KR20230027543A - Method and system for classifying false detection cases - Google Patents

Abstract

According to an embodiment of the present invention, an unexpected situation detection system includes: an unexpected situation detection device for detecting an unexpected situation by analyzing different types of data acquired through a data acquisition unit; an unexpected situation false detection classification device which inputs the unexpected situation detection information received from the unexpected situation detection device and additional information received from an external device to an artificial intelligence model which has been learned, and outputs label information of a group to which the unexpected situation detection information belongs, and generates an alarm signal according to the output label information; and a server for receiving the alarm signal from the unexpected situation false detection classification device and outputting the received alarm signal. Therefore, work efficiency of a manager of a road management institution can be improved.

Description

Method and system for classifying false detection cases {Method and system for classifying false detection cases}

A method and system for classifying unexpected false positives are posted. More specifically, an emergency detection method and system capable of increasing the efficiency of use of an emergency detection system for detecting an unexpected situation on the road and improving the work efficiency of a manager managing the unexpected situation detection system are disclosed.

As interest in road traffic safety increases, various facilities are being installed around general roads and highways. For example, a radar-based emergency detector or a Closed Circuit Television (CCTV)-based emergency detector is installed around a road, and an unexpected situation is detected using the same. Unexpected situations refer to non-repetitive and unpredictable events that occur on the road, and examples thereof include traffic accidents, road damage, obstacles on the road (falling objects, vehicles stopped on the shoulder), and the like.

In this way, when a radar-based emergency detector or a CCTV-based emergency detector is used, there is an advantage in that detection of an emergency can be automated. However, the conventional unexpected situation detection system to which this technology is applied has a problem in that a manager's work efficiency is reduced due to a high false positive rate. Taking a CCTV-based unexpected situation detector as an example, in the case of a CCTV installed in a tunnel, the rate of false positives due to the effects of soot, smoke, insects, etc. is high. In the case of CCTV installed outside the tunnel, the rate of false positives is high due to the influence of weather or a specific time such as sunrise or sunset. In the case of radar-based emergency detectors, the false positive rate is lower than that of CCTV-based emergency detectors, but the false positive rate due to large-sized dust is high.

As such, frequent alarms caused by false positives of the conventional unexpected situation detection system have a problem of reducing the manager's work efficiency, and due to this problem, the unexpected situation detection system is not efficiently used, such as keeping the unexpected situation detection system in an off state. the situation is not possible.

Republic of Korea Patent Publication No. 10-2021-0027778 (Title of Invention: Apparatus and method for analyzing abnormal behavior through object detection and tracking; Publication date March 11, 2021)

A method and system for categorizing unexpected false positives that can increase the efficiency of the emergency situation detection system and improve the work efficiency of managers of road management agencies are posted.

In order to solve the above problems, an unexpected false detection classification system according to an embodiment includes an unexpected situation detecting device for detecting an unexpected situation by analyzing different types of data obtained through a data acquisition unit; The unexpected situation detection information received from the emergency detection device and the additional information received from the external device are input to the artificial intelligence model that has been learned, and label information of a group to which the unexpected situation detection information belongs is output, and the output label an unexpected false detection classification device for generating an alarm signal according to information; and a server configured to receive the alarm signal from the unexpected situation false detection classification device and output the received alarm signal.

In order to solve the above problems, a classification method for detecting and classifying unexpected situations according to an embodiment includes the steps of detecting an unexpected situation by analyzing different types of data obtained by an unexpected situation detection device through a data acquisition unit; The unexpected false detection classification device inputs the unexpected situation detection information received from the unexpected situation detection device and the additional information received from the external device to the artificial intelligence model that has been learned, and obtains the label information of the group to which the unexpected situation detection information belongs. outputting and generating an alarm signal according to the output label information; and outputting, by a server, the alarm signal received from the device for detecting and classifying unexpected situations.

According to an embodiment, a false positive rate may be reduced by collecting various types of data related to traffic conditions using the data acquisition unit and detecting an unexpected situation based on the collected data.

According to the embodiment, an alarm signal is generated and transmitted to the manager only for the emergency detection information classified as requiring confirmation by the manager, so that the manager's work can be prevented from being deteriorated due to unnecessary alarm transmission.

