KR20160135004A - Module-based intelligent video surveillance system and antitheft method for real-time detection of livestock theft - Google Patents

Abstract

The present invention relates to a system for monitoring livestock stealing in real time using a modularization-based intelligent image monitoring system for a place for breeding all animals of a cattle shed, a pigsty, a hen house, and so on. More specifically, the present invention relates to a modularization-based intelligent image monitoring system and a method to detect, in real time, stealing a livestock of a cow, a pig, a hen, and so on which are a main livestock in a farm, not moving object detecting technologies focusing on a human by installing a network camera (IP-CCTV), which can wired or wirelessly access internet, at proper intervals inside and outside a place for breeding a livestock, and analyzing an image inputted in real time.

Description

[0001] The present invention relates to a module-based intelligent video surveillance system and an antitheft method for real-time detection of livestock theft,

The present invention relates to a housing, a pig. The present invention relates to a modularized intelligent video surveillance system and method for preventing theft of real-time domestic animals. More specifically, the present invention relates to an intelligent video surveillance system and a method thereof, By installing a network camera (IP-CCTV) and analyzing images input in real time, it is possible not to use human-centered moving object detection technology but to set an intelligent ROI (Region of Interest) through Sobel Operator It combines input image edge extraction function, lateral edge grouping function, automatic setting of interest area fence, Min-Max background modeling and GMM (Gaussian Mixture Model) Motion area noise control function, moving object extraction function, and HOG (Histogram of Oriented Gradien The present invention relates to a modular-based intelligent video surveillance system and method for real-time monitoring of theft of cattle such as cattle, pigs, and chickens, which are major livestock products, by using a modeling method.

 In recent years, various incidents / accidents such as disasters, accidents, traffic accidents such as forest fires, arson, floods, and other physical incidents such as theft, robbery, etc., have occurred in our society.

In particular, rural areas are exposed to crime more frequently than urban areas, and the arrest rate of criminals is also low.

According to the National Police Agency's "2013 Livelihood Crime Status of Livestock Products" report, 3898 cases of agricultural and livestock theft crimes occurred in 5 years (2009 ~ 2013), 742 cases in 2009, 524 cases in 2010, 1108 cases, 1043 cases in 2012, and 481 cases in 2013 (July). In the case of arrests and arrests, the arrest rate is dropping despite the increase in the number of theft crimes. As shown in Table 1 below, the average arrest rate during the five-year period was 44%.

Currently, video-based surveillance devices such as CCTV (Closed Circuit TV) are installed to prevent such crimes and to monitor theft of livestock, ).

However, current CCTV system in rural area has a problem of manually processing the incident by analyzing the stored images after theft mostly due to lack of monitoring personnel.

At the present time when the role of CCTV for rural crime is increasing, IP-CCTV is installed at appropriate intervals in the barn to prevent the manual analysis of the images stored from DVR after theft of major livestock (especially cattle) And automatically analyzes the input video to detect unauthorized theft before operating the alarm system and automatically transmits the theft signal to the related organizations (such as police stations) via wired / wireless Internet. And diffusion.

(1) Livestock Positioning System: RFID tag is installed, a tag reader is installed at a doorway of a house, a tag is recognized in a reader to acquire movement information of a cattle, It is owned by Citi Andy Co., Ltd.

(2) Household security device: Install a sound sensor in the house, extract the feature vector of the sensed audio data, learn the normal situation sound of the house through SVM (Support Vector Machine) Korea University has a technology to detect the occurrence of an abnormal situation.

(3) Management of livestock group based on RFID: RFID (Radio Frequency Identification) tag is used, plural livestock is divided into plural groups, livestock group management is performed for each group, and the identification information is periodically updated, It is a technology to detect it.

(4) Livestock Management System: It is a technology that uses a Real Time Location System (RTLS) tag, operates a wired / wireless server capable of communicating with a tag, and obtains livestock movement information through a client, which is owned by LG Innotek .

(5) Intelligent Livestock Recognition System: Color based motion area detection, object analysis, object classification candidate extraction (human, livestock products), and technology for recognizing human and livestock products through behavior classification.

(1) mobile monitoring in the dairy cattle sector: wireless mobile communication sensor (WAN) is attached to the cattle, and the location information of the cattle is transmitted wirelessly through the WAN, Well Cow is a technology that wirelessly receives location information of a cow and detects the theft.

(2) IP-CCTV Surveillance of Livestock: IP-CCTV is installed in a pre-set monitoring area in the house, wirelessly transmitting video of the monitoring area using wireless communication equipment, DARD has a technology that monitors only the wireless video transmission and does not have real-time theft detection function for image analysis.

(3) RFID based LITS system: LITS (Livestock Identification Trace system), RFID tag attached to livestock cattle, RFID scanner attached to the entrance of the house, and alert to the administrator when the cattle leaves the house. Tinus Burger .

(4) Farm security solutions: Moccall has CCTV installed at the entrance of the housing, alarming the administrator when a moving object is detected at the entrance, and storing the moving object detection image using the NVR.

Also, as shown in FIG. 1, a conventional CCTV-based video surveillance system is largely a surveillance system managed in a country and a private surveillance system managed by a general resident. The large surveillance center monitors and analyzes the CCTV directly, but when the operation price is very expensive and the surveillance site where the theft occurred can not be monitored, the incident image is analyzed and the case is resolved. In addition, the private surveillance system managed by the rural residents is more difficult to monitor than the surveillance center. In order to monitor the theft incident in real time, many operators are needed, but the reality of the rural is insufficient.

These conventional inventions have a problem that the user has to manually set the area of interest again each time the IP-CCTV is installed or the location of the IP-CCTV is moved, and in particular, the complicated environment inside the housing, There is a high possibility that a large number of image reading errors occur due to a large number of movements due to the change.

In particular, as shown in Table 2 below, (US) Buyerbeware's study shows the limitations of conventional CCTV audits.

In addition, moving object detection techniques for preventing image reading errors have a disadvantage in that they are sensitive to noise such as background shake, illumination change, and the like.

In particular, human-centered moving object sensing techniques cause very serious image reading errors when applied to animal-based sensing techniques.

Moving object classification technology to prevent image reading error mainly uses color information, and color information is most commonly used as basic data for classifying objects because it best represents the inherent color of the object. However, when using only color information, there are disadvantages to be sensitive to light, and there is a large error in extraction results.

In addition, expensive IP-CCTV surveillance system must be installed without options, which increases initial installation cost.

