KR100882509B1 - Apparatus and Method for image based-monitoring elderly people with Principal Component Analysis - Google Patents

Abstract

The present invention relates to an elderly motion image monitoring apparatus and method using a principal component element method to determine the emergency situation by independently analyzing the movement of the elderly without being affected by the angle of the camera and the size of the silhouette of the person to be detected.

The elderly motion image monitoring method according to the present invention comprises: pre-processing (S100) outputting a binary image by separating a silhouette of a person from a background of an image from image information which is captured by a surveillance camera and converted into a digital signal; PCA computational process that analyzes the posture change as a parameter by changing the proportions of the eigenvectors of the silhouette of the person by applying the PCA algorithm to the information of the silhouette of the person separated in the image preprocessing step (S100) (S100) step; An emergency situation determination step (S300) of determining whether a sudden change is detected in the ratio of an eigenvector of the silhouette of the person to determine an emergency situation; An emergency situation notification step of notifying an emergency situation to a wireless terminal to a doctor or a family when an emergency situation is determined in the emergency situation determination step (S300); Characterized in having a.

The PCA calculation process (S200) may include: a data set input step (S205) of inputting data of a two-dimensional set of x and y coordinates of an image output in the image preprocessing step (S100); An average value calculating step (S213) of obtaining a mean value from each of the data of the x coordinate and the data of the y coordinate input in the data set input step (S205); Covariance matrix having variance of x-coordinate, variance of y-coordinate, and covariance of x- and y-coordinates based on each mean value of data of x-coordinate and y-coordinate obtained in the average value calculating step (S213) a dispersion operation step S217 for obtaining a covariance matrix; Eigenvalues are obtained for the covariance matrix obtained in the variance operation step (S217). Since the image is two-dimensional, two eigenvalues are obtained, and eigenvalues and eigenvector calculations are also obtained. Step S230; A PCA output step (S250) of outputting a result of detecting two eigenvalues and an eigenvector of a silhouette of a person present in the two-dimensional image in the eigenvalue and eigenvector calculation step (S230); Is made of.

Principal Component Analysis, PCA, Elderly, Monitoring, Camera, Motion

Description

Apparatus and Method for image based-monitoring elderly people with Principal Component Analysis}

1 is a conceptual diagram schematically illustrating an elderly motion image monitoring apparatus using a principal component method according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram for schematically explaining an elderly motion image monitoring method using a principal component method according to an exemplary embodiment of the present invention.

3 is a flowchart for explaining an image preprocessing step of FIG. 2.

Figure 4 is an explanatory diagram for explaining the application of the principal component (PCA) for posture analysis of the elderly according to an embodiment of the present invention.

5 is a table summarizing the flow of the principal component analysis method applied to the fall monitoring of the elderly in FIG.

6 is a sequence diagram for explaining the PCA calculation processing step of FIG.

7 is a schematic sequence diagram of an elderly motion image monitoring method using a principal component method according to an embodiment of the present invention.

FIG. 8 illustrates an example of analyzing a person's falling posture change by applying the PCA operation processing step of FIG. 6.

The present invention relates to an elderly motion image monitoring apparatus and method using a principal component element method to determine the emergency situation by independently analyzing the movement of the elderly without being affected by the angle of the camera and the size of the silhouette of the person to be detected.

Recently, as the proportion of the elderly population increases, many elderly people live alone at home, and these elderly people sometimes experience dangerous situations due to sudden health problems or external factors in their homes. Therefore, the importance of the Personal Emergency Response System, which systematically manages and monitors the continuous health status of the elderly, is increasing. Such a system requires accurate judgment of the emergency situation.

Personal health monitoring system can be classified into three categories. First, there is a method of directly checking the health status of a doctor or a nurse, and second, a radio frequency detector (RFID), an accelerometer, a pressure sensor, etc. There is a system that monitors bio signals and walking status by wearing sensors. Third, there is a video-based surveillance system using video equipment such as infrared and CCTV cameras. The first method is a system in which a person (doctor or nurse) directly monitors the health status, but this may cause a problem of waste and shortage of medical personnel. The second method, a surveillance system that attaches a sensor to a person's body, requires additional equipment for attaching the sensor to the body and transmitting information, and the user must always wear a sensor to monitor the state of health. In addition to the inconvenience of the operation of the equipment and the operation of the equipment to prevent the malfunction of the disadvantage that is required. However, the third method, an image based system, can compensate for these disadvantages.