The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

1 is a diagram showing the configuration of a classification system for detecting false positives in an unexpected situation according to an embodiment of the present invention.
2 is a diagram showing the configuration of an emergency situation detection device according to an embodiment of the present invention.
3 is a diagram showing the configuration of an unexpected situation false detection classification device according to an embodiment of the present invention.
4 is a diagram showing the configuration of a server according to an embodiment of the present invention.
5 is a flowchart illustrating an operating method of a system for detecting false positives in an unexpected situation according to an embodiment of the present invention.

Advantages and features of the present invention, and methods for achieving them, will become clear with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in a variety of different forms. Only these embodiments are provided to complete the disclosure of the present invention and to completely inform those skilled in the art of the scope of the invention to which the present invention belongs. The invention is only defined by the scope of the claims.

Terms including ordinal numbers such as 'first' and 'second' may be used to describe various components, but the components are not limited by the terms. These terms are only used for the purpose of distinguishing one component from another. For example, a first element could be termed a second element without departing from the scope of the present invention. Similarly, the second component may also be referred to as the first component. The term 'and/or' includes a combination of a plurality of related recited items or any one of a plurality of related recited items.

It is understood that when a component is referred to as being 'connected' or 'connected' to another component, it may be directly connected or connected to the other component, but other components may exist in the middle. It should be. On the other hand, when a component is referred to as being 'directly connected' or 'directly connected' to another component, it should be understood that no other component exists in the middle.

Terms used in this specification are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless otherwise specified. In this specification, terms such as "comprises" or "comprising" are intended to indicate that there are features, numbers, steps, operations, components, parts, or combinations thereof described in the specification. However, it should be understood that it does not preclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used in a meaning commonly understood by those of ordinary skill in the art to which the present invention belongs. In addition, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless explicitly specifically defined.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the drawings, like reference numerals denote like elements.

1 is a diagram showing the configuration of a system 10 for detecting false positives in an unexpected situation according to an embodiment of the present invention.

Referring to FIG. 1 , a system 10 for classification of false detection of unexpected situations according to an embodiment may include an apparatus for detecting unexpected situations 100 , an apparatus for classifying false positive detection of unexpected situations 200 and a server 300 .

The emergency detection device 100 collects data necessary for detecting an unexpected situation and analyzes the collected data to detect an unexpected situation. Unexpected situations refer to non-repetitive and unpredictable events that occur on the road, and include, for example, a designated body of a vehicle, a traffic accident, road construction, and obstacles on the road (falling objects, vehicles stopped on the shoulder), and the like. A description of the configuration of the device 100 for detecting an emergency will be described later with reference to FIG. 2 .

The unexpected false detection classification device 200 receives the emergency detection information received from the unexpected situation detection device 100 and additional information received from an external device (not shown), classifies the input unexpected situation detection information, and , outputs the information of the classified groups. Then, an alarm signal is generated based on the output group information. The generated alarm signal is transmitted to the server 300 through a wired or wireless network. A description of the configuration of the device 200 for classifying unexpected false positives will be described later with reference to FIG. 3 .

The server 300 may transmit/receive data to and from the emergency detection device 100 and/or the unexpected false detection classification device 200 through a wired or wireless network. For example, the server 300 may receive the emergency detection information acquired by the emergency detection device 100 in real time through a wired or wireless network. Here, the emergency detection information may include data acquired by the emergency detection apparatus 100 and the type of emergency determined based on the acquired data. As another example, the server 300 may receive a notification signal generated by the device 200 for classifying unexpected false positives. The received alarm signal may be output in the form of one or more of a visual signal, an auditory signal, and a tactile signal, and a manager may check the output signal. A description of the configuration of the server 300 will be described later with reference to FIG. 4 .

2 is a diagram showing the configuration of an emergency situation detection device 100 according to an embodiment of the present invention.

Referring to FIG. 2 , the device 100 for detecting an emergency includes a data acquisition unit 110 , an emergency detection unit 120 , a storage unit 130 , a communication unit 140 and a power supply unit 150 .

The data acquisition unit 110 acquires data necessary for detecting an unexpected situation. To this end, the data acquisition unit 110 may include various types of sensors.

As an example, a sensor may include one or more microphones. One or more microphones may be installed inside the tunnel, for example, to obtain sound data inside the tunnel.

As another example, the sensor may include an image sensor. Examples of the image sensor include a CCD (solid state image sensor) sensor and a CMOS (complementary metal oxide semiconductor) sensor.

As another example, the sensor may include at least one of a Light Detection And Ranging (LiDAR) sensor and a radar (RADAR) sensor.

