KR102197620B1 - An apparatus for providing statistical information of outdoor wear, a method therefor and a computer readable recording medium on which a program for carrying out the method is recorded - Google Patents

Abstract

An apparatus for providing statistical information of outdoor clothing according to an embodiment of the present invention includes a camera unit for generating a front image, a side image, and a rear image by photographing the front, side and rear of the occupant, and the front, side and rear A recognition unit that identifies the gender and age range of the occupant and the type of outdoor clothing worn by the occupant by analyzing the image of the occupant, and statistics on the outdoor clothing worn by the occupant according to the gender and age group of the occupant. It includes a statistics processing unit to generate.

Description

An apparatus for providing statistical information of outdoor wear, a method therefor and a computer readable recording medium for providing statistical information of outdoor wear, a method therefor, and a program for performing the method therefor on which a program for carrying out the method is recorded}

The present invention relates to a technology for generating statistical information of outdoor clothing, and more particularly, an apparatus for generating statistical information of outdoor clothing, a method for the same, and a computer reading a program for performing the method is recorded. It relates to possible recording media.

Outdoor clothing is often made of down, Gore-Tex, and fleece as clothing worn to maintain comfort, protect the human body, and increase activity during outdoor activities such as hiking, exploration, fishing, and sports. 'Outdoor' means'outdoor' and'outdoor wear' and'outdoor clothing' refer to functional clothing for outdoor activities such as hiking, fishing, and camping. Since it is mainly intended to protect the human body from low temperatures, snow, rain, and wind, various expressions such as'heavy duty wear' or'beat the weather look' are used.

Korean Patent Laid-Open Patent No. 2012-0001506 published on January 04, 2012 (Name: A system that enables digital wardrobe coordination and clothing production planning using clothing recognition data)

It is an object of the present invention to provide a device capable of generating and providing statistical information on outdoor wear by residents, a method therefor, and a computer-readable recording medium in which a program for performing the method is recorded.

An apparatus for providing statistical information of outdoor clothing according to a preferred embodiment of the present invention for achieving the above-described object generates a front image, a side image, and a rear image by photographing the front, side and rear of the occupant. A camera unit, a recognition unit that analyzes the image of the front, side, and back to identify the gender and age range of the occupant and the type of outdoor clothing worn by the occupant, and the entrance according to the gender and age of the occupant It characterized in that it comprises a statistical processing unit for generating statistics on the outdoor clothing worn by the person.

The recognition unit receives the front image, the gender and age range of the occupant, and a front recognition unit that outputs a probability of the type of outdoor clothing worn by the occupant, and receives the side image, and the gender of the occupant And a side recognition unit that outputs the probability of the age group and the type of outdoor clothing worn by the occupant, the rear image is input, and the gender and age group of the occupant, and the type of outdoor clothing worn by the occupant. And a rear recognition unit for outputting a probability of, and a determination unit for synthesizing outputs of the front recognition unit, the side recognition unit, and the rear recognition unit to output a probability of a type of outdoor clothing worn by the occupant.

The recognition unit learns the front surface recognition unit using a front image of which the gender and age range of the occupant and the type of outdoor clothing worn by the occupant is known, and the gender and age range of the occupant and the outdoor worn by the occupant After learning the side recognition unit using a side image of which the type of clothing is known, and learning the rear surface recognition unit using a rear image of which the gender and age group of the occupant and the type of outdoor clothing worn by the occupant are known, All of the front, side, and back images of the gender and age range of the occupant and the type of outdoor clothing worn by the occupant are used to determine all of the front, side and rear images. It is characterized by learning.

In a method for providing statistical information of outdoor clothing according to a preferred embodiment of the present invention for achieving the above-described object, the camera unit captures the front, side and rear of the occupant and displays a front image, a side image, and a rear image. Generating, the recognition unit analyzing the front, side, and rear images to identify the gender and age range of the occupant and the type of outdoor clothing worn by the occupant, and the statistical processing unit And generating statistics on outdoor clothing worn by the occupant according to the age group.

The identifying step includes the step of receiving the front image by the front recognition unit of the recognition unit and outputting a probability of the gender and age range of the occupant and the type of outdoor clothing worn by the occupant; and side recognition by the recognition unit An additional step of receiving the side image, outputting the probability of the gender and age range of the occupant, and the type of outdoor clothing worn by the occupant, and the rear recognition unit of the recognition unit receives the rear image, and the entrance The step of outputting the probability of the gender and age range of the child and the type of outdoor clothing worn by the resident, and the determination unit of the recognition unit synthesizes the outputs of the front recognition unit, the side recognition unit and the rear recognition unit to enter the mountain And outputting the probability of the type of outdoor clothing worn by the person.

Before the generating step, the recognition unit learns the front surface recognition unit using a front image of which the gender and age range of the inmate and the type of outdoor clothing worn by the inmate are known, and the gender and age range of the inmate, and the The side recognition unit is learned by using a side image of which the type of outdoor clothing worn by the occupant is known, and the rear surface image by using a rear image of which the gender and age of the occupant and the type of outdoor clothing worn by the occupant are known. Learning the recognition unit, and the recognition unit uses all of the front, side and rear images of which the gender and age range of the occupant and the type of outdoor clothing worn by the occupant are known, and the front recognition unit, the side recognition unit, It further comprises the step of learning both the rear surface recognition unit and the determination unit.

According to another aspect of the present invention, there is provided a computer-readable recording medium in which a program for performing the method for providing statistical information of outdoor clothing according to a preferred embodiment of the present invention is recorded.

