KR102164950B1 - Method and system for multi-pedestrian tracking using teacher-student random ferns - Google Patents

Abstract

The present invention relates to a method for tracking a plurality of pedestrians using a teacher-student random fun, and more specifically, as a method for tracking a plurality of pedestrians, (1) photographing an image including a plurality of pedestrians from a camera installed in a moving vehicle. step; (2) detecting a plurality of pedestrians in the image captured in step (1); (3) extracting feature values by inputting an image including a plurality of pedestrians detected in step (2) into a deep network; (4) learning a Teacher-Student Random Ferns using the feature values extracted in step (3); And (5) tracking a plurality of pedestrians using the Teacher-Student Random Ferns learned in step (4).
In addition, the present invention relates to a plurality of pedestrian tracking systems 10 using a teacher-student random fun, and more specifically, as a plurality of pedestrian tracking systems 10, including a plurality of pedestrians in the camera installed in the moving vehicle A camera unit 100 for photographing an image; A detection unit 200 for detecting a plurality of pedestrians in the image captured by the camera unit 100; An extraction unit 300 for extracting a feature value by inputting an image including a plurality of pedestrians detected by the detection unit 200 into a deep network; A learning unit 400 for learning teacher-student random ferns using the feature values extracted by the extraction unit 300; And a tracking unit 500 for tracking a plurality of pedestrians using a teacher-student random fern learned in the learning unit 400.
According to a method and system for tracking a number of pedestrians using a teacher-student random fun proposed in the present invention, feature values of pedestrians are extracted using tiny YOLO, a kind of deep network, and the extracted feature values are used. By learning Random Ferns, real-time learning is possible, and mis-tracking due to changes in shape and size of pedestrians can be minimized.
In addition, according to the method and system for tracking a number of pedestrians using a teacher-student random fun proposed in the present invention, in order to enable real-time tracking by reducing the number of Ferns, a teacher-student random fun (Teacher- Student Random Ferns) can be used to quickly and accurately track large numbers of pedestrians in real time.

Description

A method and system for tracking multiple pedestrians using teacher-student random fun {METHOD AND SYSTEM FOR MULTI-PEDESTRIAN TRACKING USING TEACHER-STUDENT RANDOM FERNS}

The present invention relates to a method and system for tracking a plurality of pedestrians, and more particularly, to a method and system for tracking a plurality of pedestrians using a teacher-student random fun.

In order to prevent collisions between pedestrians and vehicles in Intelligent Transportation System (ITS) and Advanced Driver Assistance System (ADAS), technology for detecting and tracking pedestrians is essential.

To achieve the appropriate level of safety in an active intelligent traffic system, advanced driver assistance systems must track all pedestrians on the move to identify pedestrians at risk of entering the road in advance.

Among the various methods of tracking pedestrians, there is a method of tracking pedestrians using a Kalman filter. The Kalman filter is a recursive filter that tracks a mechanical state containing noise, and is based on measurements taken over time. The Kalman filter repeatedly performs state prediction and measurement update when the motion model and measurement model are linear, or when the motion model and measurement model follow Gaussian distribution, but it cannot be used unless the above two cases are applicable. have.

Recently, a method of tracking pedestrians using a convolutional neural network (CNN) has been studied, but it is easy to track a single pedestrian using a synthetic product neural network, but in order to track a large number of pedestrians, a large amount of There is a problem in that parameters are required and the amount of data to be processed is large, so that it is not suitable for a real-time tracking environment.

Therefore, it is required to develop a number of pedestrian tracking methods and systems that use a small amount of parameters and are suitable for a real-time tracking environment.

Meanwhile, as a prior art related to the present invention, Patent No. 10-1588648 (name of the invention: pedestrian detection and tracking method for intelligent video monitoring) has been disclosed.

