Intelligent video detection system including safety warning unit

Abstract

An intelligent video detection system including a safety warning unit is disclosed. The intelligent video detection system according to an embodiment of the present invention comprises: a shooting unit installed adjacent to a location to be monitored and capturing the location to be monitored in real time and transmitting the image to a danger state detection unit; a danger state detection unit equipped with a crowd risk detection unit that detects the risk level of a crowd of people within a preset zone based on video data transmitted from the shooting unit, a road risk detection unit that detects the risk level of a road within a preset zone, and a labor risk detection unit that detects the risk level of a work condition within a preset zone; and a safety warning unit installed at the location to be monitored and at a central control center, which outputs a preset warning signal when a dangerous condition is detected by the danger state detection unit.

Description

Technology Field

[0001] The present invention relates to an intelligent image detection system, and more specifically, to an intelligent image detection system including a safety warning unit. Background Technology

[0002] CCTV-based monitoring systems are widely utilized in various spaces, including public facilities, industrial sites, roads, and commercial facilities, and have been installed for multiple purposes such as crime prevention, traffic management, and industrial accident prevention. However, these systems have limitations as they rely heavily on simply recording footage or allowing managers to directly view it on monitor screens, making it difficult to guarantee an immediate and proactive response to real-time dangerous situations.

[0003] Conventional CCTV monitoring methods require administrators to monitor multiple screens simultaneously to identify abnormal situations, making it highly likely that critical events will be missed during times of reduced visual concentration or when fatigue is accumulated. In particular, during nighttime or periods of insufficient manpower, it is difficult for administrators to closely examine all screens, leading to the problem of failing to immediately identify dangerous situations even if they occur.

[0004] Furthermore, since CCTV monitoring is fundamentally limited to passive surveillance, managers must personally determine whether there is a risk. For example, situations such as a large number of people suddenly gathering at a specific location or a vehicle traveling on an abnormal trajectory on a road are not immediately detected until a manager makes a visual assessment. This presents a problem in that initial response may be delayed when an actual incident occurs, potentially leading to a larger scale of damage.

[0005] Conventional CCTV systems have limitations in that it is difficult to quantitatively analyze the degree of crowd concentration within a specific area. Since administrators must recognize people solely through the images appearing on the screen, it is difficult to respond in real time even when crowds gather in excessive numbers. This limitation acts as a factor that increases the risk of safety accidents, particularly in places where large crowds gather, such as concert halls, subway stations, and stadiums.

[0006] Existing CCTV systems also face limitations in road traffic monitoring. It is difficult to automatically determine whether a vehicle is driving normally, requiring administrators to continuously monitor the footage; furthermore, immediate notifications are not provided at the scene in the event of a traffic accident. Additionally, situations where animals or pedestrians suddenly enter the road cannot be detected unless an administrator personally assesses the footage, posing a high risk of traffic accidents.

[0007] Monitoring of working environments in industrial sites is also inadequate with conventional technology. Since existing CCTVs merely record workers' actions, it is difficult to automatically determine whether they are actually working in dangerous conditions. For instance, there is a limitation in that they cannot provide real-time warnings when workers continue to perform tasks even when weather or lighting conditions are unsuitable.

[0008] In addition, CCTV-based monitoring requires a large number of management personnel to work in shifts to operate continuously 24 hours a day. However, stationing personnel at all times has significant limitations in terms of cost and management efficiency, and the burden of personnel management poses a serious problem, particularly for small businesses or local governments with limited budgets.

[0009] The analysis of video data is also limited in conventional systems. Because they rely on simple recording and playback functions, it is difficult to quickly identify dangerous situations within specific timeframes, and the system remains at the level of verifying events that have already occurred. Consequently, there is a strong tendency to focus on post-incident verification rather than prevention.

[0010] Furthermore, the method of administrators directly monitoring screens exacerbates psychological burden and fatigue. Concentration declines during the process of intensively checking multiple screens for extended periods, frequently leading to a failure to recognize minor anomalies or unexpected situations in a timely manner. This limitation is cited as one of the major causes compromising the overall safety of the system.

[0011] Conventional CCTV monitoring systems ultimately harbor various problems, including the incompleteness of real-time response, the absence of quantitative risk analysis, dependence on management personnel, limitations in ensuring safety at industrial sites, and a lack of traffic accident prevention capabilities. To resolve these issues, an intelligent video detection system is required that can automatically identify dangerous situations based on video data and provide immediate warning signals, and the present invention aims to solve the problems of the conventional technology. Prior art literature

[0012] Korean Patent Publication No. 10-2767275 (Registration Date: February 7, 2025) The problem to be solved

[0013] The present invention aims to solve the problem that conventional CCTV monitoring systems rely entirely on the visual observation of managers, resulting in delayed real-time response and difficulty in immediately identifying overcrowded personnel, abnormally moving vehicles, pedestrians entering the road, and dangerous working conditions. In particular, to overcome the limitations of continuous and accurate monitoring caused by accumulated fatigue of management personnel, personnel shortages, and cost burdens, the invention aims to provide an intelligent video detection system that automatically detects dangerous situations based on video data and provides immediate warnings and notifications to support managers in responding quickly. means of solving the problem

[0014] An intelligent image detection system according to one aspect of the present invention for achieving such an objective may comprise: a shooting unit installed adjacent to a location to be monitored and capturing the location to be monitored in real time and transmitting the image to a danger state detection unit; a danger state detection unit equipped with a crowd danger detection unit that detects the degree of danger of a crowd within a preset zone based on image data transmitted from the shooting unit, a road danger detection unit that detects the degree of danger of a road within a preset zone, and a work danger detection unit that detects the degree of danger of a work state within a preset zone; and a safety warning unit installed at a location to be monitored and a central control center, which outputs a preset warning signal when a dangerous state is detected by the danger state detection unit.

[0015] In one embodiment of the present invention, the cluster risk detection unit of the risk state detection unit may be configured to include: a personnel filtering module that filters only images corresponding to people from images within a preset area; a personnel count calculation module that calculates the number of people from images filtered by the personnel filtering module; a personnel count increase / decrease tracking module that tracks the number of people changing in real time from the personnel count calculation module and monitors the increase / decrease status of the number of people over time; and a personnel overcrowding detection module that predicts the time when the number of people exceeds a reference value based on data detected through the personnel count increase / decrease tracking module, determines a risk situation due to the overcrowding state of the number of people, and transmits the information to the safety warning unit in real time if a risk situation occurs.

