KR101177710B1 - Accident Detecting System based on the Sensor Network And Method thereof - Google Patents

Abstract

The present invention relates to an accident detection, the present invention comprises the steps of capturing an image at a position where a plurality of sensor nodes are arranged, extracting object information of specific objects from the image of each sensor node, the extracted Transmitting object information to a master node, the master node integrating the object information to generate integration information, the master node transmitting the integration information to a server, and the server to the delivered integration information Performing accident occurrence inference through ROI block matching and motion vector calculation, and when the accident occurrence inference is determined as the occurrence of the accident, the sensor nodes and the at least one master node for a predetermined time based on the occurrence of the accident Requesting a video of a range and storing the sensor in the server It discloses a configuration of the Walk-based accident detection method.

Description

Accident Detecting System based on the Sensor Network And Method

The present invention relates to an accident detection, and in particular, a plurality of intelligent sensor nodes connected to a network may be installed at various locations such as an accident occurrence point and autonomously recognize and determine an occurrence of an event based on information transmitted from each sensor node. The present invention relates to an accident detection system and an incident detection method based on a sensor network.

As the industry develops, the number of vehicles, which are the means of transportation, increases, the traffic volume increases rapidly, and the speed of the roads progresses. Accidents occurring on the road are not limited to the problem between the driver and the victim, but also have a lot of influence on the vehicles in progress due to the aftermath. Such aftereffects are inevitably larger as the number of vehicles driving on the road and the speed of vehicle flow are faster. If the duration of an accident in any one lane on a three-lane road exceeds 10 minutes, the aftermath may extend up to 10 km back. Therefore, early detection of accidents or accidents on the road can be said to have a great influence on traffic communication because it enables the rapid handling of accidents and accidents. By introducing an accident detection system for accidents or accidents that occur on these roads, you can significantly reduce traffic and congestion times, and in some cases, induce decentralization of vehicles that concentrate on the accident area through bypass roads. There is an advantage.

However, the conventional accident detection system was a system that supports the monitor to determine the presence of an accident through a monitor. Therefore, the conventional accident detection system has a problem that the accident detection rate is different by the individual difference of the monitor to determine the presence of the accident through the monitor. In addition, the conventional accident detection system has a problem in that it is not possible to properly determine and take measures for the occurrence of the accident despite the investment of a large number of people and equipment due to time and space constraints. According to one study, if two or more monitors were monitored simultaneously by an integrated control center, they would not recognize 45% of violations after 12 minutes and miss 95% after 22 minutes. . Therefore, even if a large amount of money is required to build a security system and employ competent security personnel to protect corporate or social safety, it cannot prevent huge losses or security loopholes. Therefore, it is possible to develop a system that can overcome these physical and geographic constraints and enable more effective and automatic recognition and situational response to the occurrence of accidents, thereby preventing further damages that may occur in the second or third phase after the accident. It is a desperate situation.

Therefore, an object of the present invention is to install a plurality of intelligent sensor nodes having multiple autonomous growth and event recognition functions based on multiple composite sensors in a wide range of locations, and to connect each of the sensor nodes to a network, and to generate an event (fire, traffic accident, crime). Sensor nodes installed around them are able to autonomously recognize and judge the occurrence of the event, thus minimizing the information required by the monitoring personnel to detect and respond to the real-time occurrence of the event with a small number of personnel for 24 hours. An accident detection system and an accident detection method are provided.

Sensor network-based accident detection system according to an embodiment of the present invention for achieving the above object collects a certain range of images from the arrangement position, and extracts object information of specific objects included in the collected image A plurality of sensor nodes, at least one master node receiving object information from the plurality of sensor nodes and integrating respective object information to generate integration information, and receiving integration information from the at least one master node and receiving the integration information. Disclosed is a configuration including a server to infer whether an accident occurred based on.

The sensor node may include at least one image camera for collecting the image, a sensor controller which extracts object information included in the image from the image collected by the image camera, and transmits the extracted object information to the at least one master node. The sensor unit may include a communication interface and a sensor storage unit configured to store the collected image and object information.

In particular, the sensor controller may include an object detector that extracts discontinuous object information in the image as continuous dynamic object information based on an image mosaic method and a protrusion map processing method, and a preprocessor that corrects the image.

The master node may include a master communication interface for receiving the object information from the plurality of sensor nodes, and a master controller for integrating object information transmitted from the plurality of sensor nodes to generate integrated information. An information integrator for integrating object information may include an image corrector for correcting an image of the integrated information.