The present invention has been proposed to solve the above problems of the previously proposed methods, and extracts the feature values of pedestrians using tiny YOLO, a kind of deep network, and uses the extracted feature values to perform random fun ( Random Fern), which enables real-time learning to minimize mis-tracking due to change in shape and size of pedestrians, and to provide a number of pedestrian tracking methods and systems using teacher-student random fun. To do.

In addition, the present invention, in order to enable real-time tracking by reducing the number of Ferns, using a teacher-student random fern (Teacher-Student Random Ferns), it is possible to quickly and accurately track a large number of pedestrians in real time. Another object is to provide a method and system for tracking a number of pedestrians using a teacher-student random fun.

A plurality of pedestrian tracking methods using a teacher-student random fern according to a feature of the present invention for achieving the above object,

As a number of pedestrian tracking methods,

(1) taking an image including a plurality of pedestrians from a camera installed in a moving vehicle;

(2) detecting a plurality of pedestrians in the image captured in step (1);

(3) extracting feature values by inputting an image including a plurality of pedestrians detected in step (2) into a deep network;

(4) learning a Teacher-Student Random Ferns using the feature values extracted in step (3); And

(5) It is characterized in that it comprises the step of tracking a plurality of pedestrians using the teacher-student random fun (Teacher-Student Random Ferns) learned in step (4).

Preferably, the step (2),

(2-1) distinguishing between pedestrians and non-pedestrians in the image captured in step (1); And

(2-2) It may include the step of detecting the pedestrians identified in step (2-1) as the plurality of pedestrians.

Preferably, the deep network in step (3),

It can be a synthetic product neural network.

More preferably, the synthetic product neural network may be tiny YOLO.

Even more preferably, the tiny YOLO,

It may consist of 9 convolution layers, 6 max pooling layers, and 1 fully connected layer.

Preferably, in step (3),

Feature values may be extracted for each of the plurality of pedestrians detected in step (2).

Preferably, the step (4),

(4-1) learning Teacher Random Ferns using the feature values extracted in step (3); And

(4-2) It may include the step of learning a Student Random Fern using the Teacher Random Fern learned in step (4-1).

More preferably, the teacher random fern (Teacher Random Ferns),

You can have a plurality of Ferns.

More preferably, the Student Random Ferns,

The number of Ferns may be smaller than that of the Teacher Random Ferns.

Preferably, in step (5),

By using the learned teacher-student random ferns, the number of ferns may be reduced to track a plurality of pedestrians.

A plurality of pedestrian tracking systems using a teacher-student random fun according to a feature of the present invention for achieving the above object,

As a multiple pedestrian tracking system,

A camera unit for photographing an image including a plurality of pedestrians from a camera installed in a moving vehicle;

A detection unit for detecting a plurality of pedestrians in the image captured by the camera unit;

An extraction unit for extracting a feature value by inputting an image including a plurality of pedestrians detected by the detection unit into a deep network;

A learning unit for learning teacher-student random ferns using the feature values extracted from the extraction unit; And

It is characterized in that it comprises a tracking unit for tracking a plurality of pedestrians using the teacher-student random fern (Teacher-Student Random Ferns) learned in the learning unit.

Preferably, the detection unit,

A classification module for distinguishing pedestrians and non-pedestrians from the image captured by the camera unit; And

It may include a detection module for detecting the pedestrians classified by the classification module as the plurality of pedestrians.

Preferably, the deep network may be a synthetic product neural network.

More preferably, the synthetic product neural network may be tiny YOLO.

Even more preferably, the tiny YOLO,

It may consist of 9 convolution layers, 6 max pooling layers, and 1 fully connected layer.

Preferably, the extraction unit,

A feature value may be extracted for each of a plurality of pedestrians detected by the detector.

Preferably, the learning unit,

A first learning module for learning teacher random ferns using feature values extracted from the extraction unit; And

A second learning module for learning Student Random Ferns by using Teacher Random Ferns learned in the first learning module may be included.

More preferably, the teacher random fern (Teacher Random Ferns),

You can have a plurality of Ferns.