The present invention automatically generates and provides statistics on the outdoor clothing worn by the residents at the position where the residents count the numbers, so that statistics on users who enjoy outdoor clothing can be generated. Can be used to establish an effective marketing strategy.

In particular, machine learning for an artificial neural network for generating statistics according to an embodiment of the present invention is an individual learning for each of the front, side, and rear images of the inmates, and all of the front, side and rear images as a whole. Because it is achieved through the entire learning process, statistical results with high accuracy and reliability can be provided through such artificial neural networks.

1 is a block diagram of a system for providing statistical information of outdoor clothing according to an embodiment of the present invention.
2 is a block diagram illustrating a statistical device for providing statistical information on outdoor clothing according to an embodiment of the present invention.
3 to 7 are views for explaining an internal configuration of a recognition unit according to an embodiment of the present invention.
8 is a flowchart illustrating a learning method for a recognition unit according to an embodiment of the present invention.
9 is a flowchart illustrating a method for providing statistical information of outdoor clothing according to an embodiment of the present invention.

Prior to the detailed description of the present invention, terms or words used in the present specification and claims described below should not be construed as being limited to their usual or dictionary meanings, and the inventors shall use their own invention in the best way. For explanation, based on the principle that it can be appropriately defined as a concept of terms, it should be interpreted as a meaning and concept consistent with the technical idea of the present invention. Therefore, the embodiments described in the present specification and the configurations shown in the drawings are only the most preferred embodiments of the present invention, and do not represent all the technical spirit of the present invention, and various equivalents that can replace them at the time of application It should be understood that there may be water and variations.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In this case, it should be noted that the same components in the accompanying drawings are indicated by the same reference numerals as possible. In addition, detailed descriptions of known functions and configurations that may obscure the subject matter of the present invention will be omitted. For the same reason, some components in the accompanying drawings are exaggerated, omitted, or schematically illustrated, and the size of each component does not entirely reflect the actual size.

First, a system for providing statistical information of outdoor clothing according to an embodiment of the present invention will be described. 1 is a block diagram of a system for providing statistical information of outdoor clothing according to an embodiment of the present invention. Referring to FIG. 1, a system for providing statistical information of outdoor clothing according to an embodiment of the present invention includes a statistics device 100 and a statistics server 400.

The statistics device 100 includes a module (hereinafter referred to as a camera unit) that performs at least one camera function. The statistics device 100 is installed so that the camera unit faces the ticket gate 10 installed at the entrance of the hiking trail to photograph the inhabitants. In particular, since the residents must pass through the ticket gate 10 to be able to enter the mountain, the statistics device 100 can photograph the front, side, and rear of the residents while the residents pass through the ticket gate 10 through the camera unit. . The statistical apparatus 100 according to an embodiment of the present invention analyzes the front, side, and rear images of the front, side, and rear of the occupants, and analyzes the gender and age of the occupants and the outdoor clothing worn by the occupants. It identifies the type, generates statistics on the type of outdoor clothing worn according to gender and age group, and provides this to the statistics server 400. The statistics server 400 may receive statistics from the statistics device 100 installed in correspondence with the plurality of ticket gates 10 and aggregate them to construct big data for the statistics.

Then, the statistical apparatus 100 for providing statistical information of outdoor clothing according to an embodiment of the present invention will be described in more detail. 2 is a block diagram illustrating a statistical device for providing statistical information on outdoor clothing according to an embodiment of the present invention.

Referring to FIG. 2, the statistical apparatus 100 according to an embodiment of the present invention includes a communication unit 110, a camera unit 120, an input unit 130, a display unit 140, an audio unit 150, and a storage unit 160. ) And a control unit 200.

The communication unit 110 is a means for communicating with the statistics server 400. The communication unit 110 may include a radio frequency (RF) transmitter Tx for up-converting and amplifying a frequency of a transmitted signal, and an RF receiver Rx for low-noise amplifying and down-converting a received signal. Further, the communication unit 110 may include a modem that modulates a transmitted signal and demodulates a received signal. The communication unit 110 transmits statistics on outdoor clothing according to an embodiment of the present invention to the statistics server 400.

The camera unit 120 is for capturing an image and includes an image sensor. The image sensor receives light reflected from a subject and converts it into an electric signal, and may be implemented based on a Charged Coupled Device (CCD) or Complementary Metal-Oxide Semiconductor (CMOS). The camera unit 120 may further include an analog to digital converter, and may convert an electrical signal output from the image sensor into a digital sequence and output it to the controller 200.

In particular, the camera unit 120 includes a 3D sensor. The 3D sensor is a sensor for obtaining 3D coordinates for feature points of an image captured in a non-contact method. When the camera unit 120 captures an image, the 3D sensor detects the distance to each pixel of the captured image along with the captured image, generates 3D coordinates through the detected distance and transmits it to the control unit 200 do. The 3D sensor can use various types of sensors that use laser, infrared, visible light, and the like. These 3D sensors are laser type 3D scanners using any one of TOP (Time of Flight), phase-shift and Online Waveform Analysis, laser type 3D scanners using optical trigonometry, and optics using white light or modulated light. Method 3D scanner, Handheld Real Time PHOTO, optical 3D scanner, optical method using Pattern Projection or Line Scanning, laser system full body scanner, photo scanner using photogrammetry, Kinect Fusion A real time scanner to be used can be illustrated.