[0016] In one embodiment of the present invention, the road hazard detection unit of the hazard state detection unit may comprise: a vehicle filtering module that filters only images corresponding to a means of transportation including a vehicle from an image within a preset zone; an abnormal driving vehicle detection module that detects a means of transportation driving abnormally from an image filtered by the vehicle filtering module and transmits the information to a safety warning unit in real time; a traffic accident vehicle detection module that detects a means of transportation that has caused a traffic accident from an image filtered by the vehicle filtering module and transmits the information to a safety warning unit in real time; a pedestrian filtering module that filters only images corresponding to animals and people excluding means of transportation from an image within a preset zone; and an abnormal pedestrian detection module that determines whether an animal or person detected from an image filtered by the pedestrian filtering module enters the road and transmits the information to a safety warning unit in real time.

[0017] In one embodiment of the present invention, the labor risk detection unit of the risk state detection unit may comprise: a weather information detection module that detects weather information, temperature information, humidity information, and illuminance information within a preset area in real time; a suitability determination module that determines whether conditions suitable for outdoor work within a preset area are set by utilizing a manager or AI, and then compares data detected by the weather information detection module; a worker detection module that detects whether a worker is present within a preset area and detects whether a worker is performing work within a preset area; and a hazardous labor state determination module that transmits information in real time to a safety warning unit if, according to the suitability determination module, a worker is performing work within a preset area despite the conditions not being suitable for outdoor work.

[0018] In one embodiment of the present invention, the safety warning unit may comprise: a field warning output module installed adjacent to a location to be monitored, which operates by receiving an operation control signal from the safety warning unit and outputs a preset warning signal, evacuation signal, guidance broadcast, and recommendation broadcast; a control center output module installed in a central control center, which operates by receiving an operation control signal from the safety warning unit and outputs a preset warning signal, rapid response guidance signal, and response recommendation broadcast to an administrator; a wireless transmission module installed in a central control center, which receives an operation control signal from the safety warning unit and transmits an emergency situation notification signal to a mobile phone carried by an administrator; and an emergency reporting module installed in a central control center, which receives an operation control signal from the safety warning unit and reports to 119 or 112. Effects of the invention

[0019] According to the intelligent video detection system of the present invention, the recording unit, the danger state detection unit, and the safety warning unit are organically combined to significantly enhance real-time response capabilities. The recording unit records a specific area in real time, and the danger state detection unit analyzes crowding risk, road risk, and labor risk by subdividing them, thereby enabling quantitative and automated risk identification beyond simple video recording. The crowding risk detection unit can prevent safety accidents caused by overcrowding by tracking increases or decreases in personnel numbers and predicting situations where they exceed capacity, while the road risk detection unit reduces the likelihood of traffic accidents by immediately detecting abnormally moving vehicles or pedestrians entering the road. Furthermore, the labor risk detection unit automatically determines the suitability of the working environment by comprehensively considering conditions such as weather, illumination, and humidity, and immediately provides notifications to the manager and the central control center when a dangerous work situation occurs. Moreover, the safety warning unit implements a multi-layered response system including on-site warnings, evacuation guidance, control center broadcasts, notifications to the manager's mobile phone, and emergency reporting, thereby increasing the speed and accuracy of initial response in the event of an accident. Therefore, the present invention overcomes the limitations of conventional monitoring methods that rely heavily on human resources and provides the effect of substantially enhancing on-site safety in various fields, such as crime prevention, ensuring traffic safety, and preventing industrial accidents. Brief explanation of the drawing

[0020] FIG. 1 is a block diagram showing an intelligent image detection system according to one embodiment of the present invention. Figure 2 is a photograph showing the process of detecting the risk level of a cluster of people through a cluster risk detection unit of an intelligent image detection system according to one embodiment of the present invention. Figure 3 is a photograph showing the appearance displayed on the screen of the central control center when a dangerous state of a cluster of people is detected through the cluster risk detection unit of an intelligent image detection system according to one embodiment of the present invention. Figure 4 is a photograph showing the process of detecting the degree of danger of a road through a road danger detection unit of an intelligent image detection system according to one embodiment of the present invention. Figure 5 is a photograph showing the process of detecting the degree of risk of a work condition through a work risk detection unit of an intelligent image detection system according to one embodiment of the present invention. Specific details for implementing the invention

[0021] Preferred embodiments of the present invention will be described in detail below with reference to the drawings. Prior to this, terms and words used in this specification and claims should not be interpreted as being limited to their ordinary or dictionary meanings, but should be interpreted in a meaning and concept consistent with the technical spirit of the present invention.

[0022] Throughout this specification, when it is stated that one component is located "on" another component, this includes not only cases where one component is in contact with another component, but also cases where another component exists between the two components. Throughout this specification, when it is stated that a part "includes" a component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0023] FIG. 1 shows a block diagram illustrating an intelligent image detection system according to one embodiment of the present invention.

[0024] Referring to FIG. 1, the intelligent video detection system (100) according to the present embodiment is equipped with a shooting unit (110) that performs a specific role, a danger state detection unit (120), and a safety warning unit (130), thereby enabling real-time monitoring of a situation where many people are densely packed within a preset area, real-time monitoring of vehicles or people or animals driving abnormally entering a preset road area, real-time monitoring of workers working in a dangerous work environment, and the ability to quickly generate a warning notification when a dangerous situation occurs to induce a manager to respond quickly.

[0025] Hereinafter, with reference to FIGS. 1 to 5, each component constituting the intelligent image detection system (100) according to the present embodiment will be described in detail.

[0026] Detailed explanation of the camera unit (110)

[0027] The shooting unit (110) is a core component that captures the monitoring target area in real time and can be implemented in various types, such as a fixed camera, a rotating camera, a thermal imaging camera, or an infrared camera for night shooting. The shooting unit (110) is configured with a fixed angle or a variable angle depending on the characteristics of the installation location, and can precisely capture a specific point by providing a zoom-in / zoom-out function when necessary.

[0028] The recording unit (110) is designed to enable high-resolution recording, thereby clearly capturing detailed images such as crowds of people, vehicles, and the movements of workers. Unlike conventional CCTVs which are limited to simple recording purposes, the recording unit (110) is intended to provide video data in a form suitable for use by subsequent analysis modules. Therefore, the video data is transmitted as is, or, if necessary, undergoes compression and codec conversion processes to increase transmission efficiency.

