KR101958270B1

KR101958270B1 - Intelligent Image Analysis System using Image Separation Image Tracking

Info

Publication number: KR101958270B1
Application number: KR1020150171562A
Authority: KR
Inventors: 최명길
Original assignee: 홍진수
Priority date: 2015-12-03
Filing date: 2015-12-03
Publication date: 2019-03-14
Also published as: KR20170065301A

Abstract

The present invention relates to an intelligent image analysis system using image segmentation image tracking, in which intelligent image analysis using image segmentation image tracking, which analyzes and captures an image of a captured image, three-dimensionally analyzes and tracks an object existing in the image System can be provided.

Description

TECHNICAL FIELD The present invention relates to an intelligent image analysis system using image separation image tracking,

[0001] The present invention relates to an intelligent image analysis system using image segmentation image tracking, and more particularly, to an intelligent image analysis system using image segmentation image tracking, And an intelligent image analysis system using the same.

The present invention relates to an intelligent image analysis system using image segmentation image tracking.

Generally, CCTV is installed in various places around our life and it gives many benefits to make a safe life.

In particular, the development speed of CCTV is being promoted by the development of the Internet and the IT infrastructure, and it is preventing not only life crime but also crime prevention at home, school, kindergarten, hospital, Is also demonstrating its value.

In recent years, it has become an important means of responding to terrorism that is taking place all over the world.

However, in spite of this usefulness, real-time monitoring is limited.

There is a problem in that a monitoring agent for monitoring the CCTV is not only required but also the concentration of the monitoring agent is lowered with the lapse of time and the situation in which an event such as an accident or a crime occurs can not be detected in real time.

SUMMARY OF THE INVENTION The present invention has been made in an effort to solve the problems as described above, and an object of the present invention is to provide a method and apparatus for analyzing an object in an image, And to provide an image analysis system.

Another object of the present invention is to provide a method and apparatus for tracking an image in an image by three-dimensionally analyzing the image, analyzing the image, capturing the image in real time, And an intelligent image analysis system using the same.

According to an aspect of the present invention, there is provided an image processing apparatus including an image input unit for receiving input image data and then pre-processing the image data by applying a Gaussian filter;
A background data detector for detecting an input image data received by the image input unit as background data when an image is divided into parts and a velocity vector in each part is smaller than a predetermined threshold value;
A position table of a portion of the input image data received by the image input unit

The velocity vector at

An object data detector for generating object data when the calculated velocity value is greater than a threshold value set by the velocity vector;
A locus data generation unit for detecting the object data generated by the object data detection unit in the two-dimensional input image data to calculate an object size, a moving distance, and a moving direction, and generating the object object data as three-dimensional object locus data; And
An object tracking module for tracking an object through object trajectory data generated by the trajectory data generator;
And a monitoring unit for comparing the object locus data generated by the locus data generator with the preset protection zone data.

In addition, the object data detecting unit excludes background data from the input image data and detects the object.

In addition, the locus data generation unit includes a noise removal module that removes noise of object data using an image filter when analyzing object data.

The locus data generation unit generates new object data when the moving distance of the object is greater than a predetermined threshold value.

In addition, the locus data generation unit analyzes the optical flow of the object to generate object locus data.

As described above, according to the present invention, it is possible to provide an intelligent image analysis system using image segmentation image tracking in which a captured image is received, the image is analyzed, and objects existing in the image are analyzed in three dimensions and tracked in real time .

In addition, according to the present invention, there is provided an intelligent tracking system for tracking an object existing in an image by analyzing the received image, analyzing the object in three dimensions, real-time tracking, An image analysis system can be provided.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is an exemplary diagram illustrating a three-dimensional analysis of a two-dimensional image according to an embodiment of the present invention; FIG.
2 is an exemplary diagram illustrating an analysis of an object by analyzing optical flow of image data according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The advantages and features of the present invention and the manner of achieving them will become apparent with reference to the embodiments described in detail below with reference to the accompanying drawings.

The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. To fully disclose the scope of the invention to those skilled in the art, and the invention is only defined by the scope of the claims. Like reference numerals refer to like elements throughout the specification.

Hereinafter, the present invention will be described with reference to the drawings for explaining an intelligent image analysis system using image segmentation image tracking according to embodiments of the present invention.