Therefore, an image-based system is required as a personal health surveillance system for the elderly.

In particular, there is a need for an elderly motion image monitoring apparatus that performs image processing on image data input on a computer and automatically determines the movement of a person, that is, monitors the movement of the elderly on an image basis to determine an emergency situation.

However, there is a shortcoming in the conventional image-based home-monitoring system. Image-based vision systems can be very sensitive to results, depending on the camera's angle and motion, the distance to the object being monitored, and the brightness of the light.

Due to these problems, in case of motion estimation of a person using an input video signal, a preprocessing process for correcting the size of a person's silhouette is performed. In this case, the distance between the camera and the person becomes an important factor. Also, depending on the camera's point of view or angle of camera (camera phase), it can be confusing whether a person's posture is standing, falling, or shrunk.

Accordingly, the present invention provides an elderly motion image surveillance apparatus and method that compensates for these disadvantages. To this end, the present invention uses eigenvectors and eigenvalues of the Principal Component Analysis (hereinafter referred to as PCA) method, such as a person falling down independently of the distance and angle of the monitoring camera. Detect changes in posture.

The present invention does not limit activity compared to the problems of conventional image-based home-monitoring systems and has solutions for automated monitoring. In addition, the present invention applies an image processing and a PCA algorithm to generally provide a surveillance camera. Algorithms provide an algorithm for judging emergencies such as falls by independently analyzing the movements of the elderly without being affected by the limited conditions of the cameras, that is, the angle of the camera and the size of the detected human silhouette.

Briefly, the mathematical concept of the PCA used in the present invention is as follows.

PCA requires a covariance matrix of data in a statistical manner. To find the covariance matrix, we first need to find the mean and variance of the data. The average of the data is as Equation 1, and the variance can be expressed as Equation 2.

는 평균이다. Where X _i is input data, n is the number of input data,

Is the average.

Where s is the table deviation and s ² is the variance.

Since standard deviations and variances have meaning only in one dimension, we have to find the variance of the data in each other dimension independently of the data group. However, in the same way, we can obtain the information of variance in relations of different dimensions based on data of different dimensions, which is a covariance matrix. Covariance can be expressed as the variance of the data in each dimension and the variance of the data in different dimensional relationships. This meaning can be understood through Equations 3 and 4 below.

이고 Y_i의 평균은

이다.Where X _i and Y _i are input data of different dimensions, and the mean of X _i is

And the mean of Y _i is

to be.

In Equation 3 and Equation 4, it is expressed as the product of variance in each dimension. As shown in Equation 4, the covariance of the dimension of the same data group is the same as the variance.

The covariance matrix may be expressed as in Equation 5.

를 수학식 6을 이용하여 구한다. 만약 2차원이라면 2개의 λ가 획득될 것이며 이에 해당하는 각각의 벡터들이 2개를 얻을 수 있다.The present invention calculates the eigenvalues λ and eigenvectors by calculating this covariance matrix.

Is obtained using Equation 6. If it is two-dimensional, two λ will be obtained and two corresponding vectors can be obtained.

One of the characteristics of Principal Component Analysis is that for data groups distributed over the same time, vectors of directions with large variance can be obtained. As described above, an eigenvalue and corresponding eigenvectors can be obtained by matrix calculation of a covariance matrix. A vector with a higher eigenvalue becomes an important element of the data group, and a vector with a smaller eigenvalue is less important than the first vector.

The technical problem to be achieved by the present invention is to provide an elderly motion image monitoring apparatus and method for analyzing the movement of the elderly and determining the emergency situation by performing automated monitoring without restricting the activity.

Another technical problem to be achieved by the present invention is to apply image processing and PCA algorithms, without being affected by the limited conditions of Surveillance Cameras in general, that is, the angle of the camera and the size of the silhouette of the detected person. An object of the present invention is to provide an apparatus and a method for monitoring the elderly motion image using the principal component method for independently analyzing the movement of the elderly.

Hereinafter, the structure and operation of the elderly motion image monitoring apparatus and method using the principal component method according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a conceptual diagram for schematically illustrating an elderly motion image monitoring apparatus using a principal component method (PCA) according to an embodiment of the present invention, the surveillance camera unit 100, the monitoring operation processing unit 200, the central control unit 700, the output unit 800.