More preferably, the Student Random Ferns,

The number of Ferns may be smaller than that of the Teacher Random Ferns.

Preferably, the tracking unit,

By using the learned teacher-student random ferns, the number of ferns may be reduced to track a plurality of pedestrians.

According to the method and system for tracking a number of pedestrians using a teacher-student random fun proposed in the present invention, feature values of pedestrians are extracted using tiny YOLO, a kind of deep network, and randomized using the extracted feature values. By learning random ferns, real-time learning is possible, thus minimizing mis-tracking caused by changes in shape and size of pedestrians.

In addition, according to the method and system for tracking a number of pedestrians using a teacher-student random fun proposed in the present invention, in order to enable real-time tracking by reducing the number of Ferns, a teacher-student random fun (Teacher-Student) Random Ferns) can be used to quickly and accurately track a large number of pedestrians in real time.

1 is a view showing a flowchart of a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention.
2 is a diagram illustrating a multi-layer perceptron (MLP) network among deep networks.
3 is a diagram showing a detailed flow of step S200 in a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention.
4 is a diagram illustrating a step S210 of a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention.
5 is a diagram illustrating a step S300 of a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention.
6 is a diagram showing a detailed flow of step S400 in a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention.
7 is a diagram illustrating an overall process of learning a teacher random fern in a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention.
FIG. 8 is a diagram illustrating an algorithm for learning student random ferns in a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention.
9 is (a) a state of tracking a number of pedestrians using QuadMOT and (b) tracking a number of pedestrians using a method of tracking a number of pedestrians using a teacher-student random fun according to an embodiment of the present invention. A drawing shown for comparison.
10 is a diagram showing the configuration of a plurality of pedestrian tracking systems using a teacher-student random fun according to an embodiment of the present invention.
11 is a diagram showing a detailed configuration of a detection unit in a plurality of pedestrian tracking systems using a teacher-student random fun according to an embodiment of the present invention.
12 is a diagram showing a detailed configuration of a learning unit in a plurality of pedestrian tracking systems using a teacher-student random fun according to an embodiment of the present invention.

Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art may easily implement the present invention. However, in describing a preferred embodiment of the present invention in detail, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. In addition, the same or similar reference numerals are used throughout the drawings for parts having similar functions and functions.

In addition, throughout the specification, when a part is said to be'connected' to another part, it is not only'directly connected', but also'indirectly connected' with another element in the middle. Include. In addition, "including" a certain component means that other components may be further included rather than excluding other components unless specifically stated to the contrary.

Each step of the method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention may be performed by a computer device. Hereinafter, for convenience of description, the performing subject may be omitted in each step.

1 is a view showing a flowchart of a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention. As shown in Figure 1, the method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention, the step of photographing an image including a plurality of pedestrians from a camera installed in a moving vehicle (S100) , The step of detecting a plurality of pedestrians from the image captured in step S100 (S200), the step of extracting a feature value by inputting the image including the plurality of pedestrians detected in step S200 into the deep network (S300), in step S300 Using the extracted feature values, learning a teacher-student random ferns (S400), and using a teacher-student random ferns learned in step S400 It may be implemented including the step (S500) of tracking a pedestrian.

Hereinafter, before describing each step of a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention, a deep network and a random fern used in the present invention will be first described in detail. Do it.

Artificial Neural Network (ANN) is used in machine learning and cognitive science, and is a statistical learning algorithm inspired by biological neural networks (especially the brain among animals' central nervous systems). The artificial neural network refers to the overall network having problem-solving ability by changing the strength of synaptic bonding through learning by artificial neurons (nodes) that form a network through synaptic bonding. In a narrow sense, it may refer to a multilayer perceptron using error backpropagation, but this is an incorrect usage, and artificial neural networks are not limited thereto.