An intelligent image analysis system using image segmentation image tracking according to the present invention includes an image input unit for receiving input image data, a background data detection unit for detecting background data in input image data received by the image input unit, An object data detecting unit for detecting the object data from the image data and detecting the object data generated by the object data detecting unit in the two-dimensional input image data to calculate the size, the moving distance, and the moving direction of the object, An object tracing module for tracing an object through the object trajectory data generated by the trajectory data generating unit, and an object tracing module for generating the object trajectory data generated by the trajectory data generating unit, And a monitoring unit for comparing with zone data It shall be.

In the intelligent image analysis system using the image segmentation image tracking according to the present invention, the object data detection unit excludes background data from the input image data and detects an object.

In the intelligent image analysis system using image segmentation image tracking according to the present invention, the trajectory data generator includes a noise elimination module for removing noise of object data using an image filter when analyzing object data.

In the intelligent image analysis system using image segmentation image tracking according to the present invention, the trajectory data generator generates new object data when the moving distance of the object is greater than a preset threshold value.

In the intelligent image analysis system using the image segmentation image tracking according to the present invention, the locus data generation unit analyzes object optical flow to generate object locus data.

FIG. 1 is an exemplary diagram illustrating a three-dimensional analysis of a two-dimensional image according to an exemplary embodiment of the present invention. FIG. 2 illustrates an example of analyzing an object by analyzing an optical flow of image data according to an exemplary embodiment of the present invention. It is an example.

There is a method of detecting moving objects through image tracking in 3D space and a background difference method.

When detecting a moving object, there are various defects such as detection error, movement of moving object, difficulty of direction determination.

The present invention exploits a method of detecting motion regions by applying background subtraction methods in combination.

By applying these two methods, it is possible to adaptively extract objects that are not sensitive to changes in illumination and have fast or slow movements.

In the preprocessing process, a Gaussian filter is applied to obtain a clean image.

The motion object region is detected by applying the proposed method from the noise-free image.

As a background modeling method, it is modeled and updated by MOG (Mixture Of Gaussian) method to adapt to real time background change.

This program introduces an optical flow technique that enables fast detection of moving objects and at the same time determines the direction of movement.

Calculation of Optical Flow Field (Speed Field)

Optical flow is defined as a noticeable movement of image brightness.

In the image column

In the image of vision

The following two assumptions can be made when displaying the brightness of a pixel in a position.

1. Brightness

Within a large part of the image,

It is slick about. (Possibility of continuous differentiation)

2. The brightness of each point of a moving or stopping object does not change each time (constant with respect to time). Any object in the image or any point in one object

After sight

Brightness when moved to position

Taylor's deployment is as follows.

At this time,

, And defined by equation (4), the following equation (4), called optical flow constraint equation , is obtained. here

,

Respectively, of the optical flow field

,

Ingredients.

Generally, Equation 3 has more than one solution, so additional conditions are needed to solve this equation.

In this development, well-known Lucas & Kanade method is applied to calculate the optical flow field .

The Lucas & Kanade method divides the image into smaller parts and assumes that the velocity vector (optical flow vector) at each part is constant.

The solution of the optical flow constraint equation then results in the solution of the following 2x2 linear system by applying the least squares method.

In Equation (5)

Is a Gaussian window.

The Gaussian window can be represented by a combination of two separate binomial kernels.

Detection of the foreground / background using the Lucas & Kanade method

The velocity vector at

And then detects the moving object in the image.

The detection of the moving object is carried out from the optical flow field (or velocity field)

The magnitude of the velocity vector

Obtain a scalar field with

Next, a threshold value is adaptively derived from the average velocity magnitudes of the previous two frames and the current frame, and then a binary value matrix is obtained for the velocity field color field.

Dilation and erosion processing is performed on the obtained two-valued matrix and filtering processing is performed to obtain a foreground of moving objects.

Finally, calculate the speed of movement of objects from the velocity vectors of the foregrounds.

As a result, you get the foreground and moving velocity vectors of the moving objects.

As a result of the above, the motion target can be detected.

After detecting motion objects, it tracks the motion of the objects.

Tracking of motion objects is done by

Th frame

Th frame are expressed by Equations (6) and (7), respectively.

here

and

Respectively

and

The number of motion objects detected in the ith frame.