Surveillance camera unit 100 is composed of a plurality of surveillance cameras, and transmits the image of the elderly to the monitoring operation processing unit 200. An infrared camera or a CCTV camera may be used as the surveillance camera.

The monitoring operation processor 200 processes the image received from the surveillance camera unit 100 and transmits the data to the central controller 700.

Here, the data processing may be transmitted to the central control unit 700 by compressing the image data, and may be transmitted to the central control unit 700 by analyzing the elderly moving image using the principal component method. That is, the monitoring operation processing unit 200 is an image preprocessing unit (not shown) which separates the silhouette of a person from the background of the image in the input image information which is captured by the surveillance camera unit 100 and converted into a digital signal. A posture change analysis unit that analyzes the posture change by applying the change of the ratio (i.e. eigenvalues) of the magnitude of the eigenvector of the person's silhouette by applying the principal component method (PCA) algorithm to the information of the separated person's silhouette. (Not shown) may be provided. Analysis of the elderly motion image using the principal component method may be performed by the monitoring operation processor 200 or may be performed by the central controller 700.

The central control unit 700 transmits the information received from the monitoring operation processor 200 to the output unit 800. That is, the central control unit 700 transmits the information received from the monitoring operation processor 200 to the doctor's computer of the hospital, and informs the doctor or family's mobile phone or PDA in case of an emergency.

The output unit 800 consists of a doctor's computer, a doctor's or family's mobile phone or PDA, and outputs data received from the central control unit 700. That is, the output unit 800 transmits the information received from the central control unit 700 to the doctor's computer of the hospital, and in the case of an emergency, it informs the emergency of the doctor or family's mobile phone or PDA or wireless terminal.

FIG. 2 is an explanatory diagram for schematically illustrating a method for monitoring a motion of an elderly man using the principal component method according to an exemplary embodiment of the present invention.

In the image preprocessing step S100, the silhouette of the person is separated from the background of the image in the input image information which is captured by the camera and converted into a digital signal.

The posture change analysis step (S200) is to apply the principal component factor method (PCA) algorithm to the information of the silhouette of the person separated in the image preprocessing step (S100) to change the proportions of the eigenvectors of the silhouette of the person as parameters. Analyze posture changes.

If a sudden change is detected to determine whether it is an emergency (S300), if it is an emergency, inform the doctor or family, such as a mobile phone, a wireless terminal that the emergency.

FIG. 3 is a flowchart for explaining an image preprocessing step of FIG. 2.

The captured image step S105 converts the image output from the surveillance camera into a digital signal.

The median filtering step (S110) randomly inputs the image input from the captured image operation step (S105) such as salt and peppers noise generated during the transmission and coding of the image through a 5 × 5 median filter. Remove the noise.

Conventionally, the average filter is mainly used. If the noise is removed by the average filter, the influence of the noise greatly affects the surrounding area, resulting in loss of image information. Thus, the present invention uses median filters with nonlinear features. The median filter is a filter suitable for signals having nonlinear characteristics. It is applied to an image for the first time to remove noise, and at the same time, it helps edge detection and area classification by the effect of smoothing.

The difference image acquisition step S120 performs a process of separating the person moving in the background from the output of the median filtering step S110. The background model is constructed by obtaining the average value of 10 frames of the background image when a person does not enter the camera's shooting range. The present invention uses a method of separating a person from the background by a difference value from the time when a person enters an image with a pre-built background model. At the same time, the image difference of each frame is obtained to detect a person together. In the method of image difference of a frame, boundary information is detected between frames in which a human motion is performed. The difference between the background model and the frame was detected by people from the background. This is shown in Equation 7.

Human Detection = Difference in Background Model + Difference Between Frames

In the erosion operation step S130, an erosion calculation process is performed by applying a 6 × 6 filter of morphological processing.

The Gaussian Blur step S140 smoothes the residual noise blurring of the image and the boundary of the human silhouette with a convolution of the Gaussian function at the output of the erosion operation step S130.

In the expansion operation step S150, an expansion operation is performed by applying a 6 × 6 filter of the morphology technique to the output of the Gaussian Blur step S140.

In other words, the erosion calculation step (S130), the Gaussian blur step (S140), and the expansion operation step (S150) include the boundaries and noises of the backgrounds around the person in the human detection result of the difference image acquisition step (S120). For this purpose, erosion and dilation were applied by applying a 6x6 filter of Morphological processing, and Gaussian convolution in the middle of the morphology process was used to blur the residual noise of the image and human silhouette. Is to smooth the boundary.