The input unit 130 may receive a user's key manipulation for controlling the statistical apparatus 100, generate an input signal, and transmit the input signal to the control unit 200. The input unit 130 may include various keys for controlling the statistical apparatus 100. When the display unit 140 is formed of a touch screen, the input unit 130 may perform functions of various keys on the display unit 140, and when all functions can be performed only with the touch screen, the input unit 130 will be omitted. May be.

The display unit 140 is for screen display, and may visually provide a menu, input data, function setting information, and various other information of the statistical apparatus 100 to a user. In addition, the display unit 140 performs a function of outputting screens such as a boot screen, a standby screen, a menu screen, and the like of the statistics device 100. The display unit 140 may be formed of a liquid crystal display (LCD), an organic light emitting diode (OLED), an active matrix organic light emitting diode (AMOLED), or the like. Meanwhile, the display unit 140 may be implemented as a touch screen. In this case, the display unit 140 includes a touch sensor. The touch sensor detects a user's touch input. The touch sensor may be composed of a touch sensing sensor such as a capacitive overlay, a pressure type, a resistive overlay, or an infrared beam, or may be composed of a pressure sensor. . In addition to the above sensors, all kinds of sensor devices capable of sensing contact or pressure of an object may be used as the touch sensor of the present invention. The touch sensor detects a user's touch input, generates a detection signal, and transmits it to the controller 200. In particular, when the display unit 140 is formed of a touch screen, some or all functions of the input unit 130 may be performed through the display unit 140.

The audio unit 150 includes a speaker (SPK) for outputting an audio signal such as voice according to an embodiment of the present invention, and a microphone (MIKE) for collecting an audio signal such as voice. That is, the audio unit 150 may output an audio signal through the speaker SPK or transmit an audio signal input through the microphone (MIKE) to the controller 200 under control of the controller 200.

The storage unit 160 serves to store programs and data necessary for the operation of the statistical apparatus 100. Each type of data stored in the storage unit 160 may be deleted, changed, or added according to a user's manipulation.

The controller 200 may control the overall operation of the statistical apparatus 100 and a signal flow between internal blocks of the statistical apparatus 100 and perform a data processing function of processing data. In addition, the control unit 200 basically serves to control various functions of the statistical apparatus 100. The control unit 200 may be a central processing unit (CPU), a digital signal processor (DSP), or the like. The control unit 200 includes a recognition unit 210 and a statistics processing unit 220. The operation of the control unit 200 including the recognition unit 210 and the statistics processing unit 220 will be described in more detail below.

Next, the recognition unit 210 according to an embodiment of the present invention will be described in more detail. 3 to 7 are views for explaining an internal configuration of a recognition unit according to an embodiment of the present invention.

3 to 7, the recognition unit 220 includes a front recognition unit 221, a side recognition unit 212, a rear recognition unit 213 and a determination unit 214. Each of the front recognition unit 221, the side recognition unit 212, and the rear recognition unit 213 and the determination unit 214 is formed of artificial neural networks (ANNs).

In particular, as shown in Figs. 4 to 6, each of the front recognition unit 221, the side recognition unit 212 and the rear recognition unit 213 is preferably implemented as a convolutional neural network (CNN). Do. However, the front recognition unit 221, the side recognition unit 212, and the rear recognition unit 213 are not limited to a convolutional neural network, and an image is input, and the gender and age group of the inmates from the image, and the inmates wear it. If an artificial neural network (ANN) is machine-learned to identify a type of outdoor clothing, it may be applied to an embodiment of the present invention regardless of the type.

Each of the front recognition unit 221, the side recognition unit 212 and the rear recognition unit 213 includes a plurality of layers. The plurality of layers configures the next layer through a plurality of operations in which the output of any one layer is weighted. Here, the weight (W) determines the strength of the inter-layer connection.

Each of the front recognition unit 221, the side recognition unit 212 and the rear recognition unit 213 is an input layer (INL), a convolution layer (CL), a pooling layer (PL), and It includes a fully-connected layer (FCL) and an output layer (ML).

The input layer IL consists of a matrix of a predetermined size. Each element of the input layer (IL) matrix corresponds to a plurality of pixels of an input image.

4 to 6, two convolutional layers (CL: CLF, CLS) and two pooling layers (PL: PLF, PLS) are alternately repeated, but the present invention is not limited thereto. It will be appreciated by those of ordinary skill in the art that the number and arrangement order of the convolutional layers CL and the pooling layers PL may vary depending on the design of the artificial neural network. Each of the convolutional layer CL and the pooling layer PL includes a plurality of feature maps indicated by solid squares, and each of these feature maps is a matrix having a predetermined size. The value of each element of the matrix constituting the feature map is calculated by applying a convolution operation or pooling or subsampling using a kernel to the previous layer. Here, the kernel is a matrix of a predetermined size indicated by a dotted rectangle, and the value of each element of the matrix constituting the kernel becomes a weight (W).

The fully connected layer (FCL) includes a plurality of nodes (or sigmoid: C1, C2, C3......Cn), and the output layer (OUL) is a plurality of output nodes (F1 to Fm or S1 to Sm or B1 to Bm). The calculation of the fully connected layer FL is also applied to a weight W and input to a plurality of output nodes F1 to Fm or S1 to Sm or B1 to Bm of the output layer OL. Each of the plurality of output nodes (F1 to Fm, S1 to Sm, and B1 to Bm) corresponds to the gender and age group of the inhabitants and the type of outdoor clothing worn by the inhabitants. For example, each of the plurality of output nodes (F1 to Fm or S1 to Sm or B1 to Bm) is male, female, 10s, 20s, 30s, 40s, 50s, 60s or more, A company padding, B company It may correspond to padding, company C padding, ..., company A long down, company B long down, company C long down, ..., company A fleece, company B fleece, company C fleece, etc.