The lidar sensor fires high-power laser pulses, measures the time it takes for the laser to hit the target and returns, and identifies the distance and shape between objects from the measured data. Lidar is also called Time of Flight (ToF) technology.

The radar sensor emits radio waves, measures the time it takes for the radio waves to hit an object and returns, and obtains the object's distance, speed, and direction information from the measured data.

In the above, a microphone, an image sensor, a lidar sensor, and a radar sensor have been exemplified as sensors included in the data acquisition unit 110, but the types of sensors are not necessarily limited to those exemplified, and may also include an ultrasonic sensor and an infrared sensor. there is.

The emergency detection unit 120 detects an unexpected situation by analyzing the data acquired by the data acquisition unit 110 .

For example, when acoustic data is acquired by the data acquisition unit 110, the emergency detection unit 120 extracts characteristic information from the obtained acoustic data and compares the extracted characteristic information with pre-stored characteristic information. Based on the result, the type of unexpected situation, for example, whether it is a collision accident or a collision accident, is detected. If the data acquisition unit 110 acquires multi-channel sound data, the emergency detection unit 120 extracts feature information from the sound data of each channel, and stores the feature information extracted from each channel and the previously stored feature information. It is possible to detect the type of an emergency situation based on a result of comparing feature information of each channel.

As another example, when image data is acquired by the data acquisition unit 110, the emergency detection unit 120 extracts feature information from the obtained image and detects the type of the emergency based on the extracted feature information. . For example, one or more of the size of the object extracted from the image, the color of the object, the type of the object (eg, car, person, animal, obstacle), and the moving speed of the object is detected. More specifically, the emergency detection unit 120 detects an object in the image, recognizes the type of the detected object (eg, car, person, animal, obstacle), and tracks the location of the recognized object. do. The emergency detection unit 120 comprehensively considers these results to determine the type of the unexpected situation.

As another example, when 3D point cloud data is acquired by the data acquisition unit 110, the emergency detection unit 120 performs object detection and object recognition on the acquired 3D point cloud data. ) and object tracking. Object detection refers to detecting a specific object in an image, and determines only the presence or absence of a specific object in the image. Object recognition refers to discriminating whether a detected object is a person, an animal, an object, or a vehicle. Object tracking means tracking the positional change of a specific object in an image.

In order to detect an unexpected situation from data obtained from different types of sensors, the unexpected situation detection unit 120 may include one or more artificial intelligence models. That is, it includes at least one of an artificial intelligence model that detects an unexpected situation by analyzing sound data, an artificial intelligence model that detects an unexpected situation by analyzing image data, and an artificial intelligence model that detects an unexpected situation by analyzing 3D point cloud data. can do. In addition, the emergency detection unit 120 includes an artificial intelligence model that detects an unexpected situation by simultaneously analyzing two different types of data (eg, audio data and video data, image data and 3D point cloud data). You may. Unexpected situation detection information output from one or more artificial intelligence models is transmitted to the unexpected false detection classification device 200 . The unexpected situation detection information transmitted to the unexpected false detection classification device 200 may include only identification information corresponding to the type of the detected unexpected situation, and may further include data obtained from the data acquisition unit 110 in addition to the identification information. You may.

The storage unit 130 stores data, algorithms, programs, applications, and the like necessary for the device 100 to operate the emergency situation. The storage unit 130 can be read by a non-volatile memory, a volatile memory, a built-in memory, a removable external memory, a hard disk, an optical disk, a magneto-optical disk, or any form well known in the art to which the present invention belongs. A recording medium may be included. Examples of the external memory include a secure digital card (SD card), a mini SD card, and a micro SD card.

The communication unit 140 communicates with the unexpected situation false detection classification device 200 and/or the server 300 through a wired or wireless network. To this end, the communication unit 140 supports a wired communication method and/or a wireless communication method supporting a TCP/IP protocol or a UDP protocol.

The power supply unit 150 supplies power to each component of the device 100 for detecting an emergency. According to one embodiment, the power supply unit 150 may be implemented to be mechanically and electrically separable from the device 100 for detecting an emergency situation. The separated power supply unit 150 may be replaced with another redundant power supply unit 150 (not shown). According to another embodiment, the power supply unit 150 may be implemented integrally with the device 100 for detecting an emergency. In this case, the power supply unit 150 may be charged by receiving power from a separately prepared charging device (not shown) according to wired power transmission technology or wireless power transmission technology.