The server of the present invention stores a server communication interface for receiving integration information from the master node, a server control unit for performing accident inference based on the integration information, and stores the accident inference result and generates an accident when determining an accident occurrence. And a server storage unit configured to receive and store an image of a predetermined time range associated with the plurality of sensor nodes and at least one master node, and a display unit or an audio processor to output the accident inference result.

Herein, the server controller is a ROI-based rule discriminator for performing collision check by performing ROI block matching on the objects included in the integrated information, a motion vector calculator for calculating motion vectors of the objects, and ROI of the objects. And an accident inference unit that infers whether or not an accident occurs based on a collision check of the blocks and a calculation result of motion vectors. In particular, the accident inference unit may include the plurality of sensor nodes and at least one when it is determined that the accident has occurred. It is characterized by requesting the master node of the transmission of the image of the predetermined time range associated with the occurrence of the accident.

In another aspect, the present invention comprises the steps of photographing an image at a position where a plurality of sensor nodes are arranged, extracting object information of specific objects from the image of each sensor node, transmitting the extracted object information to the master node The master node integrating the object information to generate integration information, the master node transmitting the integration information to a server, and the server performing ROI block matching and motion vector calculation on the transferred integration information. In the case of inferring the occurrence of the accident through the incident, if it is determined that the accident occurred as a result of the inference of the occurrence of the incident, the plurality of sensor nodes and at least one master node requests the image of a predetermined time range based on the occurrence time of the accident to the server The configuration of the sensor network-based incident detection method comprising the step of storing The deadline.

According to an accident detection system and an accident detection method based on a sensor network according to an embodiment of the present invention, the present invention optimizes reliable information used for checking and determining an occurrence of an accident, thereby reducing concentration of the user while monitoring an accident and reducing concentration. Support to increase.

In addition, the present invention supports to reduce the cost of system management by optimizing system management by performing more efficient accident detection by dividing image analysis by layers.

The present invention is used in a variety of fields, such as driverless vehicles, defense, security, social safety, embedded SW and HW modules for situational awareness, and in particular to enable the development of products that can provide active services, not just passive. It can be used for defense unmanned border system, home network security, social safety disaster prevention system, unmanned vehicle, unmanned aircraft, unmanned vessel, and U-City industry. In addition, the present invention supports the construction of a comprehensive disaster prevention system that can prevent the accidents, disasters, fires, etc. area in advance, and prompt response in the event of a situation, and intelligent unmanned monitoring system at intersections and major roads to prevent traffic accidents. It installs and supports the construction of a system that prevents secondary accidents through wireless communication with nearby vehicles in the event of an accident. In addition, the present invention can support the intelligent unmanned surveillance system for the infrastructure and the interlocking surveillance system for individual buildings.

1 is a view schematically showing the configuration of a sensor network-based accident detection system according to an embodiment of the present invention.
2 is a block diagram showing in more detail the configuration of the sensor node according to an embodiment of the present invention.
3 is a view showing an image applied to the image mosaic according to an embodiment of the present invention.
FIG. 4 is a view illustrating an image mosaic performed on the images illustrated in FIG. 3. FIG.
5 is a view for explaining image conversion according to the application of the protrusion map according to an embodiment of the present invention.
6 is a block diagram showing in more detail the configuration of a master node according to an embodiment of the present invention.
7 is a block diagram showing in more detail the configuration of the server control unit according to an embodiment of the present invention.
8 and 9 are diagrams for explaining a collision check according to overlapping ROI blocks of the present invention.
10 and 11 are views for explaining a collision check based on the result of the motion vector calculation of the present invention.
12 and 13 are diagrams for explaining a graph-based thinking reasoning process showing the variation of the vector value according to the occurrence of the motion vector and the accident.
14 is a signal flow diagram for explaining a sensor network-based accident detection method according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, only parts necessary for understanding the operation according to the embodiment of the present invention will be described, and the description of other parts will be omitted so as not to disturb the gist of the present invention.

The terms and words used in the present specification and claims should not be construed as limited to ordinary or dictionary meanings and the inventor is not limited to the meaning of the terms in order to describe his invention in the best way. It should be interpreted as meaning and concept consistent with the technical idea of the present invention. Therefore, the embodiments described in the present specification and the configuration shown in the drawings are only the most preferred embodiments of the present invention, and do not represent all of the technical idea of the present invention, and various equivalents may be substituted for them at the time of the present application. It should be understood that there may be water and variations.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, only parts necessary for understanding the operation according to the embodiment of the present invention will be described, and the description of other parts will be omitted so as not to disturb the gist of the present invention.