The binary image step S160 binarizes the output of the dilation calculation step S150, which corresponds to the final image preprocessing process. The binarization reference value binarizes the human silhouette by setting the reference value after analyzing the histogram of the image.

The erosion calculation step (S130), the Gaussian blur step (S140), the expansion calculation step (S150), and the binary image step (S160) may be referred to as human silhouette extracting units.

With the information of the silhouette of the person acquired in real time in the image preprocessing step (S100), the principal component factor analysis (PCA) method is applied in the posture change analysis step (S100) to independently determine the distance between the camera and the person and the camera angle. Analyze your posture.

The change in the length of the eigenvector of the silhouette obtained in the preprocessing of the two-dimensional image in the image preprocessing step (S100) is an important element to detect the change in the movement of a person. The meaning of the eigenvalue corresponding to each vector indicates the importance of the distributions of the data of the silhouette existing in the image. In other words, one of the two vectors in the silhouette data group indicates that the vector has a long eigenvalue and that the data show a strong pattern in this vector direction.

The posture change analysis step (S100) observes the change in the posture of a person as an important factor of the ratio of the lengths of the eigenvectors of each person's silhouette. In other words, the ratio of the lengths of the eigenvectors can be seen as the same factors as the ratios of the eigenvalues, so that the change in human posture, that is, the sudden fall of the elderly, can be detected independently of the camera distance and shooting angle. It is composed.

Figure 4 is an explanatory diagram for explaining the application of the principal component (PCA) for posture analysis of the elderly according to an embodiment of the present invention.

4 is a vector for posture analysis using PCA when a person is standing, and the ratio between the vectors is small. This is because the vector of one component is very large. However, when a change in posture such as a person starts to fall, when the falling posture is represented as a passage of time, the posture of FIG. 4 may be represented by the postures of FIG. 4, FIG. 3, and FIG. 4. The ratio between the vectors increases while the posture changes up to 3 of 4, and when it falls down and becomes the same as the posture of 4 of FIG. 4, the ratio between the vectors decreases again. In the present invention, using the ratio between the buckers, regardless of the angle and distance of the monitoring camera, a new parameter called the ratio between the vectors for the posture analysis using the PCA in the human silhouette.

That is, as the length (size) ratio of the eigenvectors in each posture of FIG. 4 changes, the ratio of the eigenvalues also changes, and the present invention detects the fall of the elderly by the ratio of the eigenvalues.

5 is a table summarizing the flow of the principal component analysis method applied to the fall monitoring of the elderly in FIG.

The ratio of the eigenvalues of the respective vectors of the initial human silhouette according to the camera is calculated (1 in FIG. 5).

The change in the eigen value calculated as described above is observed every predetermined time (2 in FIG. 5).

Calculate whether the change in the eigen value is more than a certain standard (3 in Figure 5).

When the ratio of the eigenvalues shows a drastic change over the reference value, it is determined that the person has fallen (Fig. 5 4).

6 is a sequence diagram for explaining the posture change analysis step of FIG.

The posture change analysis step includes a data set input step S205, an average value calculation step S213, a variance calculation step S217, an eigenvalue and eigenvector calculation step S230, and an output step S250. Here, only two dimensions of the x and y axes of the image will be described, but it will be clear that the present invention is not intended to be limited to the two dimensions.

 In the data set input step (S205), data of a two-dimensional set of x and y coordinates of an image is input from the image preprocessing step (S100). Data input from the image preprocessing step (S100) refers to a binarized image of a silhouette of a person by image preprocessing, and the data set is a two-dimensional data set with x and y coordinates of the silhouette.

The average value calculating step S213 obtains a mean value from each of the data of the x coordinate and the data of the y coordinate input in the data set input step S205. In other words. The average value of the x- and y-coordinates of a silhouette in a human silhouette image.

In the variance operation step S217, all data of each of the data of the x-coordinate and the data of the y-coordinate are determined based on the mean values of the data of the x-coordinate and the y-coordinate obtained in the mean value calculation step S213. Find the variance Find the variances of the x-coordinates of the human silhouette data based on the average value of the x-coordinates, and find the variances of the y-coordinates in the same way. The covariance between the x and y coordinates is obtained. The covariance matrix is obtained from the variance of the x-coordinate, the y-coordinate, and the co-variance of the x- and y-coordinates of the human silhouette.