In the above description, the case where the device for detecting an emergency 100 includes both the data acquisition unit 110 and the emergency detection unit 120 has been described as an example, but the data acquisition unit 110 and the emergency detection unit 120 may be configured separately in terms of hardware. In this case, the unexpected situation detection unit 120 may be included in the unexpected false detection classification device 200 instead of the unexpected situation detection device 100.

In addition, in the above description, the case where the data acquisition unit 110 includes several types of sensors has been described as an example, but the data acquisition unit 110 may include only one type of sensor. In this case, a plurality of emergency detection devices 100 including different types of sensors may be installed together at specific locations around the road to obtain data, respectively.

3 is a diagram showing the configuration of an unexpected situation false detection classification device 200 according to an embodiment of the present invention.

Referring to FIG. 3 , the apparatus 200 for detecting and classifying unexpected false positives according to an embodiment includes a communication unit 210 , a classification unit 220 , a control unit 230 and a database 240 .

The communication unit 210 transmits and receives data to and from the emergency situation detection device 100, an external device (not shown), and the server 300 through a wired or wireless network. For example, the communication unit 210 receives emergency detection information from the emergency detection device 100 . The emergency detection information may include one or more of data input to the emergency detection unit 120 of the emergency detection device 100 and data output from the emergency detection unit 120 . Data input to the emergency detection unit 120 includes data obtained from the data acquisition unit 110 . Data output from the emergency detection unit 120 includes identification information corresponding to the type of the detected emergency.

As another example, the communication unit 210 receives additional information from an external device. Here, examples of external devices include a traffic information center of each city, a central traffic information center that integrates and manages traffic information of each traffic information center, and the Korea Meteorological Administration. In addition, the additional information may include, for example, weather (eg, temperature, wind direction, wind speed, rainfall, snowfall, hail), sunrise time, sunset time, fine dust concentration, traffic volume, etc. there is. However, external devices and additional information are not necessarily limited to those exemplified, and any organization that collects and provides basic data necessary to provide traffic situation information can be understood as being included in external devices, and may affect detection of unexpected situations. Any data relating to factors that may influence it may be understood to be included in additional information.

As another example, the communication unit 210 transmits an alarm signal generated by the controller 230 to be described later to the server 300 .

The classification unit 220 receives the emergency detection information received from the emergency detection device 100 and additional information received from an external device, classifies the input emergency detection information, and outputs information of the classified groups. . According to an embodiment, the input emergency detection information may be classified into three groups based on the degree of urgency. For example, when the degree of urgency is expressed as a number between 0 and 100, the entire range of the degree of urgency is three can be classified into groups. Hereinafter, for convenience of explanation, the three groups are referred to as 'first group', 'second group', and 'third group', respectively.

If the urgency of the input emergency detection information is greater than 0 and has a value less than or equal to the first reference value, the input emergency detection information is classified into the first group. In the case of the emergency detection information classified as the first group, it may be regarded as a false positive.

If the urgency of the input emergency detection information is greater than the first reference value and less than or equal to the second reference value, the input emergency detection information is classified into the second group. In the case of the unexpected situation detection information classified as the second group, it may be regarded as true detection but not an unexpected situation to the extent that an administrator intervenes.

If the urgency of the input emergency detection information is greater than the second reference value and is less than or equal to 100, the input emergency detection information is classified into a third group. In the case of the unexpected situation detection information classified as the third group, it corresponds to true detection and may be regarded as an unexpected situation requiring confirmation and/or intervention by an administrator.

That is, the classification unit 220 considers both the emergency detection information received from the emergency detection device 100 and additional information received from an external device, so that the unexpected situation detection information received from the emergency detection device 100 is It classifies whether it is a false positive or a true positive, and if it is a true positive, whether it is a situation that requires administrator's confirmation and/or intervention, or a situation that does not.

For example, reflected light or backlight may occur at sunrise or sunset, which may increase the false detection rate when an unexpected situation is detected from data obtained through an image sensor. Therefore, the classification unit 220 assigns a low weight to the emergency detection information received from the emergency detection device 100 at the sunrise or sunset time, and the information received from the emergency detection device 100 at other times. The received emergency detection information may be classified by assigning a high weight to the emergency detection information. Alternatively, the received unexpected situation detection information is divided into several time zones from sunrise time to sunset time, and the unexpected situation detection information received from the emergency detection device 100 in a time zone close to the sunrise time or sunset time is given a lower weight. The corresponding urgency can be reduced.