Korea Patent Publication No. 2012-0026889 Korean Patent No. 1170819 Korean Patent No. 1194690 Korean Patent No. 1342158 US Patent No. 07259680 U.S. Published Patent Application No. 2011-0298619 Korean Patent Publication No. 2009-0049665 Korean Patent Publication No. 2001-0067809 Korean Patent Publication No. 1998-0068917 Korean Patent No. 0700877 Korean Patent Publication No. 2009-0074387 Korean Patent No. 0908215

[One]. Chensheng Wang, Fei Wang, "A Knowledge-based Strategy for Object Recognition and Reconstruction", 2009 ITCS, pp. 377-391, July, 2009. [2]. Liang Lin, Tianfu Wu, Jake Porway, Zijian Xu, "A stochastic graph grammar for compositional object representation and recognition", Pattern Recognition, Volume 42, Issue 7, pp. 1297-1307, July 2009 [3]. Chenyang Xu and Jerry L. Prince, "Gradient Vector Flow: A New External Force for Snake", IEEE CVPR'97, pp 66-71, 1997 [4] A. Burian and P. Kuosmanen, "Tuning the smoothness of the recursive median filter," IEEE Transactions on Signal Processing, Vol. 50, No. 7, pp. 1631-1639, July 2002. [5] P. Kwan, K. Kameyama, and K. Toraichi, "A comparative labeling problem with introducing a match-all label," Proceedings of the 10th IEEE International Conference on Fuzzy Systems, Vol. 3, pp. 1384-1387, 2001. [6] Y. Ricquebourg and P. Bouthemy, "Real-time tracking of moving objects by exploiting spatio-temporal image slices," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 22, No. 8, pp. 797-808,2000. [7] W. N. M. Isa, "Analysis on spatial and temporal features of gait kinematics," Proceedings of the Fourth IEEE Workshop on Automatics Identification Advanced Technologies. pp. 130-133, 2005. [8] I. Mapoglu, D. Harwood, and L. S. Davis, "W4: Real-time surveillance of people and their activities," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 809-830, Aug. 2000. [9] M. H. Kim, Y. H. Joo, and J. B. Park, "Gesture extraction for ubiquitous robot-human interaction," Journal of Control, Automation, and System Engineering, Vol. 11, No. 12, pp. 1062-1067, Dec. 2005. [10] L. Shao, T. Kadir, and M. Brady, "Geometric and photometric invariant distinctive regions detection," Information Sciences, Vol. 177, Issue 4, pp. 1088-1122, Feb. 2007. [11] R. M. Haralick, S. R. Stemberg, and X. Zhuang, "Image analysis using mathematical morphology," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. PAMI-9, No. 4, pp. 532-550. 1987. [12]. "Recent Trends in Production and Market Forecast of Environmentally Friendly Agricultural Products at Home and Abroad", Korea Rural Economic Institute, 2009.9 [13]. "Background and Expected Effects of the Korea-US FTA" Ministry of Finance and Economy, 2006. 7 [14]. Han, Jong - Wook, and Hyun - Sook Cho, "Trends in Image Security System Technology", Journal of the Korea Institute of Information Security, Vol.19 No. 5, 2009.10].

SUMMARY OF THE INVENTION The present invention has been conceived to solve the above problems, and it is an object of the present invention to provide a module-based intelligent image analyzing apparatus capable of receiving a moving image from a IP- And a system and method for detecting theft of a livestock by a person in real time by tracking a moving object after identifying the object.

In addition, the present invention provides a Sobel operator for setting an intelligent region of interest (ROI) even if a lot of movements occur due to small objects or lighting changes in a special environment inside a house through a new object extraction algorithm or the like The background segmentation algorithm and the motion area extraction function for detecting moving objects by combining the input image edge extraction function and the lateral edge grouping function, the automatic setting of the interest area fence, the Min-Max background modeling and the GMM (Gaussian Mixture Model) Accurate readings using motion area noise control, moving object extraction, and Histogram of Oriented Gradient (HOG) modeling methods to inform administrators and security guards in real-time and real-time transmission of real-time situations via remote monitoring programs Intelligent video surveillance based on modularity for System and method therefor.

According to an aspect of the present invention, there is provided a module-based intelligent video surveillance system for preventing real-time theft of a livestock growing in a large / medium / small scale farmhouse by analyzing images inputted from IP- Wired / wireless IP-CCTV that generates real-time surveillance images of thefts and transmits them via wired / wireless; A video / wireless video receiving device for receiving video from the wired / wireless IP-CCTV; A module-based intelligent image analyzing device for receiving and analyzing images from the wired / wireless video receiving device; An alarm announcement service module for reading the theft or abnormal situation from the analyzed image from the module-based intelligent image analyzing device and dividing the alarm signal into the first intrusion prevention and the second intrusion prevention; A monitoring GUI (Graphic User Interface) capable of monitoring images analyzed by the module-based intelligent image analyzing apparatus in real time; An embedded alarm device for generating an alarm signal by reading an intruding state or a stolen state from an image analyzed by the module-based intelligent image analyzing apparatus; An administrator terminal for receiving an alarm signal from the embedded alarm device; And an affiliate institution terminal that receives an alarm signal from the embedded alarm device.

The module-based intelligent image analyzing apparatus includes an in-store interest area setting module including an input image edge extracting function, a lateral edge grouping function and an interest area fence automatic setting function through a Sobel operator for setting an intelligent ROI; A moving object recognition module in a house including a background separation algorithm function for detecting moving objects, a motion region extraction function, a motion region noise control function, and a moving object extraction function by merging Min-Max background modeling and GMM; A classification module for classifying people and livestock including geometry information extraction function, object classification algorithm function and livestock / person classification function using Histogram of Oriented Gradient (HOG) modeling method; An intrusion detection module for detecting an outsider intrusion situation including a livestock route learning function and a human path learning function for comparing with a learned route; A theft detection module for detecting a livestock stolen situation that identifies a human intrusion by data stored in a human path learning DB through a human path learning function; A moving path learning DB for storing a moving path of a person; And a livestock moving route learning DB for storing the moving route of the livestock.

의 각 픽셀 위치에서 최대 최소 조건을 만족하는 움직임 영역

를 수학식 2를 이용해 얻고, 수학식 2에 의해 추출된 이동 객체의 움직임 영역은 민감도 파라메터 δ의 영향을 받는 것을 특징으로 한다.In addition, the moving object recognition module in a house according to the present invention is characterized in that,

At the respective pixel positions of the moving region < RTI ID = 0.0 >

And the motion region of the moving object extracted by Equation (2) is affected by the sensitivity parameter?.

[Equation 1]

&Quot; (2) "

The present invention divides a target area into cells of a certain size using a HOG method to extract geometric information features of a moving object with a classification module for classifying human and livestock and obtains histograms of the directions of edge pixels for each cell And then a vector obtained by connecting the values in a line is generated.

The module-based intelligent image analyzing apparatus extracts a motion region after applying an image difference using an HSI color model dull to illumination change and a qualified HSI model normalized thereto.

In addition, the module-based intelligent image analyzing apparatus of the present invention may be configured to perform a dilation and an erosion operation to obtain a motion region minimizing noise, or to obtain information on spatial Gradient and temporal information Gradient is used to extract the object.

The module-based intelligent image analyzing apparatus includes a KLT (Kanade-Lucas-Tomasi) algorithm and a color characteristic of a moving object in order to extract a characteristic point of a moving object in order to track each moving object when a moving region is extracted And extracts feature points using the position information.

The module-based intelligent image analysis apparatus uses an active contour algorithm such as GVF (Gradient Vector Flow), which is an outline extraction algorithm.