A deep network or a deep neural network (DNN) is an artificial neural network composed of several hidden layers between an input layer and an output layer. Deep networks, like general artificial neural networks, can model complex non-linear relationships. For example, in a deep network structure for an object identification model, each object may be represented as a hierarchical structure of basic elements of an image, and in this case, additional layers may aggregate features of progressively gathered lower layers. This feature of deep networks makes it possible to model complex data with fewer units than similarly performed artificial neural networks.

FIG. 2 is a diagram illustrating a multi-layer perceptron (MLP) network among deep networks. As shown in FIG. 2, the MLP network is a neural network in which at least one intermediate layer exists between an input layer and an output layer, and an intermediate layer between the input layer and the output layer is called a hidden layer. The network is connected in the direction of the input layer, the hidden layer, and the output layer, and there is no connection within each layer and a direct connection from the output layer to the input layer, which is a feedforward network.

The MLP network has a structure similar to that of a single-layer perceptron, but by making the input/output characteristics of the intermediate layer and each unit nonlinear, the network's capability is improved and various disadvantages of the single-layer perceptron have been overcome. In the MLP network, as the number of layers increases, the characteristics of the crystal regions formed by the perceptron become more advanced. More specifically, in the case of a single layer, the pattern space is divided into two areas, a convex open area or a concave closed area is formed in the case of the second layer, and in the case of the third layer, any type of area can theoretically be formed.

In general, when input data is presented to each unit of the input layer, this signal is converted in each unit, transmitted to the intermediate layer, and finally output to the output layer. The direction of reducing the difference by comparing this output value with the desired output value The MLP network can be trained by adjusting the connection strength.

The convolutional neural network (CNN) is a type of MLP network designed to use minimal preprocessing. A composite product neural network is a neural network composed of one or several convolutional layers, a pooling layer, and a fully connected layer, and has a structure suitable for learning 2D data. Since it can be trained through a backpropagation algorithm, it can be widely used in various application fields such as object classification and object detection in images.

The convolutional layer may serve to extract features from input data. The convolutional layer may consist of a filter that extracts features and an activation function that converts a value extracted from the filter into a nonlinear value.

Synthetic product neural networks can be trained through gradient descent and backpropagation algorithms. At this time, the gradient descent method is an optimization algorithm for finding a first-order approximation, and it is a method of obtaining the slope of the function (inclination) and continuously moving it toward the lower slope until it reaches the extreme value. It refers to a statistical technique that is used, and is a method of adjusting individual weights so that a desired value is output for the same input layer.

Random Ferns, a method proposed by Ozuysal in 2007, is a modification of Bayes' theory. Random Ferns overcomes the limitations of Bayes' theory by considering the correlation between feature functions. In addition, a simple and fast calculation can be performed by implementing a feature function using the difference between two pixels. The performance of Random Ferns has better classification performance than that of Random Tree, and has the same performance as the object recognition rate of SIFT and faster operation speed than SIFT.

Random Ferns can be defined through the following process.

H multiple classes c _z , z = 1,... ,H, and N feature extraction functions f _j , j = 1,... When ,K is defined, class is classified by probability, and can be defined as in Equation 1 below.

Equation 1 may be defined as in Equation 2 below using Bayes definition.

Assuming that P(C) and P(f ₁ ,f ₂ ,...,f _k ) in Equation 2 are predetermined probability values, the multi-class c _z may be defined as in Equation 3 below.

If each feature extraction function is independently considered to obtain P(f ₁ ,f ₂ ,...,f _k ｜C=c _z ) in Equation 3, it can be calculated as in Equation 4 below.

However, in general, since some feature extraction functions show dependent features, in order to consider this, Random Ferns assumes that there is a correlation between feature extraction functions, and combines the feature extraction functions with correlations. It is referred to as a fern. When using random ferns, Equation 4 can be modified as shown in Equation 5 below.