As another example, when traffic congestion in the area where the emergency detection device 100 is installed is severe and the vehicle hardly moves, the emergency detection device 100 may recognize the congested vehicle as an obstacle on the road. Accordingly, the classification unit 220 assigns a low weight to the emergency detection information received from the emergency detection device 100 during the time zone when traffic congestion occurs, based on the traffic information of the corresponding area, so that traffic congestion does not occur. The received emergency detection information may be classified by assigning a high weight to the emergency detection information received from the emergency detection device 100 in a time zone that is not set. In addition to the method of applying weights according to traffic congestion, different weights may be applied according to the degree of traffic congestion. Specifically, when traffic congestion is determined based on the vehicle's moving speed, the vehicle's moving speed is divided into several sections, and the unexpected situation detection information received from the emergency situation detection device 100 during the time zone of the most severe traffic congestion By assigning a lower weight, the degree of urgency corresponding to the received emergency detection information may be reduced.

As another example, in the case of the emergency detection device 100 installed in an area where wild animals frequently appear, it may recognize a wild animal moving at high speed as a falling object. Therefore, the classification unit 220 detects an emergency received from the emergency detection device 100 during a time zone in which a wild animal appears frequently, based on information on wild animal appearance frequency by daytime, nighttime, sunrise time, and sunset time in the corresponding region. By assigning a lower weight to information, the degree of urgency corresponding to the received emergency detection information may be reduced.

As described above, the classification unit 220 may assign a weight according to the additional information, but may also assign a weight to the received emergency detection information itself. Specifically, among the received emergency detection information, a high weight may be assigned to overlapping emergency detection information, and a low weight may be assigned to non-overlapping emergency detection information. For example, as a result of analyzing the data obtained from the first sensor, the second sensor, and the third sensor of different types, if an unexpected situation detection result of 'falling object', 'falling object', and 'vehicle' is obtained, respectively, A first weight is applied to emergency situation detection information including identification information corresponding to 'falling object', and a second weight value lower than the first weight is applied to emergency situation detection information including identification information corresponding to 'vehicle'. Weights can be applied.

For the operation as described above, the classification unit 220 may include one or more artificial intelligence models 222 . The artificial intelligence model 222 may be trained based on machine learning. Specifically, the artificial intelligence model 222 may be trained based on deep learning.

Deep learning is a set of machine learning algorithms that attempt high-level abstractions (the task of summarizing key contents or functions in large amounts of data or complex data) through a combination of various nonlinear transformation methods. is defined Deep learning can be seen as a field of machine learning that teaches computers how to think in a large frame.

When there is some data, it is represented in a form that a computer can understand (for example, in the case of an image, pixel information is expressed as a column vector), and many studies are conducted to apply it to learning (how to better express it) techniques and how to build a model to learn them) is in progress. As a result of these efforts, various deep learning techniques have been developed. Examples of deep learning techniques include Deep Neural Networks (DNN), Convolutional Deep Neural Networks (CNN), Recurrent Neural Networks (RNN) and Deep Belief Networks (DBN). can

A deep neural network (DNN) is an artificial neural network (ANN) consisting of a plurality of hidden layers between an input layer and an output layer.

Convolutional deep neural networks (CNNs) are a type of multilayer perceptrons designed to use minimal preprocessing. A convolutional neural network consists of one or several convolutional layers and general artificial neural network layers placed on top of them, and additionally utilizes weights and pooling layers. Thanks to this structure, convolutional neural networks can fully utilize input data with a two-dimensional structure. Also, convolutional neural networks can be trained via standard back-propagation. Convolutional neural networks are easier to train than other feed-forward neural network techniques and have the advantage of using fewer parameters.

A convolutional neural network extracts features from an input image by alternately performing convolution and subsampling on the input image. A convolutional neural network includes several convolution layers, several subsampling layers (subsampling layer, local pooling layer, max-pooling layer), and a fully-connected layer. The convolution layer is a layer that performs convolution on an input image. Further, the subsampling layer is a layer that locally extracts the maximum value of the input image and maps it to a 2D image, increases the local area, and performs subsampling.

Recurrent Neural Network (RNN) refers to a neural network in which the connections between units constituting an artificial neural network constitute a directed cycle. Unlike feed-forward neural networks, recurrent neural networks can utilize memory inside the neural network to process arbitrary inputs.

Deep Belief Networks (DBN) is a generative graphical model used in machine learning. In deep learning, it means a deep neural network consisting of multiple layers of latent variables. There are connections between layers, but there is no connection between units within a layer.

