as shown, the counter line CL divides the image into two areas. At this time, the left area of the counter line CL can be set as the entrance area and the right area can be set as the exit area, or vice versa. The setting of the counter line (CL) dividing the entrance area and the exit area can be set according to input by the user. For example, the user of the management server 20 can transmit an input for setting the counter line CL to the edge device 10. Alternatively, the user can transmit an input for setting the counter line (CL) to the edge device 10 through a remote device (not shown). Upon receiving this input, the counter 300 may set a counter line CL that separates the entrance area and the exit area in the image including a plurality of frames.

As described above, when the counter line (CL) is set, the count unit 300 generates a counter line (CL) corresponding to the counter line (CL) of the human object (obj) detected through the area box (BB) in a plurality of frames in step S250. Count the number of visitors based on location. This S250 step will be described in more detail as follows.

First, the count unit 300 assigns an identifier and an initial flag to the human object according to the area where the first detected human object is located among the entrance area and exit area divided by the counter line (CL) in the preceding frame that is relatively preceding in time. Give a value. For example, assume that the left side of the counter line (CL) in the video is the entrance to the exhibition hall, and the right side is the exit. Accordingly, in FIGS. 8 and 9, it is assumed that the left side of the counter line CL is set as the entrance area (A) and the right side of the counter line CL is set as the exit area (B). Additionally, as shown, it is assumed that four human objects are detected through four area boxes (BB1, BB2, BB3, and BB4) in the preceding frame of FIG. 8. In this case, the count unit 300 assigns identifiers (ID1, ID2, ID3, ID4) to the four human objects. In addition, the count unit 300 counts the first and second human objects ( An initial flag value of '-1' is given to ID1, ID2), and the third and fourth human objects (ID3, ID3, An initial flag value of '1' is given to ID4). Here, the change frames of the first to fourth human objects (ID1, ID2, ID3, and ID4) are all set to '0'.

Next, the count unit 300 flags a change flag to the corresponding human object according to the area in which the identifier-assigned human object is located among the entrance and exit areas divided by the counter line (CL) in a subsequent frame that is relatively subsequent in time. Give a value. As described above, in FIG. 9, the left side of the counter line CL is set as the entrance area (A), and the right side of the counter line CL is set as the exit area (B). At this time, as shown in the following frame in FIG. 9, it is assumed that the human object (ID1, ID2, ID3, ID4) assigned an identifier in the preceding frame is detected through four area boxes (BB1, BB2, BB3, BB4). do. In this case, the count unit 300 counts the first, third, and fourth area boxes BB1, BB3, and BB4 in the entrance area A according to the area where the detected human object is located. A variable flag value of '-1' is given to the 3rd and 4th human objects (ID1, ID3, ID4), and the second human object (ID2) first detected through the second area box (BB2) in the exit area (B) ) is given a change flag value of '1'.

Accordingly, the counting unit 300 compares the initial flag value and the change flag value and counts the number of human objects that passed the counter line CL from the exit area (B) to the entrance area (A) as the number of visitors. For example, in the case of the first human object (ID1), comparing the initial flag value and the change flag value, it is not counted as the number of visitors because it does not pass through the counter line (CL) and continues to stay in the entrance area (A). . Additionally, in the case of the second human object (ID2), when comparing the initial flag value and the change flag value, it is not counted as the number of visitors because it passes the counter line (CL) from the entrance area (A) to the exit area (B). And in the case of the third and fourth human objects (ID3, ID4), comparing the initial flag value and the change flag value, it is found that they pass from the exit area (B) to the entrance area (A) without passing the counter line (CL). Therefore, it can be counted as the number of visitors.

Meanwhile, when deriving the area where the human object is located in step S250, the center coordinates of the area box (BB) that specifies the human object, for example, (x1, y1), (x2, y2), (x3, y3) , can be derived by comparing the coordinates of (x4, y4) and two or more points (a, b, c) of the counter line (CL).

After counting the number of visitors, the counting unit 300 sends the number of visitors counted together with data such as the identifier, initial flag value, and change flag value as described above to the management server 20 through the communication unit 11 in step S260. can be transmitted. Accordingly, the management server 20 can present the visitor's movement path and number of visitors as quantitative data. Accordingly, it is possible to identify exhibits or spaces in which visitors' interest level is low, and through this, data can be provided on which exhibits should be replaced when replacement is necessary.

Figure 10 is a diagram showing a computing device according to an embodiment of the present invention. The computing device TN100 of FIG. 10 may be a device described in this specification, such as an edge device 10 or a management server 20.

In the embodiment of FIG. 10, the computing device TN100 may include at least one processor TN110, a transceiver device TN120, and a memory TN130. Additionally, the computing device TN100 may further include a storage device TN140, an input interface device TN150, an output interface device TN160, etc. Components included in the computing device TN100 may be connected by a bus TN170 and communicate with each other.