For example, the first output node F1 or S1 or B1 may correspond to a gender male, and the second output node F2 or S2 or B2 may correspond to a gender female. Accordingly, the output value of the first output node (F1, S1, or B1) indicates the probability that the person entering the image is a male. In addition, the output value of the second output node (F2 or S2 or B2) represents the probability that the initiating person of the corresponding image is a girl. For example, if the first output value that is the output of the first output node (F1 or S1 or B1) is 0.35 and the second output value that is the output of the second output node (F2 or S2 or B2) is 0.65, the probability of excitation is 35%. And the probability of being a male is 65%.

As another example, the m-th output node (Fm or Sm or Bm) may correspond to the K company's padding. Accordingly, the output value of the m-th output node (Fm, Sm, or Bm) represents the probability that the clothing worn by the resident of the corresponding image is the K company's padding. For example, if the first output value, which is the output of the m-th output node (Fm or Sm or Bm), is 0.88, it indicates that the probability that the clothing worn by the resident of the corresponding image is padding of Company K is 88%.

Each of the plurality of layers (INL, CL, PL, FL, and OUL) includes a plurality of operations. Each of a plurality of operations of a plurality of layers (INL, CL, PL, FL, and OUL) is applied with a weight (w), and the result of the operation to which the weight (w) is applied is transferred to the next layer. That is, the operation result of the previous layer becomes the input of the next layer.

As described above, the input layer IL is a feature map that is a matrix of a predetermined size. The elements of the matrix of the input layer IL are pixel units. Each element of the matrix may be a pixel value of each pixel of the front image, and the pixel value may be input as binary data to an element of the matrix of the input layer IL.

Then, a convolution operation using each of the plurality of kernels W is performed on the input layer matrix, and the result of the operation is input to a plurality of feature maps of the first convolutional layer CLF. Here, each of the plurality of kernels W may use a matrix having a predetermined size in which an element of the matrix is a weight W. In addition, each of the plurality of feature maps of the first convolutional layer CLF is a matrix having a predetermined size.

Next, a pooling operation (subsampling) using a plurality of kernels W is performed on a plurality of feature maps of the first convolutional layer CLF. A plurality of kernels (W) is also a matrix of a predetermined size, each of which is composed of a weight (W). The result of the subsampling operation is input to a plurality of feature maps of the first pooling layer PLF. Each of the plurality of feature maps of the first pooling layer PLF is also a matrix having a predetermined size.

Subsequently, a convolution operation (convolution) is performed using a kernel (W), which is a matrix of a predetermined size in which each element of the matrix is a weight (W) with respect to the plurality of feature maps of the first pooling layer (PLF). A second convolutional layer (CLS) consisting of a map is constructed. Next, a second pooling consisting of a plurality of feature maps is performed by performing a subsampling operation using a kernel W, which is a matrix consisting of a plurality of weights w, for a plurality of feature maps of the second convolutional layer (CLS). Construct a layer (PLS). Each of the second pooling layers PLS is also a matrix having a predetermined size.

Thereafter, a convolution operation using a plurality of kernels W is performed on a plurality of feature maps of the second pooling layer PLS. A plurality of kernels (W) is also a matrix of a predetermined size whose elements are weights (w). A fully connected layer (FCL) is generated according to a result of a convolution operation using a plurality of kernels (W). In other words, the result of a convolution operation using a plurality of kernels W is input to a plurality of nodes C1 to Cn.

Each of the plurality of nodes (C1 to Cn) of the fully connected layer (FL) performs a predetermined operation using a transfer function, etc. for an input from the second pooling layer (PLS), and applies a weight (w) to the operation. Input to each node of the output layer (OUL). Accordingly, a plurality of output nodes (F1 to Fm or S1 to Sm or B1 to Bm) of the output layer (OUL) perform a predetermined operation on the value input from the fully connected layer (FCL), and the resulting output value Prints. As described above, each of the plurality of output nodes (F1 to Fm or S1 to Sm or B1 to Bm) corresponds to the gender and age of the occupant and the type of outdoor clothing worn by the occupant, respectively, and such a plurality of output nodes (F1 to Fm or S1 to Sm or B1 to Bm) Each output value is a probability value corresponding to the gender and age range of the occupant of the corresponding front image and the type of outdoor clothing worn by the occupant.

The determination unit 214 synthesizes the outputs of the front recognition unit 221, the side recognition unit 212, and the rear recognition unit 213, and the gender and age range of the residents, and the type of outdoor clothing worn by the residents. To identify. In particular, the determination unit 214 receives the weighted output of the front recognition unit 221, the side recognition unit 212, and the rear recognition unit 213, as shown in FIG. 7, and through a predetermined operation. It includes a plurality of output nodes (O1 to Om) for outputting the output value. The output values of each of the plurality of output nodes O1 to Om correspond to the gender and age group of the occupant of the corresponding front image and the type of outdoor clothing worn by the occupant, and the output value is a corresponding probability value.

The front recognition unit 221, the side recognition unit 212, and the rear recognition unit 213 and the determination unit 214 of the above-described recognition unit 222 include the gender and age range of the residents, and the outdoor worn by the residents. It is machine learning to identify the type of clothing. This learning method will be described. 8 is a flowchart illustrating a learning method for a recognition unit according to an embodiment of the present invention.