In the eigenvalue and eigenvector operation step S230, the eigenvalue for the matrix is obtained from the covariance matrix obtained in the variance operation step S217. Since the image is two-dimensional, two eigenvalues are produced. Each corresponding eigenvector is also obtained.

The output step S250 outputs a result of detecting two eigenvalues and eigenvectors of the silhouette of a person present in the two-dimensional image in the eigenvalue and eigenvector calculation step S230. A vector with a high eigenvalue is a highly influential element of the PCA results of the data set of the human silhouette. The second highest eigenvalue and the corresponding vector are the next most important factor.

Two components of this data set can be obtained as the PCA result of the silhouette of the person in the two-dimensional image obtained through this process. The present invention observes the ratio of these vectors and eigenvalues over time and detects a change in posture of a person using a sudden change in the ratio as a parameter. The present invention can detect the posture of the elderly independently of the angle and distance of the camera.

7 is a schematic sequence diagram of an elderly motion image monitoring method using a principal component method according to an embodiment of the present invention.

The video output from the surveillance camera is converted into a digital signal to remove random noise through a median filter and transmitted to the difference image acquisition unit (S115).

The difference image acquisition unit constructs a background model from the received image information (S123), and with the background image difference image information every hour to separate the person from the background with the difference value from the time when the person enters the image with the pre-built background model. The difference image information between frames is obtained and transmitted to the human silhouette extracting unit (S127).

The human silhouette extracting unit detects a person from the background through the difference image information between the background model and the frame. In the human detection result, there are boundary and noises of the backgrounds around the person. To remove this, erosion, Gaussian blur, and swelling are removed. (dilation) Binarize the result of the calculation process and transmit it to the posture change analyzer (S155).

The posture change analysis unit obtains the average and the variance of the silhouette from the received silhouette information (S210), obtains the covariance using the average and the variance of the silhouette (S220), obtains the eigenvalues and eigenvectors (S230), The ratio of the vector is obtained and transmitted to the human motion detector (S240).

FIG. 8 is an example of analyzing a person's falling posture change by applying the posture change analyzing step of FIG. 6.

As described in FIG. 4, the ratio between the eigenvectors and the eigenvectors calculated in the eigenvector calculation step (S230) until the change in the attitude of falling in the standing state of the person starts and falls. Is large and the ratio between the eigenvectors decreases when it falls down completely.

The present invention is not limited to the above described and illustrated in the drawings, and of course, more modifications and variations are possible to those skilled in the art within the scope of the following claims.

As described above, the present invention provides an elderly motion image monitoring apparatus and method for analyzing the movement of the elderly and determining an emergency situation by performing automated monitoring without restricting the activity.

In addition, the present invention is applied to the image processing and PCA algorithm, the motion of the elderly independently without being affected by the limited conditions of the surveillance cameras (Surveillance Cameras), that is, the angle of the camera and the size of the silhouette of the detected person. Provided are an elderly motion image monitoring apparatus and method using a principal component method for analyzing an emergency situation.

In other words, the method for monitoring the motion of the elderly according to the present invention applies a data analysis technique of Principal Component Analysis to obtain an image of continuously acquiring a change in the ratio between the silhouette vectors of the human body by analyzing the human posture. By capturing 2-3 images per second, posture analysis was used to analyze the occurrence of sudden falls, enabling accurate emergency judgment independent of camera constraints. If the present invention is applied to the image-based monitoring for the elderly, and connected to other network, mobile research field, improved remote medical care system and home healthcare system are possible. In other words, the present invention can be effectively applied to home monitoring, healthcare system for the elderly.

Claims (9)