The terms and words used in the present specification and claims should not be construed as limited to ordinary or dictionary meanings and the inventor is not limited to the meaning of the terms in order to describe his invention in the best way. It should be interpreted as meaning and concept consistent with the technical idea of the present invention. Therefore, the embodiments described in the present specification and the configurations shown in the drawings are merely the most preferred embodiments of the present invention, and not all of the technical ideas of the present invention are described. Therefore, It is to be understood that equivalents and modifications are possible.

Hereinafter, the configuration of the sensor network-based accident detection system of the present invention and the role of each configuration will be described in detail with reference to the accompanying drawings.

1 is a diagram schematically illustrating a configuration of a sensor network-based accident detection system according to an exemplary embodiment of the present invention.

Referring to FIG. 1, the sensor network-based accident detection system 10 of the present invention may include a configuration of sensor nodes 100, master nodes 200, and a server 300.

The incident detection system 10 of the present invention having such a configuration collects images of a set range by using an image camera provided by the corresponding sensor nodes 100 after the sensor nodes 100 are installed at various positions, respectively. can do. In this process, the sensor nodes 100 extract object information, for example, coordinate information of the object and motion coordinate information of the object in the case of a moving object, with respect to the collected image, and extract the extracted object information from the upper master nodes ( 200). Then, the master nodes 200 may generate the integrated information by integrating the object information transmitted by the sensor nodes 100 and transmit the generated integrated information to the server 300. As a result, the server 300 may infer only whether an accident occurs according to a predetermined rule based on the integrated object information, and output the inference result to a display unit to transmit information about whether an accident occurs to the server operator. In the incident detection system 10 of the present invention having such a structure, each of the sensor nodes 100 performs object information extraction, and the master nodes 200 perform information integration. It minimizes data and enables effective information processing. Accordingly, the administrator of the incident detection system 10 may achieve a quick result acquisition for the occurrence of the accident based on highly reliable information. Hereinafter, each configuration will be described in detail.

The sensor nodes 100 may be arranged in plural at various points expected by various accident occurrence points or other designers. For example, a plurality of sensor nodes 100 may be disposed in an area near an intersection. The sensor nodes 100 may capture an image of a predetermined range by using the included image camera and extract object information of objects included in the captured image. The sensor nodes 100 may transmit the extracted object information to the upper sensor node, that is, the master nodes 200. In the present invention, two sensor node groups 110 and 120 are illustrated. The first sensor node group 110 includes an eleventh sensor node SN11 and a twelfth sensor node SN12, and the second sensor node group 120 includes a twenty-first sensor node SN21 and a twenty-second sensor node ( Illustrated to include SN22). However, the sensor network-based accident detection system 10 of the present invention is not limited to the number of sensor nodes shown, and a larger number of sensor nodes may be divided and arranged by a certain group. Preferably, the sensor nodes included in one sensor group are disposed in adjacent areas, respectively, so that an image photographed has a similar range. Detailed configuration of the sensor node and the role of the configuration will be described in more detail with reference to the accompanying drawings.

The master nodes 200 may include a first master node 210 and a second master node 220 that are connected to the respective sensor node groups 110 and 120. Here, the master nodes 200 are not limited to the number as shown, but may be added as the number of sensor node groups increases or may be removed as the number of sensor node groups decreases. Referring to the drawing, the first master node 210 may be connected to the first sensor node group 110, and the second master node 220 may be connected to the second sensor node group 120. Accordingly, the first master node 210 may receive object information analyzed about image information captured by the eleventh sensor node SN11 and the twelfth sensor node SN12 included in the first sensor node group 110. Can be. In addition, the second master node 220 may receive object information analyzed about image information captured by the twenty-first sensor node SN21 and the twenty-second sensor node SN22 included in the second sensor node group 120. have. Then, the first master node 210 and the second master node 220 integrate the object information delivered by each sensor node to calculate the integrated object information. In this process, the master nodes 200 may support to perform image correction on the calculated integrated object information. In addition, the master nodes 200 may transmit the calculated integrated object information to the connected server 300. Meanwhile, in the configuration of the sensor network-based accident detection system 10, the master nodes 200 are described separately from the sensor nodes 100, but the present invention is not limited thereto. That is, a sensor node that performs the same role as the sensor nodes 100, and includes a module supporting a function of integrating object information collected and analyzed by each sensor node 100 and transmitting the same to the server 300. The sensor node can also act as a separate master node. Detailed configuration of the master node will be described in more detail with reference to the accompanying drawings.