The deep trust neural network can be used for prior learning due to its generative nature, and after learning the initial weights through prior learning, fine-tuning of the weights can be performed through backpropagation or other discrimination algorithms. This characteristic is very useful when there is little training data, because the smaller the training data, the stronger the influence of the initial weight values on the resulting model. The pre-learned initial value of the weight becomes closer to the optimal weight than the arbitrarily set initial value of the weight, which enables performance and speed improvement in the fine-tuning step.

Referring back to FIG. 3 , the artificial intelligence model 222 of the classification unit 220 may be learned with any one of the deep learning techniques described above. At this time, the artificial intelligence model 222 may be learned according to, for example, supervised learning.

Supervised learning refers to training a science and artificial intelligence model using labeling data as learning data. Labeling refers to the task of making a correct answer sheet for given input data, and labeling data refers to the correct answer sheet, that is, input data with a label.

According to the embodiment, the input data may include emergency detection information received from the emergency detection device 100 and additional information received from an external device. For example, the input data may be labeled with an urgency calculated based on the emergency detection information and the additional information as label information. As another example, information of a group to which the calculated urgency belongs may be labeled as label information in the input data.

The classification unit 220 classifies the input emergency detection information using the artificial intelligence model 222 that has been learned. Then, label information labeled in the group to which the classified emergency detection information belongs is output. At this time, the output label information may be one or more of numbers, letters, and special characters. For example, assume that the numbers '1', '2', and '3' are labeled for the first group, the second group, and the third group, respectively. In this case, if the input emergency detection information is classified into the first group, the number '1' labeled in the first group is output from the classification unit 220 . As another example, it is assumed that the first group, the second group, and the third group are labeled with 'false positive', 'true positive', and 'true positive-urgent', respectively. In this case, if the input emergency detection information is classified into the third group, the character 'false positive' labeled in the first group is output from the classification unit 220 .

The control unit 230 controls components within the unexpected false detection classification apparatus 200 according to the label information output from the classification unit 220 .

For example, when the label information output from the classification unit 220 is label information of the first group, the control unit 230 determines that the emergency detection information input to the classification unit 220 is a false positive, and Delete unexpected situation detection information.

As another example, when the label information output from the classification unit 220 is the label information of the second group, the control unit 230 determines that the accident detection information input to the classification unit 220 is a true detection, but the manager confirms and/or Intervention is judged to be an unexpected situation that is not necessary, and logging is performed. During logging, operation information of the emergency detection device 100 that has transmitted the unexpected situation detection information and operation information log information of the unexpected false detection classification device 200 may be stored.

As another example, if the label information output from the classification unit 220 is the label information of the third group, the control unit 230 determines that the input emergency detection information is a true detection, and the manager confirms and/or intervenes. It is determined that it is a situation, and an alarm signal related thereto is generated. The generated alarm signal is transmitted to the server 300 through a wired or wireless network. When the transmitted alarm signal is output through the server 300, the administrator can check the output alarm signal and take action.

The database 240 stores data, programs, applications, and algorithms necessary for the operation of the device 200 for classifying unexpected false positives. For example, the database 240 stores additional information received from an external device. In addition, the database 240 stores logging data.

In the above description, the case where the artificial intelligence model 222 of the classification unit 220 is learned according to supervised learning has been described as an example, but the artificial intelligence model 222 uses unlabeled input data as learning data and compares similar features. It can also be learned according to unsupervised learning that clusters.

4 is a diagram showing the configuration of a server 300 according to an embodiment of the present invention.

Referring to FIG. 4 , the server 300 includes an input unit 310, an output unit 320, a control unit 330, a storage unit 340, and a communication unit 350.

The input unit 310 receives commands or information from an administrator. The input unit 310 may include one or more of a microphone and a key input unit 310 for receiving an audio signal. The key input unit 310 may include a touch key and/or a mechanical key.

The output unit 320 outputs command processing results or various types of information to a manager. For example, the output unit 320 outputs an alarm signal received from the device 100 for detecting an emergency. To this end, the output unit 320 may include a display, a speaker, a haptic output unit 320 and an optical output unit 320, although not shown in the drawings. The display may be provided in a flat panel display, flexible display, opaque display, transparent display, electronic paper (E-paper), or any form well known in the art to which the present invention belongs. can A touch pad may be stacked on the display to form a touch screen, and a touch key may be implemented through the touch screen. In addition to the display and the speaker, the output unit 320 may further include any type of output means well known in the art to which the present invention pertains.