Referring to FIG. 8, the recognition unit 220 performs learning by using a front image (FIMG), a side image (SIMG), and a rear image (BIMG) in which the sex and age groups, which are training data, and the type of outdoor clothing worn are known. Perform.

First, the recognition unit 220 individually learns each of the front recognition unit 211, the side recognition unit 212, and the rear recognition unit 213 in step S110. In this step S110, the recognition unit 220 learns the front recognition unit 211 through a front image (FIMG) in which the gender and age group and the type of outdoor clothing worn are known, and the gender and age group, and the worn outdoor clothing The side recognition unit 212 is learned through a side image (SIMG) of which the type is known, and the rear recognition unit 213 is learned through a rear image (BIMG) in which the gender and age group and the type of outdoor clothing worn are known.

When individual learning in step S110, the learning method for the front recognition unit 211 will be described in more detail as follows. First, the recognition unit 220 sets an expected value according to the known gender and age range of the front image FIMG, which is the training data, and the type of outdoor clothing worn. Here, the expected value may be a target value for the gender and age group to be learned from the corresponding front image FIMG, and the type of outdoor clothing worn. For example, if the entrance of the front image (FIMG) is a male, 40s, and a video wearing K's padding, the expected value is set so that the expected value of the node corresponding to the male is higher than the expected value of the node corresponding to the female. , The expected value of the node corresponding to the 40s is set to have the highest value compared to the expected value of the node corresponding to other age groups, and the expected value of the node corresponding to company K's padding is in the node corresponding to the type of outdoor clothing different. It is set to have a higher value than that. For example, expected values of each of the nodes corresponding to men and women may be set to 0.90 and 0.10. In addition, expected values of each node corresponding to those in their 20s, 30s, 40s, 50s, 60s or more may be set to 0.04, 0.16, 0.60, 0.15, and 0.05. And, assuming that the node of the type of outdoor clothing has a node corresponding to company K's padding, company L's padding, and company M's padding, the expected value of each node corresponding to company K's padding, company L's padding, and company M's padding is 0.80 , Can be set to 0.10, 0.10.

Then, when the front image FIMG, which is the training data, is input to the front recognition unit 211, the front recognition unit 211 uses a plurality of layers INL, for a plurality of pixel values of the front image FIMG, which is the training data. A plurality of operations to which the weights W of CL, PL, FL, OUL) are applied are performed, and output values corresponding to the gender and age group and the type of outdoor clothing worn are output through the output nodes F1 to Fm. For example, each of the output nodes (F1 to Fm) corresponds to male, female, under 20s, 30s, 40s, 50s, 60s or more, padding of company K, padding of company L and padding of company M, and output nodes ( F1 to Fm), each output value is assumed to be derived as 0.50, 0.50, 0.14, 0.26, 0.40, 0.25, 0.15, 0.60, 0.20, 0.20.

In this way, before sufficient learning is performed, that is, before the learning is completed, the output value of the front recognition unit 211 may be different from the expected value. Accordingly, the recognition unit 210 compares the output value and the expected value, and corrects the weight W of the front recognition unit 211 through a back-propagation algorithm so that the difference between the expected value and the output value is minimized.

In this way, the recognition unit 210 performs a learning procedure of modifying the weight W of the front recognition unit 211 through a backpropagation algorithm when an output value is derived by inputting training data for which an expected value is set.

The recognition unit 210 repeats the above-described learning procedure for the front recognition unit 211 by using a front image FIMG, which is a plurality of different training data. In particular, when the recognition unit 210 inputs any one learning data to the front recognition unit 211, if the difference between the expected value and the output value is within a predetermined value and there is no change in the output value, the recognition unit 210 determines that learning is complete, and the learning End the procedure. The recognition unit 210 performs individual learning on the side recognition unit 212 and the rear recognition unit 213 in the same manner as the learning procedure for the front recognition unit 211 described above. However, for the side recognition unit 212, a side image (SIMG) of which gender and age group and the type of outdoor clothing worn are known as learning data is used, and for the rear recognition unit 213, gender and age group, The rear image (BIMG) of which the type of outdoor clothing worn is known is used.

When the above-described individual learning procedure is finished, the recognition unit 210 does not individually learn each of the front recognition unit 211, the side recognition unit 212 and the rear recognition unit 213 in step S120, but the front recognition unit (211), the side recognition unit 212, the rear recognition unit 213 and the determination unit 214 are learned at the same time. This step S120 may be divided into a deficiency learning step S121 and a complete learning step S123.

In the deficiency learning of step S121, the recognition unit 210 inputs a front image (FIMG) as training data to the front recognition unit 211, and inputs a side image (SIMG) as training data to the side recognition unit 212 , The rear image (BIMG), which is the training data, must be input to the rear recognition unit 213, but the training data is input to one or two of the front recognition unit 211, the side recognition unit 212, and the rear recognition unit 213. Without typing, learning is performed. On the other hand, in the complete learning of step S123, the recognition unit 210 inputs a front image (FIMG), which is training data, to the front recognition unit 211, unlike deficiency learning, and the side that is learning data to the side recognition unit 212. An image (SIMG) is input, and a rear image (BIMG), which is training data, is input to the rear recognition unit 213 to perform learning.