[0029] The imaging unit (110) may adopt a waterproof and dustproof structure so that it can operate stably in an outdoor environment depending on the installation location. In addition, an embodiment including a temperature control module or a heater device is possible to ensure normal operation even in extreme environments such as extreme heat or extreme cold. Such a structure supports long-term stable monitoring in various environments such as industrial sites, roads, and public places.

[0030] The camera unit (110) also has a variety of power supply methods. Generally, it is operated via a constant power connection, but in environments where it is difficult to connect to an external power source, it can receive power independently through a combination of a solar module and a battery. This enables continuous video recording even in areas with inadequate power infrastructure.

[0031] Consequently, the shooting unit (110) is designed to go beyond a simple video recording device to provide data quality that the danger state detection unit (120) can accurately analyze, and to respond stably to environmental changes, thereby effectively supporting the purpose of real-time safety management of the present invention.

[0032] Detailed description of the danger state detection unit (120)

[0033] The danger state detection unit (120) is a core analysis module that automatically identifies various types of danger based on images transmitted from the shooting unit (110). This module is subdivided into a cluster danger detection unit (120A), a road danger detection unit (120B), and a labor danger detection unit (120C), which independently analyze each situation and comprehensively determine whether there is a danger.

[0034] The danger state detection unit (120) applies image processing algorithms, artificial intelligence deep learning techniques, and pattern recognition technology to analyze behavior and state changes rather than simple shape distinctions. For example, it performs the function of determining whether it is normal by comprehensively analyzing human movement patterns, vehicle driving trajectories, and workers' work postures.

[0035] The danger state detection unit (120) can be combined with a GPU-based high-speed computing device to ensure real-time performance. Through this, it analyzes video data at multiple frames per second and transmits information to the safety warning unit (130) without delay when a dangerous situation occurs. Unlike the post-verification method of the prior art, the present invention enables immediate judgment and response.

[0036] In addition, the risk state detection unit (120) can simultaneously reflect a preset standard value and a policy value input by an administrator or a central control center. For example, it can be applied by setting a low personnel tolerance standard for a specific event or strengthening worker work standards in the event of bad weather, thereby enabling flexible, customized safety management.

[0037] As a result, the danger state detection unit (120) identifies various dangers in the image data in an automated manner, thereby reducing the dependence on management personnel of the existing system and enabling immediate response to dangers at the site, thereby providing a practical safety management effect.

[0038] Detailed description of the cluster risk detection unit (120A)

[0039] The cluster risk detection unit (120A) is a module that quantitatively analyzes the number of people within a specific area and determines whether there is a risk based on the state of density. It extracts images corresponding to people from the video through the person filtering module (A1) and counts them through the person count calculation module (A2). This allows for accurate determination of the number of people within the video.

[0040] The crowd risk detection unit (120A) tracks changes in the number of people over time through the person increase / decrease tracking module (A3). It can detect patterns in which the number of people rapidly increases or decreases during specific time periods, thereby enabling early prediction of situations where a large crowd gathers at a specific point. This function is effective for preventing accidents in large event venues, subway stations, stadiums, etc.

[0041] The cluster risk detection unit (120A) includes a personnel overcrowding detection module (A4) and warns in real time of situations where a threshold value is exceeded. In conventional technology, an administrator had to estimate whether there was personnel density simply by looking at the screen, but the present invention enables objective and rapid judgment through automatic calculation and prediction.

[0042] The crowd risk detection unit (120A) can continuously improve the accuracy of personnel detection through AI learning. By improving detection accuracy by considering various lighting conditions, overlap between people, and differences in the speed of movement, it provides reliable results even in real environments.

[0043] As a result, the crowd risk detection unit (120A) prevents in advance risks such as crushing, congestion, and collision accidents that may occur in situations where a large number of people gather, and supports the manager in responding immediately, thereby overcoming the limitations of qualitative judgment in conventional CCTV systems.

[0044] Detailed description of the personnel filtering module (A1)

[0045] The person filtering module (A1) performs the role of distinguishing only images corresponding to people from the video data input from the shooting unit (110). To this end, image processing algorithms and deep learning-based object recognition technology may be applied, and non-human objects such as vehicles, animals, and background objects are excluded based on the shape, movement, and body contours of the person.

[0046] The personnel filtering module (A1) is designed to take into account variables in various environments. For example, a trained dataset and pattern recognition technology are applied to reliably detect human objects even in dimly lit environments, weather conditions with snow or rain, or when people appear overlapping in a crowd. This design ensures high reliability in actual public places or industrial sites.

[0047] The personnel filtering module (A1) may also include a function for tracking human objects in multiple continuously input frames. This does not merely detect people in specific frames, but distinguishes the same person over time to prevent duplicate counting. Therefore, the accuracy of personnel count calculation is greatly improved.

[0048] In addition, the personnel filtering module (A1) is combined with a high-speed computing unit to support real-time processing. By utilizing GPU-based computing or a dedicated AI chipset, a large number of people can be processed simultaneously, which enables real-time analysis even in crowded areas such as large event venues or subway stations.

[0049] As a result, the personnel filtering module (A1) stably extracts only human objects from the captured video, thereby laying the foundation for the subsequent steps, the personnel count calculation module (A2) and the personnel count increase / decrease tracking module (A3), to operate accurately and efficiently.

[0050] Detailed description of the personnel count calculation module (A2)

[0051] The person count calculation module (A2) calculates the number of people present in a specific area by counting person objects separated through the person filtering module (A1). To this end, an algorithm that prevents duplicate counting by considering the size, location, and overlap of objects, and corrects camera angles and perspective effects may be applied.

[0052] The personnel count calculation module (A2) can calculate the number of personnel independently for each of multiple zones, not just fixed zones. For example, it can calculate the number of personnel by dividing them into seating areas of a stadium, specific sections of a subway platform, specific workspaces inside a factory, etc., thereby enabling specific and detailed risk management.

[0053] In addition, the personnel count calculation module (A2) is equipped with a real-time processing function, allowing the number of personnel at a specific point in time to be determined immediately. In conventional CCTV systems, the manager had no choice but to make a rough estimate by visually observing the screen, but the present invention enables objective and highly reliable judgment by immediately providing quantitative data.

[0054] The personnel count calculation module (A2) is linked to a database to record changes in the number of personnel by time period and can derive specific patterns based on past data. For example, by identifying in advance a tendency for the average number of personnel to increase during a specific time period, the manager can prepare in advance.