The present invention relates to a module-based intelligent video surveillance method for preventing real-time animal theft, comprising the steps of: learning a movement path of a person in a surveillance area to detect a human intrusion and storing the learned result in a human movement path learning DB; Learning the movement route of man and livestock in the surveillance area to recognize theft of the livestock and storing the learning result in the human movement route learning DB and the livestock movement route learning DB; A step in which a human intrusion situation occurs inside the housing; By combining the Min-Max background modeling and GMM, we can divide the target area into cells of a certain size by using the background separation algorithm function, the motion area extraction function, the moving area noise control function, the moving object extraction function and the HOG method for moving object detection A histogram of the direction of edge pixels is obtained for each cell, and a vector in which the values are connected in a line is generated to detect occurrence of a stolen state of the livestock leaving the fence by the intruder and the livestock; Detecting an intruding situation by recognizing an abnormal movement of a person based on a human movement path learning DB; Generating an alarm in real time from the control unit of the server to the embedded alarm device inside the housing and the smartphone of the user when an outsider situation occurs; Recognizing an anomalous situation by recognizing an abnormal movement of a person and a livestock based on a moving path learning DB of a person and a livestock; And generating alarms in real time to the related institutions through the embedded alarm device in the housing and the user's smart phone and the Internet from the control unit of the server when the stolen situation of the livestock occurs.

The step of generating a human invasion situation inside the barn of the present invention comprises: entering a person into the barn or entering the barn's outer wall; Livestock The emergence phase of our human being; An abnormal movement path recognition step of a person; The step of entering the livestock by leaving the normal travel route of the person, and the stage where the intruder and the livestock deviating from the fence occur, the intruder and the livestock are separated from the fence; A step in which a human moving path is similar to a moving path of a livestock; Leaving the livestock out of the normal travel path; And no intruder in the surveillance zone.

In the present invention, the step of storing in the human movement path learning DB and the livestock movement path learning DB may include classifying the intruder and the livestock movement path into a normal movement path and an abnormal movement path, Detecting an intruder by recognizing an abnormal movement of a person based on a moving path learning DB of a person to detect an intruder; And detecting theft of the livestock by recognizing abnormal movement of a person and a livestock on the basis of a moving path learning DB of an intruder and a livestock. However, as the data of the learning DB increases over time, the livestock and the intruder Normal / abnormal motion recognition errors are reduced.

In the present invention, the step of detecting the occurrence of theft of the livestock may include an input image edge extraction function through a Sobel operator for setting an intelligent ROI (Region of Interest) And automatically setting an edge grouping function and a fence area of interest.

In the present invention, detecting the occurrence of theft of a livestock is a moving object recognition module in a house, which combines Min-Max background modeling and GMM to perform a background separation algorithm function for detecting moving objects, And a noise control function and a moving object extraction step.

In the present invention, detecting the occurrence of a stolen livestock of a livestock includes a geometric information extraction function, an object classification algorithm function, and a livestock / person classification using a HOG modeling method as a classification module.

In the present invention, the step of detecting the occurrence of a stolen livestock includes a step of extracting a moving region after applying an image difference using a HSI color model dull to illumination change and a qualified HSI model normalized thereto.

In the present invention, the step of detecting the occurrence of the stolen state of a livestock includes a step (morphology) of obtaining a motion area minimizing noise by applying Dilatation (expansion) and Erosion (erosion) operations.

In the present invention, detecting the occurrence of a stolen state of a livestock includes an object extraction using spatial information and temporal information.

      In the present invention, the step of detecting the occurrence of a stolen livestock state includes the steps of: detecting a KLT (Kanade-Lucas-Tomasi) algorithm and a motion object And extracting characteristic points using the characteristic of the color and positional information of the characteristic points.

In the present invention, the step of detecting the occurrence of a stolen state of a livestock includes using an active contour algorithm such as GVF (Gradient Vector Flow), which is an outline extraction algorithm.

According to the present invention, The image input through the network camera installed in the place where the livestock including the house is placed is extracted by the input image edge extraction function and the horizontal mask (matrix pixel group) through the Sobel operator for setting the intelligent ROI, , It is possible to extract the horizontal edge and to set the interest area by grouping by analyzing the input image through the automatic setting function of the interest area fence so that it is not necessary to set the region of interest again even if the position of the IP- There is a castle.

In addition, the present invention provides a background separation algorithm function, a motion area extraction function, a motion region noise control function, and a moving image noise control function for detecting moving objects by combining a Min-Max background modeling algorithm and a GMM The object extraction function and the HOG modeling method can accurately and intelligently analyze even a lot of movements due to small objects or lighting changes in a special environment inside the house to monitor the theft of livestock in real time, Lt; / RTI >

In addition, the present invention provides a method for learning a human movement path, comprising: a human movement path learning module for learning a human movement path; a normal movement path learning module for learning a normal movement path in a human movement path; an abnormal movement path learning module for learning an abnormal movement path; And a normal moving course learning module that learns a normal moving course among animal movement routes and an abnormal moving course learning module that learns an abnormal moving course, and a network camera installed in a place where a livestock including a bar is raised Real-time comparison, learning and monitoring of the captured images ensures accurate detection of the perpetrator.

In addition, the present invention can modularize an expensive IP-CCTV surveillance system by modularization according to user's needs by using modularization technology, thereby reducing installation cost and enabling low-income farmers to use at low cost.

In addition, the embedded alarm device of the present invention includes a Wifi communication module for communication, a warning light for emitting an alarm signal as a light signal, a power switch for controlling a power input, a speaker for emitting an alarm warning as a voice signal, Battery, etc., it is possible to monitor not only the livestock products but also thefts of the goods in the remote places where power supply is difficult, and prevent access to wild animals that damage livestock products. In the case of large-scale livestock cultivation areas, It can also be used as a large-scale system for monitoring intruders.

FIG. 1 is a view showing a direct monitoring system for an operating and monitoring personnel according to the prior art.
FIG. 2 is a diagram showing the overall configuration of a module-based intelligent video surveillance system for real-time animal theft prevention according to the present invention.
FIG. 3 is a detailed block diagram of the module-based intelligent image analyzing apparatus of FIG. 2. FIG.
FIG. 4 is a diagram illustrating the overall configuration of a module-based intelligent video surveillance system for real-time animal theft prevention according to an embodiment of the present invention.
FIG. 5 is a diagram illustrating a detailed configuration of a module-based intelligent video surveillance system for real-time animal theft prevention according to another embodiment of the present invention.
FIG. 6 is a diagram illustrating an example of a region of interest setting according to the edge information analysis according to the present invention.
Figure 7 is a diagram illustrating an example of edge extraction in accordance with the present invention.
8 is a diagram showing an object extraction algorithm according to the present invention.
9 is a diagram illustrating a HOG Feature Construction according to the present invention.
FIG. 10 is a diagram showing details of a human intrusion situation and a domestic animal stolen state according to the present invention.
11 is a diagram showing an example of an embedded alarm device according to the present invention.
12 is a diagram showing a GUI configuration of a modular based intelligent video surveillance system for preventing real-time animal theft according to the present invention.
13 is a diagram showing a GUI for a smartphone for livestock theft monitoring according to the present invention.

For a better understanding of the present invention, a preferred embodiment of the present invention will be described with reference to the accompanying drawings. The embodiments of the present invention may be modified into various forms, and the scope of the present invention should not be construed as being limited to the embodiments described in detail below. The present embodiments are provided to enable those skilled in the art to more fully understand the present invention.