The processor TN110 may execute a program command stored in at least one of the memory TN130 and the storage device TN140. The processor TN110 may refer to a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed. Processor TN110 may be configured to implement procedures, functions, and methods described in connection with embodiments of the present invention. The processor TN110 may control each component of the computing device TN100.

Each of the memory TN130 and the storage device TN140 can store various information related to the operation of the processor TN110. Each of the memory TN130 and the storage device TN140 may be comprised of at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory TN130 may be comprised of at least one of read only memory (ROM) and random access memory (RAM).

The transceiving device TN120 can transmit or receive wired signals or wireless signals. The transmitting and receiving device (TN120) can be connected to a network and perform communication.

Meanwhile, the method according to the embodiment of the present invention described above can be implemented in the form of a program readable through various computer means and recorded on a computer-readable recording medium. Here, the recording medium may include program instructions, data files, data structures, etc., singly or in combination. Program instructions recorded on the recording medium may be those specifically designed and constructed for the present invention, or may be known and available to those skilled in the art of computer software. For example, recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. magneto-optical media), and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions may include machine language wires, such as those produced by a compiler, as well as high-level language wires that can be executed by a computer using an interpreter, etc. These hardware devices may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

Although the present invention has been described above using several preferred examples, these examples are illustrative and not limiting. As such, those of ordinary skill in the technical field to which the present invention pertains will understand that various changes and modifications can be made according to the theory of equivalents without departing from the spirit of the present invention and the scope of rights set forth in the appended claims.

10: Edge device 11: Communication unit
12: camera unit 13: storage unit
14: Control unit 20: Management server
100: learning unit 200: detection unit
300: Counter part

Claims (6)

In a method for counting the number of visitors based on an artificial neural network,
A step of sequentially inputting a plurality of frames of the image captured by the detection unit into a learning model;
The learning model calculates the center coordinates, width and height of an area box containing an object, the reliability of the area box, and the calculating an output value including a probability that the object is a human object;
wherein the detector selects an area box in which the reliability is higher than a predetermined value and the probability of being the human object is higher than a predetermined value, and detecting one or more human objects included in the plurality of frames through the area box;
A count unit setting a counter line dividing an entrance area and an exit area in an image including the plurality of frames;
the counting unit counting the number of visitors according to a position corresponding to the counter line of a human object detected through the area box in the plurality of frames;
Includes,
The step of counting the number of visitors is
The counter assigns an identifier and an initial flag value to the corresponding human object according to the area where the first detected human object is located among the entrance and exit areas divided according to the counter line (CL) in the previous frame that is relatively preceding in time. step;
The counter assigning a change flag value to the corresponding human object according to the area in which the human object to which the identifier is assigned is located among the entrance area and the exit area divided by the counter line in a subsequent frame that is relatively subsequent in time. ;
the counting unit comparing the initial flag value and the change flag value and counting the number of human objects passing the counter line from the exit area to the entrance area as the number of visitors;
Characterized by including
A method for counting the number of visitors. delete According to paragraph 1,
In the step of counting the number of visitors
Characterized in that the counting unit compares the coordinates of the center of the area box and the coordinates of two or more points of the counter line to derive the area where the human object is located.
A method for counting the number of visitors. In a device for counting the number of visitors based on an artificial neural network,
The center coordinates, width and height of the area box containing the object, the reliability of the area box, and the area box are obtained through a plurality of operations in which the learned weights between the plurality of layers are applied to the plurality of frames of the image in which the learning model is captured. When calculating an output value including the probability that the object in the object is a human object,
a detection unit that selects an area box in which the reliability is higher than a predetermined value and the probability of being the human object is higher than a predetermined value from the plurality of frames, and detects one or more human objects included in the plurality of frames through the area box; and
In the image including the plurality of frames, a counter line is set to separate the entrance area and the exit area, and the number of visitors is calculated according to the position of the human object detected through the area box in the plurality of frames corresponding to the counter line. a counting unit that counts;
Includes,
The count unit
An identifier and an initial flag value are assigned to the human object according to the area where the first detected human object is located among the entrance and exit areas divided by the counter line (CL) in the previous frame that is relatively preceding in time,
In a subsequent frame that is relatively subsequent in time, a variable flag value is assigned to the corresponding human object according to the area in which the human object to which the identifier is assigned is located among the entrance area and the exit area divided by the counter line,
Characterized in that the counting unit compares the initial flag value and the change flag value and counts the number of human objects that passed the counter line from the exit area to the entrance area as the number of visitors.
A device for counting the number of visitors. delete According to paragraph 4,
The count unit
Characterized in that the area where the human object is located is derived by comparing the coordinates of the center of the area box and the coordinates of two or more points of the counter line.
A device for counting the number of visitors.