[0055] Consequently, the personnel counting module (A2) goes beyond a simple counting function and contributes to the establishment of zone-specific risk management and preemptive response strategies, and plays a role in supporting the core functions of the cluster risk detection unit (120A).

[0056] Detailed description of the personnel increase / decrease tracking module (A3)

[0057] The personnel increase / decrease tracking module (A3) performs the function of tracking the increase / decrease trend over time based on the number of personnel input in real time from the personnel calculation module (A2). It does not stop at simply checking the current number of personnel, but analyzes changes over a certain period of time to enable early detection of the possibility of risk occurrence.

[0058] The personnel increase / decrease tracking module (A3) detects cases where the number of people increases at a rate exceeding a certain threshold and warns of the possibility of sudden congestion occurring at a specific point. For example, it can detect in advance a situation where a large crowd suddenly gathers at the entrance of a performance venue, and in this case, immediately provide a notification to the safety warning unit (130).

[0059] In addition, the personnel increase / decrease tracking module (A3) can analyze cases where the number of personnel decreases rapidly. This allows for the early detection of mass movement due to evacuation situations or accidents, enabling managers to quickly grasp the situation at the site and respond.

[0060] The personnel increase / decrease tracking module (A3) is combined with a prediction algorithm to predict the number of personnel after a certain period of time based on the current rate of increase / decrease. This allows for warnings to be given in advance before an overcrowded state occurs, and significantly improves the accident prevention effect.

[0061] Consequently, the personnel increase / decrease tracking module (A3) does not simply record the results of personnel calculation, but rather performs the role of analyzing the pattern of change to identify potential future risks at an early stage, thereby overcoming the limitations of retrospective observation in conventional CCTV systems.

[0062] Detailed description of the overcrowding detection module (A4)

[0063] The overcrowding detection module (A4) performs the function of determining in real time when the number of people in a specific area exceeds a reference value based on data provided by the number increase / decrease tracking module (A3). In conventional CCTV systems, an administrator had to estimate whether there was overcrowding by directly observing the screen, but the present invention can detect the risk in an automated manner through this module.

[0064] The overcrowding detection module (A4) not only checks whether the standard number of people is exceeded, but also analyzes the rate of increase in people and past data patterns to predict the possibility of future overcrowding. For example, it can detect a situation where the number of people increases rapidly in minutes at a specific event venue and generate a warning signal before an overcrowding condition occurs.

[0065] This module can also set customized thresholds that reflect the characteristics of each zone. It allows for flexible adjustments, such as applying relatively low occupancy standards in confined indoor spaces and high standards in large outdoor spaces. This enables safety management tailored to the specific characteristics of the space.

[0066] The overcrowding detection module (A4) immediately transmits data to the safety warning unit (130) when a dangerous situation is detected. The safety warning unit induces an immediate response through on-site warning output or manager notification, and, in some cases, can quickly execute crowd control measures in conjunction with the central control center.

[0067] Ultimately, the overcrowding detection module (A4) plays a key role in finally connecting the analysis results of the cluster risk detection unit (120A) to a risk judgment, overcoming the subjective and ex post judgment method of conventional technology and providing an active safety management effect based on real-time prediction.

[0068] Detailed description of the road hazard detection unit (120B)

[0069] The road hazard detection unit (120B) is configured to automatically determine various hazard factors that may occur within a road area based on image data collected from the shooting unit (110). This detection unit analyzes the driving status of the vehicle, whether a traffic accident has occurred, whether pedestrians or animals have entered the road in real time, and transmits this information to the safety warning unit (130).

[0070] The road hazard detection unit (120B) separates transportation objects through the vehicle filtering module (B1) and determines whether dangerous driving is occurring by analyzing the vehicle's trajectory and speed in the abnormal driving vehicle detection module (B2). Subsequently, the traffic accident vehicle detection module (B3) detects collision or rollover situations, and the pedestrian filtering module (B4) and the abnormal pedestrian detection module (B5) detect pedestrians and animals entering the road.

[0071] Unlike conventional simple video recording methods, this detection unit can process complex risk factors simultaneously. For example, it can identify in real-time complex situations where a traffic accident occurs in one lane while a pedestrian enters another lane at the same time. This establishes a multi-layered response system for accident risks.

[0072] The road hazard detection unit (120B) is designed considering the characteristics of the road environment. It operates under various conditions such as tunnels, intersections, highways, and urban roads, and infrared-based recognition technology or image correction algorithms may be applied to minimize the influence of changes in illumination or weather.

[0073] Therefore, the road hazard detection unit (120B) solves the problem of the manager's reliance on visual inspection that existing CCTV monitoring had, and provides the effect of supporting traffic accident prevention and response to unexpected situations in real time.

[0074] Detailed description of the vehicle filtering module (B1)

[0075] The vehicle filtering module (B1) performs the role of extracting only vehicles or similar means of transportation from the image data within the road hazard detection unit (120B). A deep learning-based object recognition algorithm and a vehicle-specific dataset are utilized to distinguish vehicle objects from backgrounds, pedestrians, animals, etc. in the image input from the shooting unit (110).

[0076] The vehicle filtering module (B1) identifies vehicles with high accuracy by comprehensively analyzing the vehicle's size, contour, color, license plate location, headlight pattern, etc. In particular, it applies a model trained to separate individual vehicles even in congested situations or environments where multiple vehicles appear to overlap.

[0077] This module is extensible to recognize various modes of transportation. It can distinguish and independently filter objects such as motorcycles, buses, trucks, and bicycles, in addition to automobiles, thereby enabling precise monitoring of various risk factors in the traffic environment.

[0078] The vehicle filtering module (B1) can be equipped with high-speed computation capabilities to support real-time processing. It is designed to enable stable object extraction without performance degradation, even in situations where multiple vehicles enter simultaneously on a large-scale road network or highway.

[0079] Ultimately, the vehicle filtering module (B1) is a key component that provides basic data for the subsequent module of the road hazard detection unit to operate normally, thereby solving the vehicle detection error and delay problems of the conventional technology and enabling accurate and rapid road hazard judgment.

[0080] Detailed description of the abnormal driving vehicle detection module (B2)

[0081] The abnormal driving vehicle detection module (B2) is a key component that determines whether the vehicle is following a normal driving pattern based on transportation data extracted through the vehicle filtering module (B1). This module analyzes the vehicle's speed, driving trajectory, lane keeping status, driving direction, etc., in real time to detect cases where it deviates from the criteria. In conventional CCTV systems, abnormal driving could only be confirmed by an administrator directly monitoring the screen, but in the present invention, dangerous situations can be recognized immediately through automated analysis.