Therefore, the shapes and the like of the elements in the drawings can be exaggeratedly expressed to emphasize a clearer description. It should be noted that in the drawings, the same members are denoted by the same reference numerals. Further, detailed descriptions of well-known functions and configurations that may be unnecessarily obscured by the gist of the present invention are omitted.

FIG. 2 is a diagram showing the overall configuration of a module-based intelligent video surveillance system for real-time animal theft prevention according to the present invention, and FIG. 3 is a detailed configuration of the module-based intelligent video analysis apparatus of FIG.

As shown in FIG. 2, the present invention can be applied to a wireless communication system including a wired / wireless IP-CCTV 12-1, 12-2, 12-3, a wired / wireless video receiving device 13, an NVR 13-1, An intelligent image analyzing apparatus 100, an alarm issuing service module 160, a monitoring GUI 200, an embedded alarm device 14, an administrator terminal 16 and a police / crime prevention /

As shown in FIG. 3, the module-based intelligent image analysis apparatus 100 includes an in-store interest area setting module 110, a moving object recognition module 120, a classification module 130 for classifying people and livestock An intrusion detection module 140 for detecting an outsider's intruding situation, a theft detection module 150 for detecting a livestock stolen situation, a human movement route learning DB 408, a domestic route learning DB 308, and the like .

The wired / wireless IP-CCTVs 12-1, 12-2 and 12-3 of FIG. 2 generate images for monitoring the theft of livestock in real time, transmit them by wire or wirelessly, and classify livestock / persons in the set intelligent ROI It is a device to identify human intrusion by comparing with the data stored in the moving path learning DB. It is a device for detecting high-resolution IP-CCTV Lt; / RTI >

The wired / wireless video receiving apparatus 13 receives an image for real-time monitoring of theft of livestock from the wired / wireless IP-CCTVs 12-1, 12-2, and 12-3 and transmits the video to the module-based intelligent video analysis apparatus 100 ).

The module-based intelligent image analysis apparatus 100 combines an input image edge extraction function, a lateral edge grouping function, an automatic interest area fence automatic setting function, a Min-Max background modeling, and a GMM The background separation algorithm for motion object detection, the motion region extraction function, the motion region noise control function, the moving object extraction function and the HOG modeling method are analyzed.

The alarm notification service module 160 of FIG. 2 fuses the Min-Max background modeling and the GMM with the image analyzed by the module-based intelligent image analysis device 100, reads the alarm condition using the HOG modeling method, It is a module that is divided into prevention of first intrusion and prevention of second theft.

Here, when an external invasion situation (primary alarm situation) occurs by the alarm issuing service module 160, the controller of the server generates an alarm in real time to the embedded alarm device inside the housing and the user's smartphone. In addition, if a livestock stolen (secondary alarm situation) occurs, the controller of the server generates an alarm in real time to the relevant organization (police station, etc.) through the embedded alarm device inside the house and the user's smartphone and the Internet, Prevent.

The monitoring GUI 200 shown in FIG. 2 includes an IP-CCTV number window, an intelligent area of interest information window, a monitoring video window of IP-CCTV, and a monitoring GUI operation so as to monitor images analyzed by the module-based intelligent image analyzing apparatus 100 / Stop button, the information of the surveillance video (compression method, transmission speed), the situation propagation related organization (police station, crime prevention unit, related situation room) setting window and the surveillance system logo window.

2 includes a communication module 141, a beacon lamp 142, a power source (not shown), a power source unit 142, and a power source unit 144. The communication module 141, A power switch 143 for controlling the input, a built-in battery 144 for supplying power, a speaker 145 for emitting an alarm warning to a voice signal, and the like.

The administrator terminal 16 shown in FIG. 2 is a terminal for receiving an alarm signal from the embedded alarm device, and can transmit status propagation and meta information (location, situation, time, criminal information) To the communication module 141 of the device 14.

The cannister terminal 18 shown in FIG. 2 receives an alarm signal from the embedded alarm device. It can prevent theft of the livestock farmed by the farmer by real-time video surveillance and improve the real-time killer arrest rate. In the case of cultivated areas, it can also be used as a large-scale system for monitoring access roads, vehicles, and intruders.

3, in the module-based intelligent image analyzing apparatus 100 for receiving and analyzing images from the wired / wireless video receiving apparatus 13, the intra-area-of-interest-interest setting module 110 sets an intelligent interest area It is a module that includes the input image edge extraction function, the lateral edge grouping function and the interest area fence automatic setting function through the Sobel operator.

As shown in FIG. 3, the moving object recognition module 120 recognizes a moving object in the housing, except for the background within the housing, from the input image of all the IP-CCTVs 12-1, 12-2, and 12-3 installed inside the housing, Min-Max background modeling and GMM are fused to detect all moving objects (people, lights) within a moving object, so that a background separation algorithm function for moving object detection, a motion region extraction function, a motion region noise control function, It is a module that includes extraction function.

When the classification module 130 for classifying people and livestock detects all the moving objects in the inputted images of all the IP-CCTVs 12-1, 12-2 and 12-3 installed in the housing, , It is a module that includes geometry information extraction function, object classification algorithm function and livestock / person classification function using HOG modeling method to classify moving object as livestock or person.

The intrusion detection module 140 for detecting an intruding situation, which is outside of FIG. 3 (a), is a module including a livestock moving path learning function and a human path learning function for comparing with the learned path.

Theft detection module 150 for detecting a livestock stolen situation identifies a person's intrusion from data stored in a human movement path learning DB through a human path learning function, It is a module that identifies theft of livestock with the data stored in the learning DB.

Accordingly, the image input by the above-described modules is classified into automatic setting of the ROI, moving object extraction, and livestock / intruder through S203 to S212 of FIG. 3B.

4, the human movement path learning DB 408 is a DB for storing a human movement path. The human movement path learning DB 408 includes a human movement path learning module 402 for learning a human movement path, And the abnormal movement route learning module 406 which learns the abnormal movement route, and stores and learns the generated data.

4 is a database for storing livestock movement paths. The livestock movement path learning module 308 includes a livestock movement path learning module 302 for learning a livestock movement path and a normal movement path Learning module 304 and an abnormal movement path learning module 306 that learns an abnormal movement path, and stores and learns data to be generated.

Table 3 below shows in detail the hardware and performance of the intelligent animal products monitoring and alarm device according to the present invention.

IP-CCTV (12-1, 12-2, 12-3) is installed inside the house where the images of the area where domestic animals live can be acquired to monitor intruders and livestock.

When the present invention product is installed in a house using the existing analog-CCTV without purchasing the IP-CCTVs 12-1, 12-2, and 12-3, the IoT (Internet of Things) Board to provide wired / wireless network performance such as IP-CCTVs 12-1, 12-2, and 12-3.

The wired / wireless video transmitting / receiving apparatus is installed to transmit / receive images acquired from a plurality of IP-CCTVs (12-1, 12-2, 12-3) installed in the housing to / from the NVR and the server The manager and the related organizations support the fixed / floating IP (Internet Protocol) to access the IP-CCTVs 12-1, 12-2 and 12-3.