The server 300 may infer whether an accident occurs with respect to the integrated object information transmitted from the master nodes 200 based on a predetermined rule. In this process, when the probability of occurrence of an accident is calculated to be greater than or equal to a predetermined value, the server 300 outputs the result through a display unit or an audio processor and the corresponding object information to the sensor nodes 100 and the master nodes 200. A request may be made to transmit an image within a predetermined time before and after the related image and the corresponding image. Thereafter, the server 300 may re-determine whether an accident occurs based on the received image and the integrated object information transmitted by the master nodes 200, or may store and manage the transmitted image as accident data. Detailed configuration of the server 300 will be described in more detail with reference to the accompanying drawings.

As described above, in the sensor network-based accident detection system 10 according to an embodiment of the present invention, a plurality of sensor nodes 100 extract individual object information and transmit the extracted information to a higher layer. The data processing for detection is divided into three parts to support fast and accurate processing, and in the server 300 side, only data necessary for the detection of an accident is checked whenever necessary, thereby improving data processing efficiency and management. In terms of the server 300 administrator, it is possible to determine whether an accident occurs more quickly and accurately by collecting and determining only an image having a high probability of an accident.

2 is a block diagram illustrating in detail a configuration of one sensor node among the sensor nodes 100 according to an exemplary embodiment of the present invention. For convenience of description, the reference numeral 100 assigned to the sensor nodes will be described with the same reference numeral "Sensor Node". In addition, for the convenience of the following description, reference numerals 200 assigned to the master nodes will be described with the same reference numerals for the "master node".

2, the sensor node 100 of the present invention includes a sensor communication interface 101, a sensor power supply 103, an image camera 105, a sensor storage unit 107, and a sensor controller 160. can do.

The sensor node 100 of the present invention having such a configuration captures an image of a direction installed using the image camera 105, extracts object information included in the captured image, and then extracts the corresponding object information. It can support sending to the master node connected to it. In addition, when the sensor node 100 receives an image request related to specific object information from the server 300, the sensor node 100 may support to transmit the image of a predetermined time range related to the object information to the server 300.

To this end, the sensor unit communication interface 101 forms a communication channel with the master node 200 and supports transmitting object information extracted from the image collected by the image camera 105 to the master node 200. The sensor unit communication interface 101 may be connected to the master node 200 in at least one of a wired or wireless form. That is, the sensor unit communication interface 101 may transmit image and object information of the sensor node 100 through a wired cable connected to the master node 200. In addition, the sensor unit communication interface 101 may be configured as a wireless communication module to wirelessly transmit the image and object information to the master node 200.

The sensor power supply 103 is a component for supplying power necessary for the operation of the components of the sensor node 100. The sensor power supply 103 may be configured in the form of receiving power from the permanent power source, or may be manufactured in the form of a battery and replaced at regular intervals. In addition, the sensor power supply 103 may be manufactured in the form of a rechargeable battery to transfer power supplied from the permanent power to the respective components of the sensor node 100 and to supply the charged power when the permanent power is cut off. have.

The image camera 105 photographs a background image included within a predetermined angle and objects moving on the background from the disposed position. The image camera 105 may transfer the captured image to the sensor controller 160. One video camera 105 may be included in one sensor node 100 or a plurality of image cameras 105 may be provided in one sensor node 100 according to a designer's purpose.

The sensor storage unit 107 stores an operating system for operating the sensor node 100 and various data generated during the operation of the sensor node 100. In particular, the sensor storage unit 107 performs an object information extraction algorithm for detecting object information on a specific object in the image provided by the image camera 105 and a preprocessing process for the image collected by the image camera 105. It may include a preprocessing algorithm to perform. In addition, the sensor storage unit 107 may store an image collected by the image camera 105 for a predetermined time, and may support to remove the stored image or transmit the image to a preset device after a predetermined time elapses.

The sensor controller 160 controls a signal flow required for operating the sensor node 100 of the present invention, and in particular, extracts object information from an image collected by the image camera 105, and performs a preprocessing process on the image. Can be supported. To this end, the sensor controller 160 may include an object detector 161 and a preprocessor 163.

The object detecting unit 161 may analyze the image collected by the image camera 105 to extract object information on at least one object included in the corresponding image. To this end, the object detection unit 161 performs region analysis by combining object blocks of images taken according to the image mosaic scheme as shown in FIGS. 3 and 4 by combining discontinuous images in the corresponding images. Can extract dynamic object information. 3 is a diagram in which two images for tracking the movement of a vehicle object are arranged side by side up and down in an image photographed by a specific sensor node, and FIG. 4 is a block for two parallel images of FIG. 3. The result of the image mosaic processing is shown through matching.