[0082] The abnormal driving vehicle detection module (B2) can determine traffic law violation situations in real time. For example, it automatically recognizes vehicles driving in the wrong direction, vehicles violating traffic signals, vehicles continuously deviating from their lanes, and vehicles repeatedly making sudden lane changes. This analysis goes beyond simply tracking the location of the vehicle and distinguishes between normal and abnormal conditions by comparing continuous driving patterns over time.

[0083] This module enables stable detection in various situations through AI-based driving pattern learning capabilities. For instance, it ensures high accuracy in real-world environments by learning diverse scenarios, such as sudden lane changes during high-speed driving on highways, irregular turns at urban intersections, and rapid speed reductions inside tunnels. In particular, it reduces the possibility of false detections through multi-layered learning that considers various vehicle sizes (buses, passenger cars, motorcycles, etc.) and driving conditions.

[0084] The abnormal driving vehicle detection module (B2) immediately transmits data to the safety warning unit (130) when a dangerous situation is detected. At the scene, the danger is reported to other drivers and pedestrians through warning notifications or guidance broadcasts, and the central control center provides notifications so that an administrator can respond quickly. In some cases, a response system may be activated by directly linking with the police or emergency rescue agencies.

[0085] Consequently, the abnormal driving vehicle detection module (B2) overcomes the limitations of the passive and reactive response of conventional technology and provides the effect of significantly reducing the possibility of accidents by identifying dangers on the road in real time. This contributes not only to the prevention of traffic accidents but also to the maintenance of traffic order and the strengthening of the social safety net.

[0086] Detailed description of the traffic accident vehicle detection module (B3)

[0087] The traffic accident vehicle detection module (B3) performs the role of identifying vehicles corresponding to traffic accident situations, such as collisions, rollovers, rear-end collisions, and sudden stops, among the vehicles separated through the vehicle filtering module (B1). In conventional CCTV systems, an administrator could only determine whether an accident had occurred by manually checking the screen, but the present invention can automatically detect the occurrence of an accident by analyzing the characteristics of video data.

[0088] This module detects changes in vehicle movement to determine whether a traffic accident has occurred. For example, situations where a vehicle suddenly brakes while traveling at a constant speed, or where its movement comes to a halt after a sudden change in trajectory, can be considered a collision or an accident. Additionally, changes in the vehicle's attitude—specifically, tilting or overturning—can also be recognized to determine if an accident has occurred. This analysis comprehensively utilizes pixel changes within the video and object location data.

[0089] The traffic accident vehicle detection module (B3) can identify the accident situation by analyzing the interaction between multiple vehicles. If two or more vehicles stop suddenly at the same time in a narrow space or their collision trajectories overlap, it recognizes this as a traffic accident. In particular, it can analyze complex situations such as multiple collisions or chain collisions in real time, so it can respond to complex accident situations that conventional CCTVs miss.

[0090] When an accident is detected, this module immediately transmits accident information to the safety warning unit (130). At the scene, warning signals and broadcasts are used to notify nearby drivers to prevent secondary accidents, and the central control center provides the location and situation of the accident to the manager in real time, enabling an immediate request for rescue. In addition, it can be linked with the emergency reporting module (134) to automatically transmit the accident situation to 119 or 112.

[0091] Therefore, the traffic accident vehicle detection module (B3) solves the problem of accident recognition delay that occurred in conventional technology and significantly improves the initial response speed, thereby providing the effect of minimizing human casualties and property damage caused by traffic accidents. This can be considered a core technical element that fundamentally strengthens road safety.

[0092] Detailed description of the pedestrian filtering module (B4)

[0093] The pedestrian filtering module (B4) performs the role of separating only human and animal objects, excluding vehicles, from the captured video. Since conventional CCTV systems merely displayed the entire screen, it was difficult to quickly distinguish pedestrians or animals entering the road. This module works in conjunction with the vehicle filtering module (B1) to exclude transportation objects and then detects pedestrian and animal objects in the remaining video.

[0094] The pedestrian filtering module (B4) detects pedestrians based on human-specific characteristics such as height, body contours, and gait patterns. Additionally, it recognizes the external characteristics of animals such as dogs, cats, and cows and classifies them as objects likely to enter the road. Through this, risk factors entering the road in various situations can be identified in advance.

[0095] This module utilizes AI learning models to enhance accuracy even in complex environments. It can reliably distinguish between pedestrians and animals even in the presence of environmental noise such as streetlights, shadows, rain, and snow, and enables accurate detection through object separation even when objects appear to overlap. This ensures stable operation under various conditions, including urban roads, rural roads, and tunnel entrances.

[0096] The pedestrian filtering module (B4) not only detects objects but also tracks the direction and speed of movement of specific objects to analyze the possibility of them entering the road. For example, if a person or animal moving toward the road rather than the sidewalk is detected, information is transmitted to the abnormal pedestrian detection module (B5), which leads to an immediate warning system.

[0097] Consequently, the pedestrian filtering module (B4) eliminates the reliance on visual observation of conventional CCTVs and provides the effect of preventing traffic accident risks by identifying pedestrians and animals that may enter the road in real time. This acts as a core foundation for the road hazard detection unit (120B) and effectively resolves the blind spot problem that occurred in conventional technology.

[0098] Detailed description of the abnormal gait detection module (B5)

[0099] The abnormal pedestrian detection module (B5) is a key component that analyzes pedestrian and animal objects separated by the pedestrian filtering module (B4) to determine whether the object is entering the road. In conventional CCTV systems, it was difficult to immediately identify pedestrians or animals entering the road because the administrator had to make a judgment by looking at the screen. This module automatically detects whether an object is entering the road by comprehensively analyzing the object's location, direction of movement, and speed of movement.

[0100] The abnormal walking detection module (B5) is based on technology that tracks the movement trajectory of pedestrians. For example, if a pedestrian moves toward the road in a section that is not a sidewalk or crosswalk area, it is identified as abnormal walking. In the case of animals, it also recognizes and warns of cases where there is a high probability of moving toward a vehicle traffic area. These functions are effective in recognizing unexpected situations in advance and preventing traffic accidents.

[0101] This module is designed to operate accurately under various environmental conditions. It reliably recognizes the outlines and movements of pedestrians even at night or in weather conditions such as fog, rain, or snow, and enables accurate identification through AI-based object tracking technology even in complex scenes where vehicles and pedestrians are mixed. This is a significant advantage that overcomes the visual limitations of conventional CCTVs.