The NVR 13-1 has a plurality of connection channels for storing a plurality of real-time images acquired through a wired / wireless video transmitting / receiving apparatus, and has a frame rate of 10 fps (frame per second) or more It has a transmission speed. In addition, the NVR can be stored separately through a storage device (HDD1) for storing images in an ordinary manner and an intelligent image storage device (HDD2) for storing an intruder and a livestock as theft occurs.

The embedded alarm device 14 is equipped with a siren / beacon and is installed inside the housing to generate an alarm inside the housing and wireless communication functions for Internet communication with the servers 100, 160, and 200.

Accordingly, the present invention analyzes real-time images input from a plurality of IP-CCTVs 12-1, 12-2, and 12-3 and transmits the embedded alarm device 14, a smart phone of an administrator, (Police station, autonomous crime prevention unit, situation room, etc.) through an internet communication, so that the H / W of the above-described configuration can be built inside the house to detect the stolen state of the livestock in real time.

hardware Hardware Performance IP-CCTV - Wired / wireless network camera with fixed / dynamic IP support
- Supports resolution (740x480) or higher
- Transfer rate 20fps or more
- Minimum illumination 0.0007Lux or more IoT (Internet of Things) board -Image related S / W can be loaded with MCU function
- Gives wired / wireless network performance such as IP-CCTV to already installed analog CCTV
Wired / wireless video
Receiving device
- Wired / wireless video reception (Wireless Router)
- Fixed / Dynamic IP support
- Multiple video transmission / reception possible NVR
(Network Video Recorder) - Network camera connection (4 channels / 8 channels) or more
- At least 10 fps transmission speed
- 2 or more storage devices (HDD)
-VGA / HDMI Live Monitor Output Embedded
Alarm device - Wired / wireless communication function
- Sirens and warning lights Server
computer - Desktop PC available as a server
- Wired / wireless internet communication function
-CPU i3 / RAM 8GB / VGA GTX 760 or higher

FIG. 4 is a diagram illustrating a detailed configuration of a module-based intelligent video surveillance system for real-time animal theft prevention according to an exemplary embodiment of the present invention. FIG. 6 is a diagram showing a detailed configuration of an intelligent video surveillance system based on a modularity, and FIG. 6 is a view showing an interface configuration of a modular based intelligent video surveillance system for real time animal theft prevention according to the present invention.

In FIG. 4, the process of learning the moving path of man and livestock classifies the moving path of man and livestock into the normal moving path and the abnormal moving path, and stores the learning results in the moving path learning DBs 308 and 408, respectively.

First, in order to detect an intruder of an outsider, an infiltration situation is detected by recognizing an abnormal movement of a person based on a travel route learning DB 408 of a person.

Then, based on the moving path learning DBs 308 and 408 of the human and livestock, the abnormal movement of the human and the livestock is recognized to detect the theft of the livestock.

Therefore, as the amount of data of the learning DB is increased on the basis of the movement path learning DBs 308 and 408 of the human and livestock over time, errors in recognition of normal / abnormal movement of livestock and people are reduced.

Also, as shown in FIG. 4, in the case of theft situation, the abnormal movement of the human / livestock is recognized and the moving path learning based theft recognition method using the situation recognition algorithm is used. In case of the intrusion situation, We use the intrusion detection method based on the path learning method using recognition algorithm.

In other words, the learning process of the human and livestock movement path within the housing according to the present invention includes: 1) classifying the movement path of man and livestock into the normal movement path and the abnormal movement path, and storing the learning results in the movement path learning DB .

2) Next, in order to detect the intruder, it recognizes the abnormal movement of the person based on the moving route learning DB of the person and detects the intrusion situation.

3) Detect theft of livestock by recognizing abnormal movement of people and livestock based on DB of moving route of people and livestock.

Therefore, as the learning process becomes longer over time, the data of the route learning DB increases, and consequently, errors in recognition of normal / abnormal movement of livestock and people are reduced.

As shown in FIG. 5, an alarm condition is read from an image analyzed by a module-based intelligent image analyzing apparatus, an alarm signal is divided into an intrusion prevention 1 step and a theft prevention 2 step. When the intruding situation occurs (S102), the moving object is extracted / classified by Min-Max background modeling or the like (S103). If the alarm situation is the first stage (S104) An alarm is issued (S106).

In the case of the alarm situation 2 (S107), the alarm alarm device is activated (S108), an alarm is issued to the manager (S109), and an alarm is issued to the relevant authority (S110).

FIG. 6 shows an example of an intelligent interest area setting method, FIG. 7 shows an example of edge extraction according to the present invention, FIG. 8 shows an object extraction algorithm according to the present invention, Is a diagram showing a HOG Feature Construction according to the present invention.

How to set up an intelligent ROI (Region of Interest); Sobel Operator

  There has been a problem that the user has to reset the area of interest each time the IP-CCTV is installed or the location of the IP-CCTV is moved.

In order to solve this problem, the present invention proposes a method of setting an intelligent ROI.

The intelligent ROI module according to the present invention analyzes the input image to set the ROI, thereby changing the position of the IP-CCTV 12-1, 12-2, and 12-3, There is no.

As shown in Fig. 6, the housing is generally designed in a rectangular structure for convenience of maintenance and management.

Therefore, the in-store interest area setting module 110 can automatically set the area of interest in the livestock by extracting edge areas from the input image and selecting edges grouped with the side information.

As shown in FIG. 7, the ROI setting module 110 extracts an edge having horizontal information in the image through a Sobel operator.

In other words, among various edge extraction masks such as Sobel, Pristwit, Roberts, Tsurei Chen, and Sta Kastick, the Sobel mask can extract edges in all directions, and in particular, can use a horizontal mask The horizontal edges are extracted, and the projected pixel values are relatively averaged, so that they are relatively strong in noise and more sensitive to the diagonal edge than the vertical and horizontal edges. Therefore, it is advantageous to identify objects (people, cattle, pigs, chickens, etc.) whose internal pattern is simple but unique with unique contour information without modifying the shape of the object.

Min-Max background modeling and GMM fusion; Moving object detection method

As shown in FIG. 8, the present invention can be applied to all the objects moving within a barn, except for the background within the barn in the input image of all the IP-CCTVs 12-1, 12-2, and 12-3 installed inside the barn (Person, light, etc.) can be detected.

In order to continuously monitor various moving objects, the moving object detection technology using the moving object recognition module 120 according to the present invention extracts a moving area excluding the background of the image, Detects objects.

In the present invention, a complicated environment in which a lot of movements due to a small object or illumination change occurs in a special environment inside a house requires a special modeling due to a lot of movements due to a small object or illumination change unlike a simple background. Hereinafter, embodiments of modeling applicable to the intelligent animal products monitoring and alarm apparatus according to the present invention will be described in detail.

For example, in the present invention, it is very difficult to extract a moving object using a simple moving object sensing technology. Therefore, a Min-Max background modeling method or a GMM method using the moving object recognition module 120 is used.