In addition, the object detection unit 161 may perform object coordinate detection and dynamic object information detection using a saliency map (protrusion map) to detect an object as shown in FIG. 5. Referring to FIG. 5, the object detecting unit 161 performs image processing by performing gray level conversion, or “Convert gray,” on an image “Camera input image” collected by the image camera 105, and then image-processed image. Calculate a "Static saliency map" for, and then calculate a "Dynamic saliency map." The object detector 161 performs a "Removing outlier" to remove the surrounding objects except the main object after calculating the static protrusion map and the dynamic protrusion map. Object detection after "labeling" is supported. The object detector 161 may control the object information to be transmitted to the master node 200 after performing the above-described protrusion map-based object information detection. At this time, the object detecting unit 161 assigns a unique identification number to the detected objects while performing a labeling process, calculates dynamic object information using the location information of the object having the corresponding unique identification number, and then assigns it to the master node 200. You can control the delivery. In particular, the object detector 161 of the present invention may generate a static protrusion image from three layers of pyramid images using Gaussian blurring, and generate a dynamic protrusion image using five static protrusion images. Thereafter, the object detector 161 may remove an abnormal point and detect an object by using a morphology.

The preprocessor 163 may be configured to remove noise, correct pixels, and correct spectra for an image collected by the image camera 105 or a detected object image before or after the object detector 161 performs object detection. Spatial corrections can be performed and the contrast can be improved. In addition, the preprocessor 163 may support spectroscopic correction on the image collected by the image camera 105 so that the same image may be analyzed regardless of the change in light. The preprocessor 163 may perform spatial correction as necessary. The preprocessor 163 may not be included in the sensor node 100 but provided in the master node 200 to be described later, and may be performed together with image correction.

As described above, the sensor node 100 according to an embodiment of the present invention calculates object information on main objects included in the image collected by the image camera 105, and masters the calculated object information and surrounding environment information. A function of delivering to the node 200 may be performed. The sensor node 100 may be implemented by an embedded system scheme.

6 is a block diagram illustrating a configuration of one master node among the master nodes 200 according to an embodiment of the present invention in more detail.

Referring to FIG. 6, the master node 200 of the present invention may include a master communication interface 201, a master power supply unit 203, a master storage unit 207, and a master control unit 260.

The master node 200 having such a configuration collects object information included in a specific image from the sensor nodes 100 through the master communication interface 201 and supports integrating the collected object information. In this case, the master node 200 may support to perform image matching while performing certain matching in the object information integration process.

In more detail, the master communication interface 201 is a similar communication module that is the same as or compatible with the sensor interface communication interface 101 provided in the sensor node 100, for example, in the form of a wired cable or a wireless communication module. The object information and the image transmitted by the sensor node 100 may be received. When the master communication interface 201 receives object information and an image from the sensor node 100, the master communication interface 201 may transfer the corresponding information to the master controller 260.

When the master controller 260 receives object information on specific objects from the sensor nodes 100, the master controller 260 may support to temporarily store the corresponding object information in the master storage unit 207. The master controller 260 may support to integrate the temporarily stored object information. To this end, the master controller 260 may include an information integrator 261 and an image corrector 263. The information integrator 261 classifies the object information delivered by the plurality of sensor nodes 100 for each object, and controls to integrate the classified information. To this end, the information integrator 261 may perform information integration on specific objects by referring to unique identification numbers of the object information transmitted by the sensor nodes 100. The image correction unit 263 is configured to minimize distortion information of an image generated in the information integration process. The image correction unit 263 may extract a feature point using a Harris corner and operate a direct linear transformation (DLT). The image correction unit 263 may support to minimize distortion of an image by using a stitching technique. The master controller 260 may control to transmit the image corrected and information-integrated data to the server 300.

The master power supply unit 203 is configured to supply power required for operating the master node 200. The master power supply 203 may be configured in a similar form to the sensor power supply 103.

As described above, the master node 200 of the present invention integrates and corrects the object information transmitted by the sensor node 100 and delivers the corresponding data to the server 300.

7 is a block diagram illustrating in more detail the configuration of the server 300 according to an embodiment of the present invention.

Referring to FIG. 7, the server 300 includes a server communication interface 301, a server storage unit 307, an audio processor 309, an input unit 313, a display unit 311, and a server controller 360. It may include.

The server 300 of the present invention having such a configuration receives the integrated information delivered by the master node 200 based on the server communication interface 301 and performs inference as to whether an accident occurs based on the integrated information. can do. Hereinafter, each configuration will be described in detail.