In Equation 5, F _k = (f _a(k,1) ,f _a(k,2) ,... ,f _a(k,S) } = 1,… ,M denotes the k-th ferns, and contains S feature extraction functions. a(k,j) has a range of 1,... As a random permutation function of ,N, S feature extraction functions are randomly selected. Through this, random ferns perform classification using the results of M ferns.

Hereinafter, each step of a method for tracking a plurality of pedestrians using a teacher-student random fun proposed in the present invention using the above-described deep network and random ferns will be described in detail.

In step S100, an image including a plurality of pedestrians may be captured by a camera installed in a moving vehicle. In order for an active intelligent traffic system to achieve an appropriate level of safety, the advanced driver assistance system must track all pedestrians in motion and identify pedestrians who are at risk of entering the road in advance. It is possible to take an image including a number of pedestrians through the camera.

In step S200, a plurality of pedestrians may be detected in the image captured in step S100. 3 is a diagram showing a detailed flow of step S200 in a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention. As shown in Figure 3, step S200 of the method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention, the step of distinguishing a pedestrian and a non-pedestrian in the image taken in step S100 ( S210), and detecting a pedestrian identified in step S210 as a plurality of pedestrians (S220).

4 is a diagram illustrating a step S210 of a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention. As shown in FIG. 4, in step S210, pedestrians and non-pedestrians may be distinguished from the image captured in step S100. At this time, the pedestrian can be a person, and the non-pedestrian can be a power pole, a tree, or a building.

In step S220, the pedestrian identified in step S210 may be detected as a plurality of pedestrians. In the method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention, since the feature values of each pedestrian must be extracted by detecting a plurality of pedestrians and inputting them to a deep network, in step S220, A pedestrian identified in S210 may be detected as a plurality of pedestrians, and the feature values of each pedestrian may be extracted by inputting it into the deep network of step S300 described below.

5 is a diagram illustrating a step S300 of a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention. As shown in FIG. 5, in step S300 of a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention, an image including a plurality of pedestrians detected in step S200 is input to a deep network. Thus, the feature value can be extracted.

More specifically, in step S300 of the method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention, tiny YOLO, a kind of synthetic multiplication neural network, is used as a deep network, in step S200. A feature value for each pedestrian may be extracted from an image including a plurality of detected pedestrians.

Tiny YOLO may be composed of 9 convolution layers, 6 max pooling layers, and 1 fully connected layer. At this time, the characteristic value of the pedestrian can be extracted through the fully connected layer, which is the last layer of tiny YOLO.

In step S400, a teacher-student random ferns may be learned using the feature values extracted in step S300. 6 is a diagram showing a detailed flow of step S400 in a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention. As shown in FIG. 6, step S400 of the method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention includes a teacher random fun using the feature values extracted in step S300. ) Learning (S410), and learning a Student Random Ferns using the Teacher Random Ferns learned in step S410 (S420).

Teacher-Student Random Ferns is a tracker made up of Random Ferns. Teacher Random Ferns, because it is constructed based on a large amount of learning data, has high tracking performance, but it may be difficult to track pedestrians in real time due to a slow tracking speed. Accordingly, a method for tracking a number of pedestrians using a teacher-student random fun according to an embodiment of the present invention is a fun method while maintaining the tracking performance of a teacher random fern using a student random fern (Student Random Ferns). By reducing the number of (Ferns), pedestrians can be tracked faster and more accurately than before.

FIG. 7 is a diagram illustrating an overall process of learning a teacher random fern in a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention. As shown in FIG. 7, in a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention, a Teacher Random Ferns detects and detects a plurality of pedestrians in step S200. An image including a plurality of pedestrians may be input to the deep network in step S300 and learned using the extracted feature values. In this case, the Teacher Random Ferns may have a plurality of Ferns, for example, 1 to L (L is a natural number) and L Ferns.

In step S410, teacher random ferns may be learned using the feature values extracted in step S300. More specifically, teacher random ferns may be trained using tiny YOLO, one of the characteristic values of the pedestrians extracted in step S300 and the synthetic product neural network.