An image preprocessing step of outputting a binarized image by separating a silhouette of a person from a background of the image from the image information which is captured by the surveillance camera and converted into a digital signal; The PCA algorithm is applied to the information of the silhouette of the person separated in the image preprocessing step (S100) to extract a change in the ratio of the eigenvectors of the silhouette of the person as a parameter and analyze the posture change. Posture change analysis step (S200); An emergency situation determination step (S300) of determining an emergency situation when a sudden change in the magnitude ratio of the eigenvectors of the human silhouette is detected; An emergency situation notification step of notifying an emergency situation to a wireless terminal to a doctor or a family when an emergency situation is determined in the emergency situation determination step (S300); Elderly motion image monitoring method comprising the. Surveillance camera unit consisting of a plurality of surveillance cameras, for taking an image of the elderly and outputting; A monitoring operation processor 200 for analyzing the motion image of the elderly using the principal component method of the image received from the surveillance camera unit 100; In the elderly motion image monitoring apparatus having a; central control unit 700 for transmitting the information received from the monitoring operation processing unit 200 to the doctor's computer of the hospital, in case of emergency situation to inform the wireless terminal of the doctor or family, The monitoring operation processing unit 200 An image preprocessor configured to separate a silhouette of a person from the background of the image in the input image information which is captured by the surveillance camera and converted into a digital signal; Posture change analysis to analyze the posture change by extracting the change in the ratio of the magnitude of the eigenvector of the person's silhouette as a parameter by applying PCA algorithm to the information of the person's silhouette separated in the image preprocessing step (S100) part;  Elderly motion image monitoring device comprising the. The method of claim 1, The image preprocessing step (S100), A captured image step (S105) of converting an image output from the surveillance camera into a digital signal; A median filtering step of removing random noise such as salt and peppers noise generated during a transmission and coding process of an image through a median filter of the image input from the photographed image step S105 ( S110); From the output of the median filtering step S110, in order to separate a person moving in the background, obtaining image information of the difference between the background model and the input frame and acquiring the difference image between the background model and the frame detecting the person (S120). ; An erosion calculation step (S130) of performing an erosion operation by applying a filter of a morphological technique from the output of the difference between the background model and the frame (S120); A Gaussian Blur step (S140) of smoothing the boundary between a silhouette of a person and a residual noise blur of an image by a convolution of a Gaussian function at the output of the erosion operation (S130); A dilation calculation step (S150) of performing an expansion operation by applying a morphology technique filter to the output of the Gaussian Blur step (S140); A binary image step (S160) of binarizing the output of the dilation operation step (S150); Elderly motion image monitoring method comprising the. The method of claim 3, In the acquiring difference image between the background model and the frame (S120), the background model is An elderly motion image monitoring method comprising: building and calculating an average value of ten frames of a background image when a person does not fall within a shooting range of a camera. delete The method of claim 1, The posture change analysis step (S200) step, A data set input step (S205) of inputting data of a two-dimensional set of x and y coordinates of an image output in the image preprocessing step (S100); An average value calculating step (S213) of obtaining a mean value from each of the data of the x coordinate and the data of the y coordinate input in the data set input step (S205); Covariance matrix having the variance of the x coordinate, the variance of the y coordinate, and the covariance of the x and y coordinates based on each mean value of the data of the x coordinate and the y coordinate data obtained in the average value operation step S213 a dispersion operation step S217 for obtaining a covariance matrix; An eigenvalue calculation step for obtaining an eigenvalue for the covariance matrix obtained in the variance operation step (S217), but obtaining two eigenvalues because the image is two-dimensional; An output step (S250) of outputting an eigenvalue of a silhouette of a person present in a 2D image in the eigenvalue calculation step; Elderly motion image monitoring method comprising a. delete An eigenvalue ratio calculation step of calculating a ratio of eigenvalues of a silhouette of a person; An eigenvalue ratio observation step of observing a change in the ratio of the eigenvalues calculated in the eigenvalue ratio calculation step every predetermined time; An eigenvalue ratio conversion determination step of determining whether a change in the eigenvalue ratio is greater than or equal to a predetermined reference in the eigenvalue ratio observation step; A fall determination step of judging that a person has fallen if the ratio of the eigen values shows a sudden change of more than a reference value in the eigen value ratio change determination step; Elderly movement image monitoring method characterized in that it comprises at least. A data input step of converting an image output from the surveillance camera into a digital signal to remove random noise through a median filter; Obtaining a difference image between the background model and the frame, which obtains the difference image information between the background model and the input frame every hour from the time when a person enters the image from the image information received in the data input step; In order to detect a person from the background through the difference image information between the background model and the input frame received in the difference image acquisition step between the background model and the frame, and to remove the boundary and noises of the background around the person in the detection result, A person silhouette extraction step of performing erosion, Gaussian blur, and dilation operation and binarizing the resulting information of the silhouette of the person; The average and the variance of the silhouette are obtained from the information of the silhouette of the person received in the human silhouette extraction step, the covariance is obtained using the average and the variance of the silhouette, the eigenvalue is obtained using the covariance, and the ratio of the eigenvalues is calculated. A posture change analysis step of obtaining and detecting a human motion; Elderly motion image monitoring method comprising the.

Images