The server communication interface 301 communicates with the master communication interface 201 to receive the integrated information transmitted by the master node 200. The server communication interface 301 may transfer the received integrated information to the server controller 360. The server communication interface 301 may be configured with various types of communication modules, such as wired or wireless, according to a connection method with the master node 200.

The server storage unit 307 may store various application programs necessary for driving the server 300, and may store integrated information delivered by the master node 200. In particular, the server storage unit 307 of the present invention can store a variety of application programs necessary for inference on the detection of accidents from the integrated information delivered by the master node 200. For example, the server storage unit 307 may store an ROI based rule determination algorithm, a motion vector calculation algorithm, an accident inference algorithm, and the like for the integrated information. The ROI-based rule determination algorithm may include routines for determining whether to reason about the currently delivered integrated information according to overlapping information of regions designated as ROIs among objects included in the integrated information. The ROI relates to the ROI, and certain objects, for example, vehicle objects, in the image may be the ROI. The motion vector calculation algorithm is an algorithm for calculating a motion vector of objects included in an image. The accident inference algorithm may include routines for assisting in determining whether an accident occurs by performing probabilistic reasoning on all data included in the integrated information that is currently transmitted when an accident inference is requested by the ROI-based rule determination algorithm. . Each of the above algorithms is loaded into the server controller 360 to support performance of an accident detection. Meanwhile, the server storage unit 307 may receive and store a schedule image and surrounding information related to an accident from the sensor nodes 100 and the master nodes 200 according to the accident inference result.

The audio processor 309 is configured to output various audio signals necessary for operating the server 300. The audio processor 309 may include a speaker. In particular, the audio processor 309 may support the server controller 360 to output audio data corresponding to an alarm when it is determined that the integrated information transmitted from the master node 200 corresponds to an accident occurrence. Then, the operator operating the server 300 may recognize the output audio data to more closely check the images according to the occurrence of the accident.

The input unit 313 is a component for generating an input signal for operating the server 300. In practice, the sensor network-based accident detection system 10 may automatically transmit and analyze data of the sensor node 100, the master node 200, the server 300, and perform accident detection accordingly. The input unit 313 for inputting additional information may be omitted. Meanwhile, the input unit 313 may generate an input signal for manually setting the data analysis and request for the incident detection process and an input signal for automatically setting according to the user request. When an input signal for automatic setting is generated by the input unit 313, the accident detection system 10 may automatically detect an image collected by the sensor nodes 100 according to preset routines even if no input signal is generated. Detection and alarm can be performed. When an input signal for manual setting is generated by the input unit 313, the accident detection system 10 is based on automatic setting, but re-inspects specific integrated information according to a user's request, or sensor nodes according to the request. The image request may be performed to the 100 and the master nodes 200.

The display unit 311 is configured to output various screens required in the server 300 operation process. The display unit 311 may include at least one display area or at least one display device. In the case of including a plurality of display areas or display devices, the display unit 311 may support to output the integrated information transmitted from each master node 200 to the plurality of display areas. In this case, the display unit 311 may output the ROI-based rule determination result, the motion vector calculation result, the accident inference result, and the like, which are performed by the server 300, to a predetermined screen area.

The server controller 360 initializes the components constituting the server 300 of the present invention, and controls various signal flows required for an incident detection operation according to an embodiment of the present invention. To this end, the server controller 360 may include a ROI-based rule determination unit 361, a motion vector calculator 363, and an accident inference unit 365.

The ROI-based rule determination unit 361 sets objects detected in the integrated information transmitted by the master nodes 200 as ROI regions, and checks whether the overlapping degree of each ROI region has a predetermined difference or more than a predetermined value. do. For example, as illustrated in FIGS. 8 and 9, the ROI-based rule determination unit 361 may designate each vehicle object as an ROI block, and determine collision range by determining an overlapping range of the ROI blocks. 8 and 9, since the corresponding ROI blocks are detected to overlap each other by a predetermined portion or more, the ROI-based rule determination unit 361 may determine that the integrated information is information having a high probability of accident occurrence. Accordingly, the ROI-based rule determination unit 361 requests the accident inference unit 365 to proceed with the accident inference with respect to the data included in the integrated information, or the motion vector to perform the motion vector calculation before the request for the inference of the accident. The calculation unit 363 may request a motion vector operation related to the integrated information.