However, these methods are disadvantageous in that they are sensitive to slight environmental changes, that is, noise such as background shake, illumination change, and the like. In order to solve such a problem, the present invention explains Min-Max background modeling and GMM fusion method improving the existing moving object detection technology as follows.

As shown in FIG. 8, a new object extraction algorithm is developed by merging Min-Max background modeling and GMM according to the present invention and improving color information used for background modeling.

A new object extraction algorithm that combines Min-Max background modeling and GMM is a method for setting the sensitivity of a moving object by introducing user parameters into an RGB color model used for background modeling.

의 각 픽셀 위치에서 최대 최소 조건을 만족하는 움직임 영역

를 수학식 2를 이용해 얻는다.From the Min-Max background modeling result of the moving object recognition module 120,

At the respective pixel positions of the moving region < RTI ID = 0.0 >

Is obtained by using Equation (2).

The motion region of the moving object extracted by Equation (2) is affected by the sensitivity parameter?.

Histogram of Oriented Gradient (HOG); Moving object classification method

When a person and a cow are recognized in the input image of the IP-CCTVs 12-1, 12-2, and 12-3, a situation occurs depending on the type, Is required.

9, the module-based intelligent image analysis apparatus 100 according to the present invention performs all the movement from the input image of all the IP-CCTVs 12-1, 12-2, and 12-3 installed in the housing, When an object is detected, the moving object classifies whether the moving object is a livestock or a person because the situation is different depending on the type of each moving object.

Conventional moving object classification techniques mainly use color information, and color information is most commonly used as basic data for classifying objects because it best represents an inherent color of an object. However, when using only color information, there are disadvantages to be sensitive to light, and there is a large error in extraction results.

Accordingly, in the present invention, geometric information of a moving object is extracted from a module-based intelligent image analysis apparatus 100 by a moving object classification method through a classification module 130 for classifying people and livestock, and based on the extracted geometric information, Is classified as a person or animal.

Other

The motion region extraction method extracts the motion region around the livestock and removes the environmental factors such as water, shaking, shadows, and rain among the extracted motion regions.

Because the places where the livestock are located usually have many natural objects, there are fluctuations and changes of natural objects caused by water, rain, and winds.

Therefore, the present invention extracts the motion region after applying the image difference using the HSI color model dull to illumination change and the qualified HSI model normalized thereto. In addition, the HSI color model is composed of color, hue, saturation, and intensity, so it can be used less sensitive to illumination than RGB color model because it can be used separately RGB color model is converted into HSI color model and used.

The HSI color model does not need to know what percentage of blue or green it needs to produce a color when using color space. It simply adjusts the saturation to turn the deep red into pink. To brighten the darkness, adjust the brightness. The HSI is modeled as a cylindrical coordinate system.

Specifically, the geometric information-based object classification method of the classification module 130 for classifying a person and a livestock according to the present invention extracts geometric information features of the extracted moving objects by using a HOG method, , A histogram of the direction of edge pixels is obtained for each cell, and a vector is generated by connecting the values in a line.

That is, since HOG uses the information on the contour of the object, it is advantageous to identify the object (human, cow, pig, chicken, etc.) whose internal pattern is simple but inherent unique contour information.

In addition, the module-based intelligent image analyzing apparatus extracts a motion region after applying an image difference using an HSI color model dull to illumination change and a qualified HSI model normalized thereto.

In addition, the module-based intelligent image analyzing apparatus obtains a motion region minimizing noise by applying dilation and erosion operation.

In addition, the module-based intelligent image analysis apparatus extracts objects using spatial information and temporal information.

In addition, the module-based intelligent image analyzing apparatus may further include a KLT (Kanade-Lucas-Tomasi) algorithm and a color characteristic of a motion object to extract feature points of motion objects for tracking each motion object when the motion region is extracted And extracts feature points using the position information.

In addition, the module-based intelligent image analysis apparatus uses an active contour algorithm such as GVF (Gradient Vector Flow), which is an outline extraction algorithm.

FIG. 10 is a view showing details of a human intruding situation and a livestock stolen situation according to the present invention, and FIG. 11 is a view showing an example of an embedded alarm device according to the present invention.

10, in the input images of all the IP-CCTVs 12-1, 12-2, and 12-3 installed inside the place where the livestock including the housing is housed in a large scale, the detected persons and the animals Are required to be continuously monitored and a technique capable of recognizing the situation according to each situation when their behavior or abnormal movement occurs.

In the present invention, the intrusion detection module 140 for detecting the intruder situation of the outsiders recognizes the situation where the intruder enters the house and the situation where the livestock is stolen. First, the situation of outsiders enters into a place where a large number of livestock are raised by an outsider in a day / night environment where the user has no work of congratulation or the end of the day.

For example, theft of the livestock detected by the intrusion detection module 140 for detecting an outsider's intruding situation may be detected when the livestock is detected in the area where the moving path of the livestock is abnormal and the livestock can not move, I know. In addition, even if a livestock is moved by a moving object other than a livestock, it is detected as a stolen condition.

Specifically, if the human intrusion situation (1-1) perceived by the theft detection module (150) for detection of a livestock stolen situation is examined in detail, a person enters the house (1-1-1) (1-1-2), entering the livestock into the us (step 1-1-3) and departing from the normal travel route of the livestock, In the case of a livestock stolen situation (1-2), a large number of animals in the area of interest are constituted by the abnormal moving route recognition (1-1-4), and the continuous intruder detection step 1-2-1 ), Intruders and livestock are separated from the fence, the migration route of the human is similar to the livestock (1-2-2), and in other cases (1-3), the livestock departs from the normal route, No invader phase (1-3-1), and so on.

Accordingly, in the present invention, in order to detect an intruder from outside, a human movement path is learned in a surveillance area, and the learning result is stored in a human movement path learning DB. And it recognizes the abnormal movement of people based on the human movement path learning DB and detects the invasion situation.

In order for the intrusion detection module 140 for detecting the intruding situation detection of the outside of FIG. 3 to recognize the stolen of a livestock, it is necessary to learn a moving route of a person and a livestock in a surveillance area, Save to DB.

The intrusion detection module 140 for detecting the intruder presence of the outsiders recognizes the theft situation by recognizing the abnormal movement of the person and the livestock based on the travel route learning DB.

       In the present invention, when an intruder of an outsider and a livestock stolen situation occurs, the intruder detection module 140 for detecting an outsider's intruding situation and the theft detection module 150 for detecting a livestock stolen situation detect and prevent such a situation in real time In order to be alert and respond according to each situation, technology is needed.

In addition, there is a need for a direct alarm device to warn criminals in places where large animals are being raised.

As shown in FIG. 5 and FIG. 11, the alarm response method of the present invention includes an embedded alarm device 14 (FIG. 5) installed in an intruding area when an intruder situation (primary alarm situation) occurs in a vulnerable area in a place where a livestock is mass- And the module-based intelligent image analyzing apparatus 100 notifies the operator's terminal 16 of the operator of the situation through the wired / wireless Internet function.

In the event that a stolen state (secondary alarm situation) occurs, the embedded alarm device 14 installed in the intrusion zone is operated, and the module-based intelligent image analysis apparatus 100 controls the administrator's terminal (16) and informs the relevant authorities (police station, government office, autonomous crime prevention unit, etc.) to send the situation information and meta information (location and appearance information of criminals, time of theft, An escape route, a moving route of a cow, etc.).