[0102] When the abnormal pedestrian detection module (B5) detects a dangerous situation, it immediately provides a notification to the safety warning unit (130). At the site, the danger can be warned to pedestrians and drivers through warning broadcasts or traffic light control devices, and at the central control center, an administrator can quickly recognize the situation and respond. This multi-layered response significantly enhances road safety.

[0103] Ultimately, the abnormal pedestrian detection module (B5) resolves the problem of delayed identification of pedestrians and animals entering the road that occurred in conventional technology, and operates in a configuration that maximizes the effect of preventing traffic accidents through real-time automatic identification. This is a core module that enhances the overall reliability of the road hazard detection unit (120B).

[0104] Detailed description of the work risk detection unit (120C)

[0105] The labor risk detection unit (120C) is a module that determines whether a worker is working under safe conditions within a specific area based on video collected from the shooting unit (110) and external environment data. While conventional CCTVs were limited to simply capturing the presence of a worker, the present invention automatically detects the risk by comprehensively analyzing the working environment and the condition of the worker.

[0106] The work hazard detection unit (120C) utilizes temperature, humidity, illuminance, and weather data provided by the weather information detection module (C1). It compares this data with preset safety standards to determine whether the work environment is suitable. For example, in the event of dangerous weather conditions such as heatwaves or strong winds, it determines that the work is unsafe and issues a warning.

[0107] This detection unit recognizes the presence of workers within the area and whether they are actually performing work through the worker detection module (C3). By analyzing the worker's location, movement, and use of work tools, it can determine whether they are simply in a waiting state or working in an actual dangerous environment. This provides the manager with more specific response guidelines.

[0108] The work risk detection unit (120C) is linked with the hazardous work status judgment module (C4) to immediately transmit a notification to the safety warning unit (130) when a worker is working in an unsafe environment. The warning can be transmitted in various ways, such as on-site broadcasting, notifications to managers, and reports to the central control center, and if necessary, leads to an immediate work stoppage instruction.

[0109] Accordingly, the work risk detection unit (120C) provides an intelligent safety management function that simultaneously considers the work environment and the condition of the worker, which could not be solved by conventional technology. This has technical features that directly contribute to the prevention of industrial accidents and the protection of the worker's life.

[0110] Detailed description of the weather information detection module (C1)

[0111] The weather information detection module (C1) is configured to provide core input data to the labor risk detection unit (120C) and detects weather conditions in a specific area in real time. Since existing CCTVs only recorded the external environment as video, they could not directly detect elements important to the working environment, such as temperature, humidity, and illuminance. This module overcomes these limitations.

[0112] The weather information detection module (C1) measures temperature, humidity, wind speed, rainfall, illuminance, etc. through sensors. In addition, if necessary, it can collect accurate weather data by linking with an external weather API. This multi-data collection structure reflects the weather conditions at the site more precisely.

[0113] This module links the measured data with the work risk detection unit (120C) and compares it with safety standards. For example, if the illuminance drops below a certain standard, it determines that the work risk has increased, and warns of the risk of heatstroke if work is performed for a long time in a high-temperature and high-humidity environment. This structure provides significant benefits for industrial safety management.

[0114] The weather information detection module (C1) is implemented with a structure that is waterproof, dustproof, and durable depending on the installation environment. It can operate stably in various environments such as outdoor industrial sites, construction sites, and road work zones, and guarantees data reliability even with long-term use.

[0115] Ultimately, the weather information detection module (C1) detects external environmental information in real time that conventional CCTVs could not provide, and is utilized as an important criterion for judging the safety of workers. Through this, the suitability of the work environment can be automatically determined, and preemptive response for accident prevention is enabled.

[0116] Detailed explanation of the suitability determination module (C2)

[0117] The suitability determination module (C2) performs the function of automatically determining whether the area is suitable for worker work based on various environmental data, such as temperature, humidity, illuminance, and wind speed, provided by the weather information detection module (C1). Conventional CCTV systems could not directly evaluate the safety of environmental conditions because they simply recorded video, but this module determines work suitability in real time by comparing it with industrial safety regulations or safety standards set by work managers.

[0118] This module is fundamentally based on thresholds pre-entered by the manager or the central control center. For example, it sets the environment as unusable if the temperature exceeds a certain standard, or identifies it as a hazardous environment if the illuminance drops below a certain level. Furthermore, different thresholds can be applied depending on the characteristics of the work site, enabling the implementation of customized judgment systems suitable for high-temperature and low-temperature workplaces, as well as outdoor and indoor environments.

[0119] The suitability determination module (C2) does not stop at simply comparing current values ​​but also considers the trend of data change. For example, if the temperature is continuously rising or the humidity is rapidly increasing, it can determine that there is a high probability of danger occurring even if the threshold has not yet been exceeded, and issue a preliminary warning. This enables preemptive response before an accident occurs.

[0120] This module can also be combined with AI-based prediction algorithms. By analyzing historical weather data and work environment records, it learns hazardous situations that may occur when specific conditions are combined, and based on this, can predict the suitability of work for the next few hours in advance. This capability enables work managers to establish more proactive safety measures.

[0121] Ultimately, the suitability determination module (C2) goes beyond simple monitoring and rapidly determines whether the working environment is safe through real-time analysis and prediction of environmental data. This overcomes the limitations of environmental risk assessment inherent in conventional CCTV systems and provides an important function that guarantees the life and safety of workers.

[0122] Detailed description of the worker detection module (C3)

[0123] The worker detection module (C3) is configured to determine whether a worker is present within a preset area and whether the worker is actually performing work. While conventional CCTVs could only confirm the presence of a person by simply capturing video, this module precisely analyzes the worker's location, behavioral patterns, and use of work tools through video analysis. Through this, it is possible to identify not only whether a person is present but also whether actual work activities are taking place.

[0124] This module tracks workers' physical movements by applying a deep learning-based human body recognition algorithm. It can accurately determine whether work is being performed by learning work actions that are repeatedly carried out in specific areas (e.g., bending, lifting, using tools, etc.). Additionally, it may include functions to detect workers' attire, whether safety helmets are being worn, and whether protective equipment is being used, thereby enabling verification of compliance with work safety regulations.

[0125] The worker detection module (C3) recognizes each worker individually even when multiple workers are present simultaneously. It improves recognition accuracy by utilizing object separation technology even when they appear to overlap. This means that the status of each worker can be monitored independently even when multiple tasks are being performed simultaneously in a large-scale workplace.