In addition, the velocity vector of each motion object is expressed by Equation (8) and Equation (9).

the problem is

Each object in the < RTI ID = 0.0 >

Is the target of the ith frame or the object of the new appearance.

In the present invention, the positional relationship, the magnitude, and the moving velocity vector of the objects are used for object detection.

This means that the moving direction of the moving objects does not change suddenly in a real situation.

There is a certain limit to the movement speed of motion objects. In other words, the distance of the moving object between two neighboring frames can not be larger than a certain value.

It is based on the assumption that the variance of the foreground area of the track being tracked is not very large.

Actual

Each target in the ith frame

Tracking in the following sequence.

Frame

The closest distance to

And his distance

.

if

Is greater than a predetermined threshold value

If

And ends the trace.

And determined from above

The difference between the areas of the threshold values

If larger

And ends the trace.

Finally

And determined from above

Are compared with each other. For the comparison of the direction of movement, the angle between the velocity vectors of the two objects is used.

At this time

The threshold value < RTI ID = 0.0 >

If greater than

If this is not the case,

end

Th frame

As shown in FIG.

Determination method of tracking object considering moving direction and moving distance.

First of all, the directionality and the moving distance are examined only for those whose history number exceeds the predetermined threshold value in the temporary target list. If the condition is satisfied, it is confirmed as a target of tracking.

An average moving distance, an average moving distance in the Y axis direction, an average moving direction angle Deviation vehicle How to obtain.

The total sum of the distances between the first and subsequent points, the sum of the moving distances in the Y-axis direction, and the sum of the angular deviations are obtained with the history information for the temporary object.

Next, we divide these values by the number considered and obtain the averages, respectively, and use the general Euclidean distance as the travel distance.

That is, Equation 13 is obtained.

The moving distance in the Y-axis direction

.

That is, in Equation 14,

Remember that absolute values are not absolute distance values because they do not reflect actual mobility.

The directional angle deviation is expressed as an angle between two vectors.

That is, it is obtained by the following equation (15).

In Equation (15)

Is the vector between the first and last points,

: Represents a vector between the first and subsequent points.

Mobility test with three features.

First, it is assumed that a block that is not the Y-axis moving direction is moved when the expression (16) is satisfied.

Also, for a block moving in the Y-axis direction, although the movement distance is small (however, it must be larger than a certain threshold value), it is regarded as a moving object if the number of history is significantly increased.

That is, Equation 17 is obtained.

In the case of the present invention, the above four threshold values

Is dynamically determined according to the number of temporary target histories and the engine image size.

By separating the foreground, background, and difference of the input image into frames by frame and dynamically discriminating the difference between each frame, the object in the image can be tracked.

.

Claims

An image input unit for receiving input image data and then pre-processing the image data by applying a Gaussian filter;
A background data detector for detecting an input image data received by the image input unit as background data when an image is divided into parts and a velocity vector in each part is smaller than a predetermined threshold value;
A position table of a portion of the input image data received by the image input unit

The velocity vector at

An object data detector for generating object data when the calculated velocity value is greater than a threshold value set by the velocity vector;
A locus data generation unit for detecting the object data generated by the object data detection unit in the two-dimensional input image data to calculate an object size, a moving distance, and a moving direction, and generating the object object data as three-dimensional object locus data; And
An object tracking module for tracking an object through object trajectory data generated by the trajectory data generator;
And a monitoring unit for comparing the object sign data generated by the sign data generating unit with the protection zone data, which is a predetermined protection zone.

◈ Claim 2 is abandoned due to payment of registration fee.

2. The apparatus of claim 1, wherein the object data detector comprises:
And an object is detected by excluding background data from the input image data.

◈ Claim 3 is abandoned due to the registration fee.

The apparatus according to claim 1,
And a noise removal module for removing noise of object data by using an image filter when analyzing object data.

◈ Claim 4 is abandoned due to the registration fee.

The apparatus according to claim 1,
And generating new object data when the moving distance of the object is larger than a preset threshold value.

◈ Claim 5 is abandoned due to the registration fee.

The apparatus according to claim 1,
And analyzing the optical flow of the object to generate object trajectory data.