In step S420, student random ferns may be learned using the teacher random ferns learned in step S410. More specifically, using the Teacher Random Ferns learned through step S410, it is possible to learn by dividing whether a pedestrian appears first or more than once.

FIG. 8 is a diagram illustrating an algorithm for learning a student random fern in a method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention. Through the algorithm of FIG. 8, in the method for tracking a number of pedestrians using a teacher-student random fun according to an embodiment of the present invention, a student random fern when a pedestrian first appears can be learned. At this time, the above algorithm may be repeated as many as the number of pedestrians detected through step S200 to learn Student Random Ferns.

When a pedestrian appears more than once, if the pedestrian detected in the current frame and the pedestrian learned by the Student Random Ferns in the previous frame match, data connection is performed between the two frames, and the Student Random Ferns Random Ferns) has been updated and can be learned.

As shown in the last part of the algorithm of FIG. 8, the number of Ferns in Student Random Ferns in a plurality of pedestrian tracking methods using a teacher-student random fun according to an embodiment of the present invention is May be less than Teacher Random Ferns.

In step S500, a plurality of pedestrians may be tracked using a teacher-student random fern learned in step S400. More specifically, in step S500 of the method for tracking a plurality of pedestrians using a teacher-student random fun according to an embodiment of the present invention, a teacher-student random fern learned in step S400 is used. You can track a large number of pedestrians by reducing the number of ferns.

9 is (a) a state of tracking a number of pedestrians using QuadMOT and (b) tracking a number of pedestrians using a method of tracking a number of pedestrians using a teacher-student random fun according to an embodiment of the present invention. It is a drawing shown to compare appearances. At this time, if the student random ferns are determined to be the same pedestrians in the subsequent frame, they are displayed as boxes of the same color. Referring to (a) of FIG. 9, when a number of pedestrians are tracked using QuadMOT, if the camera movement is large or pedestrians overlap each other, the pedestrian tracking is omitted (yellow arrow in the fourth image). , There is a problem tracking other pedestrians (red arrow in the third video).

However, looking at (b) of FIG. 9, in the case of tracking pedestrians using a plurality of pedestrian tracking methods using a teacher-student random fun according to an embodiment of the present invention, a student random fun (Student Random Ferns) is By newly updating and learning, the phenomenon of missing pedestrian tracking or tracking other pedestrians in the middle does not occur.

10 is a diagram showing the configuration of a plurality of pedestrian tracking system 10 using a teacher-student random fun according to an embodiment of the present invention. As shown in FIG. 10, a plurality of pedestrian tracking systems 10 using a teacher-student random fun according to an embodiment of the present invention includes a camera unit 100, a detection unit 200, and an extraction unit 300 , It may be configured to include a learning unit 400 and a tracking unit 500.

More specifically, a plurality of pedestrian tracking systems 10 using a teacher-student random fun according to an embodiment of the present invention, a camera unit 100 for photographing an image including a plurality of pedestrians from a camera installed in a moving vehicle. ), a detection unit 200 that detects a plurality of pedestrians from an image captured by the camera unit 100, an image including a plurality of pedestrians detected by the detection unit 200 is input to a deep network to extract feature values The extraction unit 300, the learning unit 400 for learning teacher-student random ferns using the feature values extracted from the extraction unit 300, and the teacher learned in the learning unit 400 -It may be configured to include a tracking unit 500 for tracking a plurality of pedestrians using a student random fun (Teacher-Student Random Ferns).

11 is a diagram showing a detailed configuration of a detection unit in a plurality of pedestrian tracking system 10 using a teacher-student random fun according to an embodiment of the present invention. As shown in FIG. 11, in the plurality of pedestrian tracking systems 10 using a teacher-student random fun according to an embodiment of the present invention, the detection unit 200 is in the image captured by the camera unit 100 It may be configured to include a classification module 210 that distinguishes between pedestrians and non-pedestrians, and a detection module 220 that detects pedestrians separated by the classification module 210 as the plurality of pedestrians.