In more detail, the deficiency learning in step S121 is as follows. First, the recognition unit 220 performs deficiency learning using learning data excluding one or two of the front image (FIMG), the side image (SIMG), and the rear image (BIMG), and the known gender and The expected value is set according to the age group and the type of outdoor clothing worn. Here, the expected value may be a target value for a gender and age group to be learned from the corresponding learning data, and a type of outdoor clothing worn. For example, if the entry into the training data is a male, in their 40s, and an image wearing K company's padding, the expected value is set so that the expected value of the node corresponding to the male is higher than the expected value of the node corresponding to the female. The expected value of the node corresponding to is set to have the highest value compared to the expected value of the node corresponding to other age groups, and the expected value of the node corresponding to company K's padding is higher than that of the node corresponding to other types of outdoor clothing. Is set to have. For example, expected values of each of the nodes corresponding to men and women may be set to 0.90 and 0.10. In addition, expected values of each node corresponding to those in their 20s, 30s, 40s, 50s, 60s or more may be set to 0.04, 0.16, 0.60, 0.15, and 0.05. And, assuming that the node of the type of outdoor clothing has a node corresponding to company K's padding, company L's padding, and company M's padding, the expected value of each node corresponding to company K's padding, company L's padding, and company M's padding is 0.80 , Can be set to 0.10, 0.10. In this way, the method of setting the expected value is the same as in individual learning.

After setting the expected value, the recognition unit 210 converts the learning data excluding one or two of the front image (FIMG), the side image (SIMG), and the rear image (BIMG) into the front recognition unit 211 and the side recognition unit. (212) and input to the rear recognition unit 213. For example, the recognition unit 210 inputs a front image (FIMG) to the front recognition unit 211, inputs a side image (SIMG) to the side recognition unit 212, and a rear image (BIMG) to the rear recognition unit. It may not be entered in (213). As another example, the recognition unit 210 may input only the side image SIMG to the side recognition unit 212 and may not input the front image FIMG and the rear image BIMG.

Among the front recognition unit 211, the side recognition unit 212, and the rear recognition unit 213, the module into which the learning data is input is a plurality of front image (FIMG), side image (SIMG), and rear image (BIMG), which are training data. Through the output nodes (F1 to Fm or S1 to Sm or B1 to Bm) by performing a plurality of operations to which the weights (W) of a plurality of layers (INL, CL, PL, FL, OUL) are applied to the pixel values of Print the output value.

And the output values of these output nodes (F1 to Fm or S1 to Sm or B1 to Bm), as shown in FIG. 7, a weight (W) is applied to the output nodes (O1 to Om) of the determination unit 214 It is input, and the output nodes O1 to Om output an output value through an operation.

For example, each of the output nodes O1 to Om corresponds to male, female, under 20s, 30s, 40s, 50s, 60s or more, padding of K company, padding of company L, and padding of company M, and output nodes ( F1 to Fm), each output value is assumed to be derived as 0.50, 0.50, 0.14, 0.26, 0.40, 0.25, 0.15, 0.60, 0.20, 0.20.

As described above, before sufficient learning is performed, that is, before learning is completed, the output value may be different from the expected value. Therefore, the recognition unit 210 compares the output value and the expected value, and uses a back-propagation algorithm to minimize the difference between the expected value and the output value. The front recognition unit 211, the side recognition unit 212, and the rear recognition unit (213) and the weight (W) of the determination unit 214 are corrected. In this way, the recognition unit 210 inputs training data excluding one or two of the front image (FIMG), the side image (SIMG), and the rear image (BIMG) in which the expected value is set, and when the output value is derived, the backpropagation algorithm Through the front recognition unit 211, the side recognition unit 212, the rear recognition unit 213, and performs a deficiency learning procedure of modifying the weight (W) of the determination unit 214.

The recognition unit 210 may repeatedly perform the deficiency learning procedure using a plurality of different learning data. In addition, when the difference between the expected value and the output value is within a predetermined value while the output value does not change during the repeated deficiency learning, the recognition unit 210 determines that the learning has been completed, and ends the deficiency learning procedure.

The complete learning in step S123 is performed using all of the front image (FIMG), the side image (SIMG), and the rear image (BIMG) as training data, and the remaining procedures are performed in the same manner as the deficiency learning in step S121 described above.

That is, the recognition unit 210 sets an expected value, and transmits training data including all of a front image (FIMG), a side image (SIMG), and a rear image (BIMG) to the front recognition unit 211 and the side recognition unit 212. And after input to the rear recognition unit 213, through a back-propagation algorithm so that the difference between the expected value and the output values of the output nodes O1 to Om of the determination unit 214 is minimal. 211), the weight (W) of the side recognition unit 212, the rear recognition unit 213 and the determination unit 214 is corrected. In addition, when the difference between the expected value and the output value is within a predetermined value while the output value does not change during the repeated complete learning, the recognition unit 210 determines that the learning has been completed, and ends the complete learning process.

When the learning process is finished, the recognition unit 210 is the gender and age group from the front image (FIMG), the side image (SIMG), and the rear image (BIMG) of which the type of outdoor clothing worn is unknown. You can identify the type of outdoor clothing.

Then, a method for providing statistical information of outdoor clothing according to an embodiment of the present invention will be described. 9 is a flowchart illustrating a method for providing statistical information of outdoor clothing according to an embodiment of the present invention.

Referring to FIG. 9, the control unit 200 of the statistics device 100 photographs the front, side and rear of the occupant passing through the ticket gate 10 through the camera unit 120 in step S210, and a front image in step S220. (FIMG), lateral image (SIMG), and rear image (BIMG) are generated.