The embedded alarm device 14 of FIG. 2 is installed in a vulnerable area in the housing and generates alarm information when an outbreak situation (alarm condition 1 step) occurs.

For example, the embedded alarm device 14 includes a Wifi communication module 141 for communication, a warning light 142 for emitting an alarm warning as a light signal, a power switch 143 for controlling power input, A speaker 145 for emitting a signal, Detection of criminals through real-time video surveillance including built-in battery 144, power indicator LED 145, and speakers 146 and 147, which provide power, prevents theft of farm animals raised by farmers and improves real- It can also be used as a large-scale system for monitoring access roads, vehicles and intruders in the case of large-scale livestock cultivation areas.

The monitoring GUI for intelligent video surveillance for the final intelligent video surveillance shown in FIG. 12 includes: (1) an IP-CCTV number window; (2) an intelligent area of interest information window; (3) a monitoring video window of IP- (Compression method, transmission speed) of the surveillance image, ⑥ setting window for the situation propagation related organization (police station, crime prevention unit, related situation room), and ⑦ surveillance system logo window.

Therefore, the monitoring GUI 200 according to the present invention can monitor the intelligent image monitoring by outputting the analyzed image from the module-based intelligent image analyzing apparatus 100 in real time.

FIG. 13A shows a smartphone monitoring GUI for livestock theft monitoring according to the present invention. The IP-CCTV monitoring video information window 161, the interest area information window 162, the related institution communication button 163, CCTV image setting and selection button 166, an embedded alarm device ON / OFF button 167, a next IP-CCTV monitoring image 163, A confirmation window 168, and the like.

Accordingly, in order to monitor the inside and the inside of the house by the smartphone GUI, the images of the IP-CCTVs 12-1, 12-2, and 12-3 installed in the housing are connected to the wired / wireless video transmitting / Whenever an intruder of an outsider or a burglar alarm is detected, an alarm is generated on the manager's smartphone as shown in FIG. 13 (b), and brief meta information of the offender is displayed.

According to the above-described configuration, An input image is extracted through a Sobel operator to set an intelligent ROI (Region of Interest), which is a new object extraction algorithm, through a network camera installed in a place where a large number of livestock including a house are grown. And a lateral segmentation function, an interest area fence automatic setting function, a Min-Max background modeling, and a GMM, a background segmentation algorithm for moving object detection, a motion region extraction function, a motion region noise control function, HOG (Histogram of Oriented Gradient) modeling method accurately and intelligently analyze even if a lot of movements occur due to small objects or lighting changes in a special environment inside the house to reduce the monitoring time of livestock theft and protect property of farmers .

In addition, the present invention includes a human movement path learning module 402 for learning a human movement path, a normal movement path learning module 404 for learning a normal movement path in a human movement path, and an abnormal movement path learning module 406) and a livestock movement path learning module (302) for learning a livestock movement path, a normal lane learning module (304) for learning a normal lane travel path among the livestock movement paths, and an abnormal lane learning module ), It is anticipated that real-time comparison, learning and monitoring of the images generated by the network camera installed in the place where large-sized livestock including the housing is raised will prevent theft of farm animals raised by accurate criminal detection and improve the real-time killer arrest rate .

In addition, the present invention can modularize the expensive IP-CCTV surveillance system by modularization according to the user's needs by using the modularization technology, thereby reducing installation cost and enabling low-income farmers.

The embedded alarm device 14 includes a Wifi communication module 141 for communication, a warning lamp 142 for emitting an alarm warning as a light signal, a power switch 143 for controlling power input, A loudspeaker 145 that emits a signal, and a built-in battery 144 that provides power, as well as livestock products that can monitor theft of a product in a remote area that is difficult to supply power, It can also be used as a large-scale system that monitors access roads, vehicles, and intruders in the case of large-scale livestock cultivation areas.

In the technical aspect of the present invention, there is a moving area extraction and tracking technology, an object classification technique for moving objects, an object tracking technology, an alarm condition input interface technique, a crime situation transmission and alarm agent program technology,

From the economic and industrial point of view, the economic aspects include the intelligence and modernization of the domestic video security system, the improvement of the quality of life of farmers and the improvement of productivity, the upgrading of existing CCTV installation farmers, In the industrial aspect, it is possible to develop a single integrated system when the sensor network such as the crime rate reduction effect on agricultural products and the temperature / humidity related to crop growth is added.

As a commercialization method of the present invention, it can be installed as a system for preventing damage to wild animals (wild boar, elk), installed in agricultural and horticultural farms far away from home, and can be expanded and developed as an integrated development system by linking with a crop growth system. It can be developed as a large-scale system that monitors access roads, vehicles, intruders, etc. in the case of a large agricultural and fishery cultivation area. It can be modified and developed as a system that notifies the manager directly by connecting the alert agent program with the CDMA module.

12-1, 12-2, 12-3: wired / wireless IP-CCTV
13: wired / wireless video receiving device
14: Embedded alarm device
16: Administrator terminal
18: Related institution terminal
100: Module based intelligent image analysis device
110: Interest zone setting module in the house
120: moving object recognition module in the house
130: Classification module for classifying people and livestock
140: Intrusion Detection Module for Outside Infiltration Detection
150: Theft detection module for detection of livestock stolen situation
160: Alarm notification service module
200: Monitoring GUI
308: Livestock Movement Path Learning DB
408: Human movement path learning DB

Claims (20)

A modular based intelligent video surveillance system for real-time prevention of theft of livestock growing in a large scale, medium scale, or small scale farmhouse by analyzing images input in real time from IP-CCTV,
Wireless and IP-CCTV that generates real-time surveillance of theft of livestock and transmits it by wired / wireless;
A video / wireless video receiving device for receiving video from the wired / wireless IP-CCTV;
A module-based intelligent image analyzing device for receiving and analyzing images from the wired / wireless video receiving device;
An alarm announcement service module for reading the theft or abnormal situation from the analyzed image from the module-based intelligent image analyzing device and dividing the alarm signal into the first intrusion prevention and the second intrusion prevention;
A monitoring GUI capable of monitoring images analyzed by the module-based intelligent image analyzing apparatus;
An embedded alarm device for generating an alarm signal by reading an intruding state or a stolen state from an image analyzed by the module-based intelligent image analyzing apparatus;
An administrator terminal for receiving an alarm signal from the embedded alarm device; And
And an affiliated institution terminal for receiving an alarm signal from the embedded alarm device, the intelligent video surveillance system based on a modularity for real time animal theft prevention. The method according to claim 1,
The module-based intelligent image analysis apparatus includes an input image edge extraction function, a lateral edge grouping function, and an interest area fence automatic setting function through a Sobel operator for setting an intelligent ROI (Region of Interest) An interest zone setting module in the bar;
A moving object recognition module in a house including a background separation algorithm function for detecting moving objects, a motion region extraction function, a motion region noise control function, and a moving object extraction function by merging Min-Max background modeling and GMM;
A classification module for classifying intruders and livestock including geometry information extraction function and object classification algorithm function using Histogram of Oriented Gradient (HOG) modeling method and livestock / intruder classification function;
An intrusion detection module for detecting an outsider intruder situation including a livestock route learning function and an intruder's route learning function for comparing with a learned route;
A theft detection module for detecting a livestock robbery situation in which an intruder's intrusion is identified by data stored in a human movement route learning DB through an intruder's route learning function;
A path learning DB for storing an intruder's travel route; And a livestock route learning DB for storing a livestock movement route, and a module-based intelligent video surveillance system for real time livestock theft prevention. 3. The method of claim 2,
The moving object recognition module in the house can be classified into a moving image monitoring object