[0126] In addition, this module links real-time collected data with the suitability determination module (C2). For example, if a worker continues to perform work in a situation where weather conditions are determined to be dangerous, this can be identified as a hazardous work state. Through such organic inter-module linkage, it can function as an actual safety management system, going beyond simple recognition functions.

[0127] Consequently, the worker detection module (C3) goes beyond simply detecting the presence of a person to detect whether work is being done and whether safety equipment is being worn, thereby enhancing safety at the industrial site. This overcomes the limitations of conventional CCTVs and plays an essential role in implementing a worker-tailored safety management system.

[0128] Detailed explanation of the hazardous work status determination module (C4)

[0129] The hazardous work status determination module (C4) performs the function of finally determining whether a worker is working in a hazardous environment by combining environmental suitability data provided by the suitability determination module (C2) and worker work status data from the worker detection module (C3). In conventional technology, the work environment and the worker's status had to be checked separately, and a manager had to manually combine them, but the present invention automatically integrates and analyzes them.

[0130] This module immediately identifies hazardous conditions, such as when a worker continues to operate during a heatwave or handles heavy equipment in low-light conditions. It can accurately detect actual "hazardous work situations" that cannot be identified simply by the presence of a dangerous environment or a worker. This represents a level of precise safety management capability that conventional CCTVs could not achieve.

[0131] The dangerous work status determination module (C4) immediately transmits data to the safety warning unit (130) when a dangerous situation is detected. At the site, workers can be directly warned through methods such as warning broadcasts, warning lights, and work stop signals, and the central control center receives notifications so that a manager can respond immediately. In some cases, automatic reporting to external emergency rescue agencies such as 119 is also possible.

[0132] This module goes beyond simply warning of hazardous conditions to evaluate the degree of risk in multiple stages. For example, it adjusts the intensity of the response by classifying levels into "Caution," "Warning," and "Immediate Stop." This reduces confusion caused by unnecessary warnings and enables a more rapid and powerful response in actual emergency situations.

[0133] Consequently, the hazardous work condition judgment module (C4) performs the final judgment function to practically guarantee the safety of workers. This overcomes the limitations of conventional CCTVs, which were limited to simple monitoring, and enables the implementation of a practical industrial safety management system by comprehensively considering the work environment and work activities.

[0134] Detailed explanation of the safety warning unit (130)

[0135] The safety warning unit (130) is a core component that receives a danger signal transmitted from the danger state detection unit (120) and outputs various forms of warnings and notifications corresponding to it. In conventional CCTV systems, an administrator had to manually take action after directly viewing the screen and recognizing an abnormal situation, but the present invention drastically reduces response time by establishing an automatic warning system through the safety warning unit (130).

[0136] The safety warning unit (130) has a structure capable of simultaneously controlling a warning device installed at the site and an output device of the central control center. That is, when a danger is detected in a specific area, it immediately provides a warning signal directly to workers, pedestrians, and drivers at the site, while simultaneously providing a notification to the central control center so that a manager can respond quickly. Through this, the response to danger is not limited to a single path but is achieved in a multi-layered structure.

[0137] This configuration is designed with a modular structure, allowing specific output modules to be configured to operate depending on the on-site situation. For example, in a simple overcrowding situation, only on-site broadcasting is activated, while in the event of a traffic accident, on-site notifications, central control center notifications, and emergency reporting functions are activated simultaneously. This enables customized responses for each situation.

[0138] The safety warning unit (130) is also designed to be network-based and supports both wired and wireless communication. The communication network can be redundant or wireless networks such as LTE and 5G can be used in parallel so that signal transmission between the field device and the central control center can be made without delay. Through this, warnings are reliably transmitted even in unexpected situations.

[0139] Consequently, the safety warning unit (130) plays a practical central role in connecting the risk detection results to actual response rather than merely being data. This overcomes the passive limitations of conventional technology and establishes a real-time automatic response system, thereby providing the effect of minimizing casualties and property damage.

[0140] Detailed description of the field warning output module (131)

[0141] The on-site warning output module (131) is installed adjacent to the monitoring target location and is configured to receive a control signal from the safety warning unit (130) and immediately output a warning signal at the site. This module directly delivers a warning to users and workers in the area where the danger occurred, thereby inducing immediate evacuation or cautionary action.

[0142] The on-site warning output module (131) may include various output means. For example, a warning light, a siren, an electronic display, a voice broadcasting device, etc. In certain situations, only a simple visual warning may be provided, and in emergency situations, a siren and voice broadcasting may be used simultaneously to induce rapid evacuation of personnel at the site.

[0143] This module can support multilingual voice guidance or visual icon display functions to enhance user understanding. For example, in workplaces with a large number of foreign workers, it can be configured to output announcements in multiple languages, such as English and Chinese, in addition to Korean, and can provide visual warning signals for the hearing impaired.

[0144] The field warning output module (131) may include its own power supply system or be equipped with an emergency battery to maintain the warning function for a certain period of time even during a power outage. This ensures that the warning function can continue even if the power supply is cut off in a dangerous situation.

[0145] Therefore, the on-site warning output module (131) directly conveys the dangerous situation to the on-site user, thereby securing immediate on-site response capabilities that conventional CCTVs could not provide. This contributes decisively to the prevention of safety accidents and plays an important role in minimizing damage in emergency situations.

[0146] Detailed description of the control center output module (132)

[0147] The control center output module (132) is installed in the central control center and is configured to receive a control signal from the safety warning unit (130) and output various types of notifications so that the manager can respond quickly. In conventional CCTV systems, the manager had to manually view the screen to identify abnormal situations, but this module automatically receives a danger detection signal, allowing the manager to respond immediately.

[0148] This module is integrated with the central control center's monitors, warning notification windows, and broadcasting systems to provide simultaneous visual and auditory alerts in the event of a dangerous situation. For instance, if the area is overcrowded, it outputs a warning window and sound; if a traffic accident occurs, the location is displayed on a map and the video is automatically zoomed in. This allows administrators to intuitively grasp the situation.

[0149] The control center output module (132) can output situation-specific customized messages. It outputs a caution message in simple warning situations and an immediate response instruction message in emergency situations to help the manager distinguish priorities. This significantly improves the response efficiency of the central control center.