12 is a diagram showing a detailed configuration of the learning unit 400 in a plurality of pedestrian tracking system 10 using a teacher-student random fun according to an embodiment of the present invention. As shown in FIG. 12, in a plurality of pedestrian tracking systems 10 using a teacher-student random fun according to an embodiment of the present invention, the learning unit 400 includes feature values extracted from the extraction unit 300 Using the first learning module 410 for learning Teacher Random Ferns, and the Teacher Random Ferns learned in the first learning module 410, Student Random Ferns Ferns) may be configured to include a second learning module 420 for learning.

For a plurality of pedestrian tracking systems 10 using a teacher-student random fun according to an embodiment of the present invention, since it has been sufficiently described in relation to a method for tracking a plurality of pedestrians using a teacher-student random fun, detailed description will be omitted. To

As described above, according to the method and system 10 for tracking a number of pedestrians using a teacher-student random fun proposed in the present invention, feature values of pedestrians are extracted using tiny YOLO, a type of deep network, By learning Random Ferns using the extracted feature values, real-time learning is possible, and mis-tracking due to change in shape and size of pedestrians can be minimized. In addition, according to the present invention, in order to enable real-time tracking by reducing the number of Ferns, a number of pedestrians can be quickly and accurately tracked in real time by using a Teacher-Student Random Ferns. have.

The present invention described above can be modified or applied in various ways by those of ordinary skill in the technical field to which the present invention belongs, and the scope of the technical idea according to the present invention should be determined by the following claims.

10: Multiple pedestrian tracking system
100: camera unit
200: detection unit
210: division module
220: detection module
300: extraction unit
400: learning department
410: first learning module
420: second learning module
500: tracking unit
S100: Step of capturing an image including a plurality of pedestrians from a camera installed in a moving vehicle
S200: detecting a plurality of pedestrians in the image captured in step S100
S210: Step of distinguishing a pedestrian and a non-pedestrian in the image taken in step S100
S220: Step of detecting a pedestrian identified in step S210 as a plurality of pedestrians
S300: Step of extracting feature values by inputting an image including a plurality of pedestrians detected in step S200 into a deep network
S400: Learning a Teacher-Student Random Ferns using the feature values extracted in step S300
S410: Learning a Teacher Random Ferns using the feature values extracted in step S300
S420: Learning a Student Random Ferns using the Teacher Random Ferns learned in Step S410
S500: Step of tracking a number of pedestrians using the teacher-student random ferns learned in step S400

Claims (20)