Then, the recognition unit 210 of the control unit 200 analyzes the front image (FIMG), the side image (SIMG), and the rear image (BIMG) according to what was previously learned in step S230 (the embodiment of FIG. 8) Identify the child's gender and age range, and the type of outdoor clothing worn. In more detail, the recognition unit 210 transmits a front image (FIMG), a side image (SIMG), and a rear image (BIMG) to the front recognition unit 211, the side recognition unit 212 and the rear recognition unit 213. Enter. Then, the front recognition unit 211, the side recognition unit 212, and the rear recognition unit 213 each have a plurality of pixel values of a front image (FIMG), a side image (SIMG), and a rear image (BIMG). A plurality of operations to which the weights W of the layers INL, CL, PL, FL, and OUL are applied are performed, and the output values are output through the output nodes F1 to Fm or S1 to Sm or B1 to Bm. And the output values of these output nodes (F1 to Fm or S1 to Sm or B1 to Bm), as shown in FIG. 7, a weight (W) is applied to the output nodes (O1 to Om) of the determination unit 214 It is input, and the output nodes O1 to Om output an output value through an operation. For example, each of the output nodes O1 to Om corresponds to male, female, under 20s, 30s, 40s, 50s, 60s or more, padding of K company, padding of company L, and padding of company M, and output nodes ( F1 to Fm), each output value is assumed to be derived as 0.30, 0.70, 0.14, 0.26, 0.40, 0.25, 0.15, 0.10, 0.10, 0.80. Accordingly, since the node corresponding to the male is 0.30, the node corresponding to the female is 0.70, the probability of being male is 30%, and the probability of being female is 70%, so the recognition unit 210 determines the inmate as female. Likewise, since the probability of being in the 40's is the highest at 40%, the recognition unit 210 determines the person in the 40's. In addition, since the probability of the company M's padding is the highest at 80%, the recognition unit 210 determines the type of outdoor clothing worn by the corresponding resident as the company M's padding.

Finally, the statistics processing unit 220 of the control unit 200 accumulates other data and generates statistics according to the recognition unit 210 recognizes. That is, the statistics processing unit 220 generates statistics on outdoor clothing worn by the inmate according to the gender and age group of the inmate recognized by the recognition unit 210.

Additionally, the statistics processing unit 220 may transmit the generated statistics to the statistics server 400 through the communication unit 110.

Meanwhile, the methods according to the embodiments of the present invention described above may be implemented in the form of programs readable by various computer means and recorded on a computer readable recording medium. Here, the recording medium may include a program command, a data file, a data structure, or the like alone or in combination. The program instructions recorded on the recording medium may be specially designed and configured for the present invention, or may be known and usable to those skilled in computer software. For example, the recording medium includes magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic-optical media such as floptical disks ( magneto-optical media), and hardware devices specially configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of the program instruction may include not only machine language wires such as those made by a compiler, but also high-level language wires that can be executed by a computer using an interpreter or the like. These hardware devices may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

The present invention has been described above using several preferred embodiments, but these embodiments are illustrative and not limiting. As such, those of ordinary skill in the art to which the present invention pertains will understand that various changes and modifications can be made according to the equivalence theory without departing from the spirit of the present invention and the scope of the rights presented in the appended claims.

100: statistics device
110: communication unit 120: camera unit
130: input unit 140: display unit
150: audio unit 160: storage unit
200: control unit 210: recognition unit
220: statistics processing unit 400: statistics server

Claims (7)