The control unit 330 connects and controls components within the server 300 . For example, each component is controlled so that the alarm signal received from the unexpected false detection classification device 200 is output through the output unit 320 . As another example, the controller 330 generates a response signal including the determination information when determination information is input by a manager. The control unit 330 includes a central processing unit (CPU), a micro processor unit (MPU), a micro controller unit (MCU), a graphic processing unit (GPU), or any type of processor well known in the art of the present invention. It can be.

The storage unit 340 stores data, programs, and applications necessary for the server 300 to operate. The storage unit 340 may include a non-volatile memory, a volatile memory, a hard disk, an optical disk, a magneto-optical disk, or any type of computer-readable recording medium well known in the art to which the present invention belongs. .

The communication unit 350 communicates with the device 200 for classifying unexpected false positives through a wired/wireless network. For example, the communication unit 350 receives an alarm signal from the device 200 for classifying false positive detection of unexpected situations, and transmits a response signal thereto to the classification device 200 for false positive detection of unexpected situations. To this end, the communication unit 350 supports a wired communication method and/or a wireless communication method.

5 is a flowchart illustrating a method for classifying false detection of an unexpected situation according to an embodiment of the present invention.

First, the device 100 for detecting an emergency acquires data through the data acquisition unit 110 including different types of sensors (S510). The data obtained in step S510 is data obtained from one or more sensors selected from among a microphone, a CCD sensor, a CMOS sensor, a lidar sensor, a radar sensor, an ultrasonic sensor, and an infrared sensor, and includes one or more of sound data, image data, and 3D point cloud data. can include

Thereafter, the device 100 for detecting an emergency detects an unexpected situation by analyzing the obtained data (S520). The step S520 includes inputting the obtained data into one or more artificial intelligence models, extracting feature information from the data, and detecting a type of emergency situation based on the extracted feature information.

Then, the unexpected false detection classification device 200 receives emergency detection information and additional information from the unexpected situation detection device 100 and an external device, respectively (S530). The additional information is information provided from external devices such as the traffic information center of each city, the central traffic information center, or the Korea Meteorological Administration, and the weather in the area where the emergency detection device 100 is located (eg, temperature, wind direction, wind speed, rainfall, snowfall, hail), sunrise time, sunset time, fine dust concentration, and traffic volume. However, the additional information is not limited to the exemplified ones, and any data related to factors that may affect the detection of an unexpected situation may be understood to be included in the additional information.

The received unexpected situation detection information and additional information are input to the artificial intelligence model 222 of the unexpected false detection classification device 200 (S540). At this time, the artificial intelligence model 222 may be pre-learned according to supervised learning using labeling data.

Next, the artificial intelligence model of the classification unit 220 classifies the input emergency detection information according to the degree of urgency and outputs label information of the classified groups (S550). The step S550 includes classifying the urgency of the input emergency detection information into three groups based on a first reference value and a second reference value greater than the first reference value, and the urgency of the input emergency detection information. If belongs to the first group less than the first reference value, outputting the label information of the first group, and the urgency of the input emergency detection information is greater than the first reference value and less than or equal to the second reference value in the second group. If belonging to, outputting label information of a second group, and if the urgency of the input emergency detection information belongs to a third group greater than a second reference value, outputting label information of a third group. do.

Thereafter, the device 200 for classifying false detection of unexpected situations generates an alarm signal according to the output label information (S560). The step S560 includes generating an alarm signal when label information of the third group is output from the classification unit 220 .

Thereafter, the device 200 for classifying unexpected false positives transmits the generated alarm signal to the server 300 (S570).

The server 300 outputs the alarm signal received from the device 200 for classifying unexpected false positives (S580). As a result, the manager can check the output alarm signal and take action. For example, a patrol team is immediately dispatched to the location where an unexpected situation is detected.

In the above, with reference to FIGS. 1 to 5 , the unexpected situation false detection classification system and method according to an embodiment of the present invention have been described. In the unexpected false detection classification system 10 according to an embodiment, for example, an alarm signal is generated only when the unexpected situation detection information input to the unexpected false detection classification device 200 is classified into the third group. heard and explained.

Although not shown in the figure, according to the unexpected false detection classification system 10 according to another embodiment, the unexpected situation detection information input to the unexpected false detection classification device 200 is classified into a first group or a second group. In this case, a verification request signal requesting verification of the classification result by a manager may be arbitrarily generated. The verification request signal may include emergency detection information input to the classification unit 220 of the unexpected false detection classification apparatus 200 and label information output from the classification unit 220 .