As a number of pedestrian tracking methods,
(1) taking an image including a plurality of pedestrians from a camera installed in a moving vehicle;
(2) detecting a plurality of pedestrians in the image captured in step (1);
(3) extracting feature values by inputting an image including a plurality of pedestrians detected in step (2) into a deep network;
(4) learning a Teacher-Student Random Ferns using the feature values extracted in step (3); And
(5) including the step of tracking a plurality of pedestrians using the teacher-student random ferns learned in step (4),
In step (3),
Extracting feature values for each of the plurality of pedestrians detected in step (2),
The step (4),
(4-1) learning a Teacher Random Ferns having a plurality of Ferns by using the feature values extracted in Step (3); And
(4-2) Using the Teacher Random Ferns learned in step (4-1), the Student Random Ferns with a smaller number of Ferns than the Teacher Random Ferns Random Ferns),
In step (4-2),
Using the Teacher Random Ferns learned through the above step (4-1), the case of the pedestrian appearing first or more than once is divided and learned,
When a pedestrian appears more than once, if the pedestrian detected in the current frame and the pedestrian learned by the Student Random Ferns in the previous frame match, data connection is performed between the two frames, and the Student Random Ferns Random Ferns) is newly updated and learned,
Using the Student Random Ferns, it is possible to track pedestrians faster and more accurately than before by reducing the number of Ferns while maintaining the tracking performance of the Teacher Random Ferns.
In step (5),
Using the learned teacher-student random fun (Teacher-Student Random Ferns), characterized in that to track a plurality of pedestrians by reducing the number of ferns, a number of pedestrian tracking method using a teacher-student random fun.
The method of claim 1, wherein the step (2),
(2-1) distinguishing between pedestrians and non-pedestrians in the image captured in step (1); And
(2-2) A method of tracking a plurality of pedestrians using a teacher-student random fun, comprising the step of detecting the pedestrians identified in step (2-1) as the plurality of pedestrians.
The method of claim 1, wherein the deep network in step (3),
A method of tracking multiple pedestrians using a teacher-student random fun, characterized in that it is a synthetic product neural network.
The method of claim 3, wherein the composite product neural network,
Tiny YOLO, characterized in that, a number of pedestrian tracking method using a teacher-student random fun.
The method of claim 4, wherein the tiny YOLO,
A method for tracking multiple pedestrians using a teacher-student random fun, characterized by consisting of 9 convolution layers, 6 max pooling layers, and 1 fully connected layer .
delete delete delete delete delete As a plurality of pedestrian tracking system 10,
A camera unit 100 for capturing an image including a plurality of pedestrians from a camera installed in a moving vehicle;
A detection unit 200 for detecting a plurality of pedestrians in the image captured by the camera unit 100;
An extraction unit 300 for extracting a feature value by inputting an image including a plurality of pedestrians detected by the detection unit 200 into a deep network;
Including a tracking unit 500 for tracking a plurality of pedestrians using the teacher-student random fun (Teacher-Student Random Ferns) learned in the learning unit 400,
The extraction unit 300,
Extracting feature values for each of the plurality of pedestrians detected by the detection unit 200,
The learning unit 400,
A first learning module 410 for learning teacher random ferns having a plurality of ferns using the feature values extracted by the extraction unit 300; And
Using Teacher Random Ferns learned in the first learning module 410, a Student Random Ferns with a smaller number of Ferns than the Teacher Random Ferns Including a second learning module 420 to learn,
The second learning module 420,
Using the teacher random ferns learned through the first learning module 410, the first appearance of the pedestrian and the case of appearing more than two times are divided and learned,
When a pedestrian appears more than once, if the pedestrian detected in the current frame and the pedestrian learned by the Student Random Ferns in the previous frame match, data connection is performed between the two frames, and the Student Random Ferns Random Ferns) is newly updated and learned,
Using the Student Random Ferns, it is possible to track pedestrians faster and more accurately than before by reducing the number of Ferns while maintaining the tracking performance of the Teacher Random Ferns.
The tracking unit 500,
Using the learned teacher-student random fun (Teacher-Student Random Ferns), characterized in that to track a plurality of pedestrians by reducing the number of ferns, a plurality of pedestrian tracking system using a teacher-student random fun.
The method of claim 11, wherein the detection unit 200,
A classification module 210 that distinguishes between pedestrians and non-pedestrians in the image captured by the camera unit 100; And
A plurality of pedestrian tracking systems using a teacher-student random fun, characterized in that it comprises a detection module 220 that detects the pedestrians classified by the classification module 210 as the plurality of pedestrians.
The method of claim 11, wherein the deep network,
A plurality of pedestrian tracking systems using a teacher-student random fun, characterized in that the synthetic product neural network.
The method of claim 13, wherein the composite product neural network,
Tiny YOLO, characterized in that, a number of pedestrian tracking system using a teacher-student random fun.
The method of claim 14, wherein the tiny YOLO,
Multiple pedestrian tracking system using teacher-student random fun, characterized by consisting of 9 convolution layers, 6 max pooling layers, and 1 fully connected layer .
delete delete delete delete delete

Images