A motion area satisfying a maximum / minimum condition at each pixel position

And the motion region of the moving object extracted by Equation (2) is influenced by the sensitivity parameter?. The modularization-based intelligent video surveillance system for real-time animal theft prevention is characterized by:
[Equation 1]

&Quot; (2) "

3. The method of claim 2,
The classification module for classifying the intruder and the livestock divides the object area into cells of a predetermined size using the HOG to extract the geometric information characteristic of the moving object to be monitored in real time and obtains a histogram of the direction of the edge pixels for each cell And generates a vector by connecting the values in a line. The modularization-based intelligent video surveillance system for real-time animal theft prevention. The method according to claim 1,
The module-based intelligent image analyzing device extracts a moving area after applying an image difference using a HSI color model dull to illumination change and a qualified HSI model normalized thereto. The modular based intelligent image Surveillance system. The method according to claim 1,
Wherein the module-based intelligent image analyzing apparatus obtains a motion region minimizing noise by applying a dilation and an erosion calculation process to a module-based intelligent video surveillance system for real-time animal theft prevention. The method according to claim 1,
Wherein the module-based intelligent image analyzing apparatus extracts an object using spatial information and temporal information. The intelligent video surveillance system is based on a modularity for real-time animal theft prevention. The method according to claim 1,
The module-based intelligent image analyzer analyzes the KLT (Kanade-Lucas-Tomasi) algorithm and the color characteristics and position of clothes of moving objects to extract feature points of moving objects for tracking of each moving object when the moving region is extracted And extracting feature points using the information. The intelligent video surveillance system based on modularity for real time animal theft prevention. The method according to claim 1,
Wherein the module-based intelligent image analyzing apparatus uses an active contour algorithm such as GVF, which is an outline extraction algorithm, as a structural information, and a module-based intelligent image surveillance system for real-time animal theft prevention. A modular based intelligent video surveillance method for real-time animal theft prevention,
In order to detect intruder intrusion, intruder 's movement route is learned in the surveillance area, and the result is stored in the human movement route learning database. In order to recognize the theft of livestock, the intruder and livestock movement route are learned and learned Storing the result in a human movement path learning DB and a livestock movement path learning DB;
A step in which a human intrusion situation occurs inside the housing;
Min-Max background modeling and GMM are merged to create a target area using a background separation algorithm function, motion region extraction function, motion region noise control function, moving object extraction function and Histogram of Oriented Gradient A step of obtaining a histogram of the direction of edge pixels for each cell and generating a vector in which the values are connected in a line to detect the occurrence of a stolen state of the livestock leaving the fence by the intruder and the livestock;
Detecting an intruding situation by recognizing an abnormal movement of a person based on a human movement path learning DB; Generating an alarm in real time from the control unit of the server to the embedded alarm device inside the housing and the smartphone of the user when an outsider situation occurs;
Recognizing the unauthorized movement of a person and livestock based on the moving path learning DB of a person and a livestock and recognizing the theft situation; And
When a livestock stolen occurs, an alarm is generated in real time from the control unit of the server to the related organization through the embedded alarm device inside the housing and the user's smart phone and the Internet. The modularization based on real time livestock theft prevention Intelligent video surveillance method. 11. The method of claim 10,
The step of causing a human intrusion situation inside the housing includes:
A step in which a person enters the inside of the barn or enters the outside wall of the barn;
Livestock The emergence phase of our human being;
An abnormal movement path recognition step of a person;
A step where a person deviates from the normal movement path and enters into a livestock; and a stage where a livestock stolen where a person and a livestock deviate from the fence occurs is a step in which a person and a livestock are separated from the fence;
A step in which a human moving path is similar to a moving path of a livestock;
Leaving the livestock out of the normal travel path;
And an intruder in the surveillance area. The method of claim 1, 11. The method of claim 10,
The step of storing in the human path learning DB and the livestock path learning DB,
Classifying the movement path of a person and a livestock into a normal movement path and an abnormal movement path, and storing learning results in the respective DBs;
Detecting an intruding situation by recognizing an abnormal movement of a person based on a moving path learning DB of a person to detect a person; And
And detecting theft of the livestock by recognizing abnormal movement of people and livestock based on the moving path learning DB of the human and livestock. However, as the data of the moving path learning DB increases over time, And a normal / abnormal movement recognition error is reduced. A modular based intelligent video surveillance method for real time animal theft prevention. 11. The method of claim 10,
The step of detecting occurrence of a stolen state of a livestock includes:
A step of automatically setting an input image edge extraction function, a lateral edge grouping function and an interest area fence through a Sobel operator for setting an intelligent ROI (Region Of Interest) A modular based intelligent video surveillance method for real - time animal theft prevention. 11. The method of claim 10,
The step of detecting occurrence of a stolen state of a livestock includes:
Extracting a moving object extraction function, a moving region extraction function, a moving area noise control function, and a moving object for a moving object detection by fusing a Min-Max background modeling and a GMM with a moving object recognition module in a house A Modularized Intelligent Video Surveillance Method for Real - Time Livestock Theft Prevention. 11. The method of claim 10,
The step of detecting occurrence of a stolen state of a livestock includes:
The method comprising: extracting geometric information using an HOG modeling method, classifying an object, and classifying a livestock / intruder into a class module; and modularizing the intelligent video monitoring method for real time animal theft prevention. 11. The method of claim 10,
The step of detecting occurrence of a stolen state of a livestock includes:
And a step of extracting a moving region after applying an image difference using a HSI color model dull to illumination change and a qualified HSI model normalized thereto. 11. The method of claim 10,
The step of detecting occurrence of a stolen state of a livestock includes:
The method comprising the steps of: obtaining a motion region minimizing noise by applying a dilation and erosion computation operation to the real-time livestock theft. 11. The method of claim 10,
The step of detecting occurrence of a stolen state of a livestock includes:
And extracting objects using spatial information and temporal information. 2. The method according to claim 1, wherein the spatial information is temporal information. 11. The method of claim 10,
The step of detecting occurrence of a stolen state of a livestock includes:
In order to extract the feature points of moving objects in order to track each motion object when the motion region is extracted, feature points are extracted using KLT (Kanade-Lucas-Tomasi) algorithm and color and position information of clothes of moving objects The method of claim 1, wherein the at least one of the at least two of the at least two of the at least two of the at least two of the at least one of the at least two of the at least two of the at least one of the plurality 11. The method of claim 10,
The step of detecting occurrence of a stolen state of a livestock includes:
Using an active contour algorithm such as GVF, which is a contour extraction algorithm.

Images