The motion vector calculator 363 is configured to calculate motion vectors of objects included in the integrated information transmitted from the master nodes 200. The motion vector calculator 363 may perform a collision check on whether an accident occurs based on the motion vector calculated from the dynamic information of the object. For example, as shown in FIGS. 10 and 11, the motion vector calculator 363 continuously calculates motion vectors of two dynamic objects included in schedule integration information, and stores the motion vectors in one coordinate. Simultaneous output allows you to calculate the probability that two objects will collide with each other. In particular, FIG. 10 illustrates the extent to which the X, Y coordinates of the center point of the ROI region of the object included in the integrated information change from frame to frame. As shown in the drawing, the motion vectors of the two objects are moved to one point, which indicates that the collision probability between the objects is very high. The motion vector calculator 363 may transfer the calculated results of the motion vectors to the accident inference unit 365.

The accident inference unit 365 may perform inference inference based on the information on the ROI blocks transmitted by the ROI-based rule determination unit 361 and the motion vector results of the objects transmitted by the motion vector calculator 363. Can be. When an accident occurs, the objects related to the accident change the direction and speed of movement greatly. Therefore, the accident inference unit 365 may calculate the difference in the motion vector between two or more objects related to the accident, and then reflect the change amount, that is, increase or decrease, in the incident inference. An example of an event situation analysis using the motion vector of the accident inference unit 365 is illustrated in FIGS. 12 and 13. 12 and 13, the graphs having the relatively high vertices show the amount of change in the motion vector difference between the two objects. In the graphs having the relatively low vertices, the vertices show the moment of accident occurrence. The accident inference unit 365 of the present invention may detect whether an accident occurs for various integrated information transmitted by the master nodes 200 by using the graphs based on the motion vectors.

On the other hand, when the accident inference unit 365 determines that an accident occurs with respect to the integrated information, the surrounding image related to the information, for example, the image 10 seconds before the accident after the accident after the sensor node 100 and the master node 200 Can ask. The accident inference unit 365 controls to store the information in response to the request from the sensor node 100 and the master node 200 in the server storage unit 307, and then proceeds to process the accident. It can be provided as decision information.

In the above, the components of the sensor network-based accident detection system 10 and the roles of the components have been described. Hereinafter, an accident detection method operated based on the accident detection system will be described in more detail with reference to the accompanying drawings.

14 is a diagram illustrating a signal flow between components for explaining an accident detection method of a sensor network based accident detection system according to an exemplary embodiment of the present invention.

Referring to FIG. 14, in the accident detection method of the present invention, first, the sensor node 100 performs image collection and object information detection in step S101. To this end, the sensor node 100 may be controlled to initialize the image camera 105 by receiving power from the sensor power supply 103. The sensor node 100 may collect an image of a predetermined range indicated by the image camera 105 and detect object information on the collected image. In this process, the sensor node 100 may perform image processing and object information extraction according to an object mosaicing method based on an image mosaic and a saliency map.

Thereafter, the sensor node 100 may transmit the detected object information to the master node 200 in step S103. In this process, the sensor node 100 may transmit the surrounding environment information of the object to the master node 200 together. In addition, the sensor node 100 may perform object information extraction on the collected image frames in real time and transmit discontinuous information of the corresponding objects to the master node 200 in real time. However, in the case where it is impossible to process a high-capacity data according to the hardware characteristics of the sensor node 100, the sensor node 100 extracts object information on an image frame at predetermined time intervals, and extracts the extracted object information at regular intervals from the master node. And transmit to 200. Meanwhile, the sensor node 100 may generate continuous dynamic object information by integrating respective discontinuous object information and then transfer the same to the master node 200. In this case, the sensor node 100 may assign a unique identification number to an object included in the image and generate continuous dynamic object information through information tracking on an object having the corresponding unique identification number.

The master node 200 integrates object information received from the plurality of sensor nodes 100 in step S105, and performs image correction as necessary. For example, when the sensor nodes 100 transmit discontinuous object information, the sensor node 100 may convert the discontinuous object information into continuous dynamic object information. In addition, the master node 200 supports integrating object information of each sensor node 100 when the sensor nodes 100 deliver continuous dynamic object information, and in this process, the image of the integrated information. Calibration can be performed. The image correction process may be a process of correcting image distortion, lack of pixels, and bending of corners of the integrated information.

In addition, the master node 200 may transmit the integrated information to the server 300 in step S107. Then, the server 300 performs accident inference based on the integrated information delivered in step S109. To this end, the server 300 checks the ROI block designation for objects and collisions according to the overlapping degree of the ROI blocks, or sets the center point of the ROI block in X and Y coordinates, and then frames the motion vectors of the plurality of objects. Changes can be used to check for conflicts. In addition, the server 300 may perform accident inference by using the overlap of the ROI blocks and the change amount of the motion vector of the objects designated as the ROI blocks.