In the device for providing statistical information of outdoor clothing,
A camera unit for generating a front image, a side image, and a rear image by photographing the front, side and rear of the inmate;
A front recognition unit for processing the front image, each of which is an artificial neural network, a side recognition unit for processing the side image, a rear recognition unit for processing the rear image, the front recognition unit, the side recognition unit, and the rear surface A recognition unit including a determination unit for processing an output of the recognition unit, and identifying a gender and age range of the occupant and a type of outdoor clothing worn by the occupant by analyzing the front image, the side image and the rear image; And
Includes; a statistical processing unit that generates statistics on the outdoor clothing worn by the inmate according to the gender and age of the inmate,
The recognition unit
The front image recognition unit is learned using the front image of the occupant's gender and age group and the type of outdoor clothing worn by the occupant, and a side image of which the gender and age group of the occupant and the type of outdoor clothing worn by the occupant are known After learning the side recognition unit by using and performing individual learning of learning the rear recognition unit using the rear image of which the gender and age of the occupant and the type of outdoor clothing worn by the occupant are known,
Preparing learning data excluding one or two of the front image, the side image, and the rear image, setting an expected value for the prepared learning data, excluding one or two of the side image and the rear image Input the learning data to a corresponding module among the front recognition unit, the side recognition unit, and the rear recognition unit, and the module to which learning data is input from the front recognition unit, the side recognition unit and the rear recognition unit, and the determination unit Backpropagation algorithm so that the difference between the expected value and the output value is minimized by comparing the output value and the expected value when outputting the output value by performing a plurality of operations in which the weights (W) of a plurality of layers are applied to a plurality of pixel values of the training data Through the front recognition unit, the side recognition unit, the rear recognition unit, and performing deficiency learning to correct the weight of the determination unit,
Preparing training data including all of the front image, the side image, and the rear image, setting an expected value for the prepared training data, and storing training data including all of the front image, the side image, and the rear image Input to the front recognition unit, the side recognition unit and the rear recognition unit, and the front recognition unit, the side recognition unit, the rear recognition unit, and the determination unit apply weights of a plurality of layers to a plurality of pixel values of the training data. When the output value is output by performing a plurality of operations, the output value and the expected value are compared, and the front recognition unit, the side recognition unit, the rear surface recognition unit and the determination are performed through a backpropagation algorithm so that the difference between the expected value and the output value is minimized. Performing perfect learning to modify negative weights
Device for providing statistical information. The method of claim 1,
The recognition unit
A front face recognition unit configured to receive the front image and output a probability of the gender and age range of the occupant and the type of outdoor clothing worn by the occupant;
A side recognition unit configured to receive the side image and output a probability of the gender and age range of the occupant and the type of outdoor clothing worn by the occupant;
A rear recognition unit configured to receive the rear image and output a probability of the gender and age range of the occupant and the type of outdoor clothing worn by the occupant; And
And a determination unit for synthesizing the outputs of the front recognition unit, the side recognition unit, and the rear recognition unit to output a probability for the type of outdoor clothing worn by the occupant.
Device for providing statistical information. The method of claim 1,
The recognition unit
Learning the front face recognition unit using a front image in which the gender and age range of the occupant and the type of outdoor clothing worn by the occupant are known,
Learning the side recognition unit by using a side image in which the gender and age range of the inmate and the type of outdoor clothing worn by the inmate are known,
After learning the rear surface recognition unit using a rear image in which the gender and age range of the occupant and the type of outdoor clothing worn by the occupant are known,
Using all of the front, side, and rear images of the gender and age range of the resident and the type of outdoor clothing worn by the resident
Characterized in that learning all of the front recognition unit, the side recognition unit, the rear recognition unit, and the determination unit
Device for providing statistical information. In a method for providing statistical information of outdoor clothing,
Generating a front image, a side image, and a rear image by a camera unit photographing the front, side, and rear of the occupant;
A front recognition unit for processing the front image, each of which is an artificial neural network, a side recognition unit for processing the side image, a rear recognition unit for processing the rear image, the front recognition unit, the side recognition unit, and the rear surface Identifying a gender and age range of the occupant and a type of outdoor clothing worn by the occupant by analyzing the front image, the side image, and the rear image by a recognition unit including a determination unit processing an output of the recognition unit;
Comprising, by a statistical processing unit, generating statistics on the outdoor clothing worn by the occupant according to the gender and age of the occupant,
Before the step of generating the front image, the side image and the back image,
The recognition unit learns the front surface recognition unit using a front image in which the gender and age range of the occupant and the type of outdoor clothing worn by the occupant is known, and the gender and age range of the occupant and the outdoor clothing worn by the occupant Individual learning to learn the side recognition unit by using a side image of which the type of is known, and to learn the rear surface recognition unit using a rear image of which the gender and age of the occupant and the type of outdoor clothing worn by the occupant are known. Performing;
The recognition unit prepares training data excluding one or two of the front image, the side image, and the rear image, sets an expected value for the prepared learning data, and sets one or two of the side image and the rear image A module in which learning data excluding dogs is inputted to a corresponding module among the front recognition unit, the side recognition unit, and the rear recognition unit, and learning data is input from the front recognition unit, the side recognition unit, and the rear recognition unit, and When the determination unit outputs an output value by performing a plurality of operations in which the weights (W) of a plurality of layers are applied to a plurality of pixel values of the training data, the difference between the expected value and the output value is minimized by comparing the output value and the expected value. Performing deficiency learning of correcting weights of the front recognition unit, the side recognition unit, the rear recognition unit, and the determination unit through a back propagation algorithm;
Learning that the recognition unit prepares training data including all of the front image, the side image, and the rear image, sets an expected value for the prepared training data, and includes all of the front image, the side image, and the rear image Data is input to the front recognition unit, the side recognition unit, and the rear surface recognition unit, and the front recognition unit, the side recognition unit, the rear surface recognition unit, and the determination unit include a plurality of layers for the plurality of pixel values of the learning data. When the output value is output by performing a plurality of calculations to which the weight is applied, the front recognition unit, the side recognition unit, and the rear recognition unit are compared through a backpropagation algorithm so that the difference between the expected value and the output value is minimized. And performing complete learning of correcting the weight of the determination unit.
It characterized in that it further comprises
Methods for providing statistical information. The method of claim 4,
The identifying step
Receiving the front image of the recognition unit, and outputting a probability of a gender and age range of the occupant and a type of outdoor clothing worn by the occupant;
Receiving the side image by the side recognition unit of the recognition unit, and outputting a probability of a gender and age range of the occupant and a type of outdoor clothing worn by the occupant;
Receiving the rear image and outputting a probability of a gender and age range of the occupant and a type of outdoor clothing worn by the occupant; And
And outputting a probability of a type of outdoor clothing worn by the occupant by synthesizing the output of the front recognition unit, the side recognition unit, and the rear recognition unit by the determination unit of the recognition unit.
Methods for providing statistical information. The method of claim 4,
Before the step of generating,
The recognition unit learns the front surface recognition unit by using a front image of which the gender and age range of the occupant and the type of outdoor clothing worn by the occupant is known, and the gender and age range of the occupant and the outdoor worn by the occupant Learning the side recognition unit using a side image of which a type of clothing is known, and learning the rear surface recognition unit using a rear image of which the gender and age group of the occupant and the type of outdoor clothing worn by the occupant are known; And
The recognition unit uses all of the front, side, and rear images of the gender and age range of the occupant and the type of outdoor clothing worn by the occupant to determine the front surface recognition unit, the side surface recognition unit, the rear surface recognition unit, and the determination The step of learning all of the parts; characterized in that it further comprises
Methods for providing statistical information. A computer-readable recording medium on which a program for performing the method for providing statistical information according to any one of claims 4 to 6 is recorded.

Images