The generated verification request signal is transmitted to the server 300 through a wired/wireless network, and when the verification request signal is output through the server 300, the manager determines whether the emergency detection information and label information matched thereto are appropriate. That is, it is determined whether the classification result by the unexpected situation false detection classification device 200 is appropriate, and the determination result is input using the input unit 310 . Then, a response signal including the determination result is generated and transmitted to the unexpected false detection classification device 200 through a wired/wireless network.

Then, the device 200 for classifying unexpected false positives may update the artificial intelligence model 222 of the classification unit 220 based on the response signal received from the server 300 .

The embodiments of the present invention disclosed in the present specification and drawings are only presented as specific examples to easily explain the technical content of the present invention and help understanding of the present invention, and are not intended to limit the scope of the present invention. It is obvious to those skilled in the art that other modified examples based on the technical idea of the present invention can be implemented in addition to the embodiments disclosed herein.

10: Unexpected situation false detection classification system
100: unexpected situation detection device
200: Unexpected situation false detection classification device
300: server

Claims (10)

an emergency detection device for detecting an unexpected situation by analyzing different types of data acquired through the data acquisition unit;
The unexpected situation detection information received from the emergency detection device and the additional information received from the external device are input to the artificial intelligence model that has been learned, and label information of a group to which the unexpected situation detection information belongs is output, and the output label an unexpected false detection classification device for generating an alarm signal according to information; and
a server that receives the alarm signal from the unexpected situation false detection classification device and outputs the received alarm signal;
Including, an emergency false detection classification system. According to claim 1,
The data acquisition unit includes one or more sensors,
The sensor includes one or more of a microphone, a CCD sensor, a CMOS sensor, a LiDAR sensor, a radar sensor, an ultrasonic sensor, and an infrared sensor,
Unexpected false positive classification system. According to claim 1,
The above additional information
Including information related to factors that may affect the emergency detection of the emergency detection device,
Unexpected false positive classification system. According to claim 3,
The above additional information
Including one or more of the weather, sunrise time, sunset time, fine dust concentration, and traffic volume in the area where the emergency detection device is located,
Unexpected false positive classification system. According to claim 1,
The sudden signal false positive classification apparatus includes a classification unit including the artificial intelligence model,
The artificial intelligence model is pre-trained to classify the given data into one of a first group, a second group, and a third group,
Contingency false positive classification system. According to claim 5,
The classification unit
When the urgency of the emergency detection information input to the artificial intelligence model is less than or equal to a first reference value, outputting label information of the first group;
When the urgency of the emergency detection information input to the artificial intelligence model is greater than the first reference value and less than a second reference value greater than the first reference value, outputting label information of the second group;
Outputting label information of the third group when the urgency of the emergency detection information input to the artificial intelligence model is equal to or greater than the second reference value.
Unexpected false positive classification system. According to claim 6,
The sudden signal false positive classification device includes a control unit,
The control unit
When the label information of the first group is output from the classification unit, it is determined that the input emergency detection information is a false positive;
When the label information of the second group is output from the classification unit, it is determined that the input accident detection information corresponds to true detection, but it is an accident that does not require confirmation or intervention by a manager;
When the label information of the third group is output from the classifying unit, it is determined that the input emergency detection information corresponds to true detection and is an unexpected situation requiring confirmation or intervention by a manager.
Unexpected false positive classification system. According to claim 7,
The control unit
When the label information of the second group is output from the classification unit, logging is performed;
generating an alarm signal and transmitting it to the server when the label information of the third group is output from the classification unit;
Unexpected false positive classification system. According to claim 7,
The control unit
When the label information of the first group or the label information of the second group is output from the classification unit, a verification request signal requesting verification by a manager for the output label information is randomly generated and transmitted to the server;
When a response signal to the verification request signal is received from the server, updating the artificial intelligence model based on the response signal,
Unexpected false positive classification system. detecting an unexpected situation by analyzing different types of data acquired through a data acquisition unit by an emergency detection device;
The unexpected false detection classification device inputs the unexpected situation detection information received from the unexpected situation detection device and the additional information received from the external device to the artificial intelligence model that has been learned, and obtains the label information of the group to which the unexpected situation detection information belongs. outputting and generating an alarm signal according to the output label information; and
outputting, by a server, the alarm signal received from the classification device for detecting false positive events;
Including, unexpected false positive classification method.

Images