After that, when the accident inference result is determined to be an accident, the server 300 includes image information and other information about the objects at the time of the accident to the sensor nodes 100 and the master nodes 200 in step S111. Request for collection information can be performed.

Then, the sensor nodes 100 and the master nodes 200 may transmit the collection information according to the request of the server 300 to the server 300 in step S113. The server 300 may store the collection information transmitted from the sensor nodes 100 and the master nodes 200 in the server storage unit 307 and provide the information for later processing of the accident.

As described above, the sensor network-based accident detection system and the incident detection method according to an embodiment of the present invention provide each server to extract information about objects in the image, and provide them to the server after the master nodes integrate them. By optimizing the amount of data processing to be processed in the server, it is possible to support more efficient system management and incident detection management.

While the present invention has been described with reference to several preferred embodiments, these embodiments are illustrative and not restrictive. As such, those of ordinary skill in the art will appreciate that various changes and modifications can be made according to equivalents without departing from the spirit of the present invention and the scope of rights set forth in the appended claims.

10: accident detection system 100: sensor node
101: sensor unit communication interface 103: sensor power supply unit
107: sensor storage unit 110: the first sensor node group
120: second sensor node group 160: sensor control unit
161: object detection unit 163: preprocessing unit
200, 210, 220: master node 201: master communication interface
203: master power supply unit 207: master storage unit
260: master control unit 261: information integration unit
263: image correction unit 300: server
301: server communication interface 307: server storage unit
309: audio processor 311: display unit
313 input unit 360 server control unit
361: ROI based rule determination unit 363: motion vector calculation unit
365: accident reasoning unit

Claims (8)

A plurality of sensor nodes which collect a range of images at the arranged position and extract object information of specific objects included in the collected images;
At least one master node receiving object information from the plurality of sensor nodes and integrating object information to generate integration information;
A server that receives the integrated information from the at least one master node and infers whether or not an accident has occurred through ROI block matching and motion vector calculation on the integrated information;
Sensor network-based accident detection system comprising a. The method of claim 1,
The sensor node is
At least one image camera for collecting the image;
A sensor controller configured to extract object information included in the corresponding image from the image collected by the image camera;
A sensor unit communication interface configured to transmit the extracted object information to the at least one master node;
A sensor storage unit storing the collected image and object information;
Sensor network-based accident detection system comprising a. The method of claim 2,
The sensor controller
An object detection unit extracting discontinuous object information in the image as continuous dynamic object information based on an image mosaic method and a protrusion map processing method;
A preprocessor for correcting the image;
Sensor network-based accident detection system comprising a. The method of claim 1,
The master node
A master communication interface for receiving the object information from the plurality of sensor nodes;
And a master controller configured to generate object information by integrating object information transmitted by the plurality of sensor nodes.
The master control unit
An information integration unit for integrating the object information;
An image corrector configured to correct an image of the integrated information;
Sensor network-based accident detection system comprising a. The method of claim 1,
The server
A server communication interface for receiving integration information from the master node;
A server controller configured to perform accident inference based on the integrated information;
A server storage unit for storing the accident inference result and receiving and storing an image of a predetermined time range related to an accident occurrence from the plurality of sensor nodes and at least one master node when determining an accident occurrence;
A display unit or an audio processor which outputs the accident inference result;
Sensor network-based accident detection system comprising a. The method according to claim 5,
The server control unit
A ROI-based rule determination unit performing a collision check by performing ROI block matching on the objects included in the integrated information;
A motion vector calculator for calculating motion vectors of the objects;
An accident inference unit that infers whether an accident has occurred based on a collision check of the ROI blocks of the objects and a calculation result of motion vectors;
Sensor network-based accident detection system comprising a. The method of claim 6,
The accident reasoning unit
Sensor network based incident detection system, characterized in that for requesting the plurality of sensor nodes and at least one master node to transmit an image of a predetermined time range associated with the accident occurs if it is determined that the accident occurred. Photographing an image at a position where a plurality of sensor nodes are arranged;
Extracting object information of specific objects from the captured images of the respective sensor nodes;
Transmitting the extracted object information to a master node;
The master node integrating the object information to generate integration information;
Transmitting, by the master node, the integration information to a server;
The server performing accident occurrence inference through the ROI block matching and motion vector calculation on the transmitted integrated information;
Requesting the plurality of sensor nodes and at least one master node for an image in a predetermined time range based on an accident occurrence time and storing the image in the server when it is determined that the accident has occurred;
Sensor network-based accident detection method comprising a.

Images