[0150] This module can also be integrated with administrator terminals or mobile devices. Even when an administrator is absent from the central control center or away for field inspections, risk alerts can be received in real-time on mobile devices, minimizing response gaps. This fundamentally resolves the "response delay in the absence of an administrator" problem that occurred in conventional technologies.

[0151] Ultimately, the control center output module (132) performs the function of enabling the manager of the central control center to immediately grasp the situation and respond. This can be considered a core module that overcomes the limitations of passive observation in conventional CCTV systems and connects dangerous situations to a practical response system.

[0152] Detailed description of the wireless transmission module (133)

[0153] The wireless transmission module (133) performs the function of immediately transmitting a notification to a mobile phone or portable terminal held by a central control center manager after receiving a control signal from the safety warning unit (130). Conventional CCTV systems could not deliver immediate notifications to the manager, so there was a problem of delayed response when the manager was not present, but the present invention overcomes this limitation through the wireless transmission module (133).

[0154] The wireless transmission module (133) can support various communication methods. It can provide not only general SMS text notifications, but also push notifications via a dedicated application, email integration, and real-time data transmission functions based on LTE and 5G. Through this, the manager can always recognize dangerous situations in real time, regardless of whether they are at the site or outside.

[0155] This module can also apply differential notification methods based on the level of risk. For example, it provides only standard push notifications in simple overcrowding situations, while providing emergency alerts accompanied by mandatory sounds or vibrations in situations involving traffic accidents or critical worker risks. This multi-layered notification system helps managers intuitively perceive the severity of the situation and respond immediately.

[0156] The wireless transmission module (133) can apply a communication method that enhances security. The transmitted data is encrypted to minimize the possibility of external hacking or tampering, and only designated recipients can receive notifications after undergoing terminal authentication by the administrator. This is an important factor in ensuring the reliability of the safety alarm system.

[0157] Consequently, the wireless transmission module (133) enables the central control center manager to perceive dangerous situations in real time at the same level as the field, regardless of where the manager is located, thereby resolving the problem of response delay in conventional technology. This eliminates response gaps caused by the absence of the manager and enables a more continuous and flexible safety management system.

[0158] Detailed description of the emergency reporting module (134)

[0159] The emergency reporting module (134) receives a control signal from the safety warning unit (130) and performs the function of automatically reporting to a pre-set emergency rescue agency (e.g., 119, 112). In conventional CCTV systems, when an accident occurs, the administrator must make a phone call to report it, so there were frequent cases of initial response failure due to delays in reporting. This module overcomes these limitations and enables a rapid rescue request.

[0160] The emergency reporting module (134) immediately transmits the situation to an emergency rescue agency when a dangerous situation is detected. The transmitted information may include location information, type of accident, captured video images of the scene, degree of danger, etc. For example, in the event of a traffic accident, the GPS coordinates of the location and an image of the accident vehicle are automatically transmitted, and in the event of a worker accident, a report is made along with environmental data of the workplace.

[0161] This module goes beyond simple reporting and can also receive response signals from rescue agencies after the report and transmit them to the central control center. For example, when a dispatch from 119 is confirmed, the relevant information is immediately displayed to the manager, which can reduce unnecessary duplicate responses. This has the effect of strengthening the cooperation system between the field and rescue agencies.

[0162] The emergency reporting module (134) can communicate in various ways. It can be linked with a wired telephone network, utilize an internet-based VoIP or wireless communication network, and can be designed to continue making emergency reports via a satellite communication module even when the power supply is cut off. This multi-communication network support structure ensures that the report does not fail under any circumstances.

[0163] Therefore, the emergency reporting module (134) automatically reports to the rescue agency without manual intervention by the manager when a dangerous situation occurs, thereby resolving the structural limitations of the conventional CCTV system. This provides the important effect of securing the golden time immediately after an accident occurs and minimizing casualties and property damage.

[0164] As explained above, the intelligent video detection system of the present invention can resolve the problem of real-time response delays caused by the excessive reliance of conventional CCTV monitoring systems on the administrator's visual observation. By combining a recording unit and a danger state detection unit to perform automatic analysis based on video data, it provides a system that enables the administrator to immediately recognize and warn of dangerous situations even if they miss a screen or lose focus. This enhances real-time responsiveness and effectively overcomes the structural limitations of existing systems.

[0165] In addition, the present invention can automatically calculate whether there is overcrowding of multiple people through a crowd risk detection unit and predict risks in advance by analyzing trends in the increase or decrease of the number of people. In conventional technology, since managers had to detect crowd density situations simply by watching a screen, it was difficult to immediately determine when a threshold was exceeded. The present invention provides an automatic detection function based on quantitative data, thereby significantly reducing the possibility of safety accidents in large event venues or congested transportation facilities.

[0166] The present invention rapidly detects abnormally moving vehicles, vehicles involved in traffic accidents, and pedestrians or animals entering the road through a road hazard detection unit. While conventional CCTVs required an administrator to manually monitor the screen to identify such situations, the present invention automatically generates warnings by determining traffic conditions in real-time from video data. This enhances the ability to respond to sudden accident risks and ensures the effectiveness of traffic accident prevention.

[0167] Furthermore, the present invention can determine the safety status of workers in industrial sites in real time through a work hazard detection unit. It automatically evaluates the suitability of working conditions by comprehensively analyzing environmental information such as weather, temperature, humidity, and illuminance, and issues a warning if a worker performs work under unsuitable conditions. Conventional technology had limitations in that it merely recorded images of workers without recognizing the presence of hazardous environments; however, the present invention overcomes this limitation, providing a substantial effect in preventing industrial accidents.

[0168] The safety warning unit dramatically improves the response system by providing multi-layered functions such as on-site warning output, central control center broadcasting, administrator mobile phone notifications, and emergency reporting. In conventional technology, administrators had to respond manually after personally verifying dangerous situations, but since the present invention automatically provides warnings and notifications simultaneously with the detection of danger, the speed of initial response can be significantly increased. This is a key advantage that minimizes the spread of damage in emergency situations.

[0169] Furthermore, the present invention reduces reliance on management personnel and ensures operational efficiency through an automated system. Conventional CCTV monitoring required multiple managers to work in shifts, but the present invention enables effective management of large areas with a small number of personnel through an automatic detection and warning system. This contributes not only to reducing personnel operating costs but also to maintaining consistency in management quality.

[0170] Ultimately, the present invention resolves the problems associated with conventional CCTV monitoring systems, such as delays in real-time response, limitations in determining overcrowding, insufficient prevention of traffic accidents, inadequate determination of working environment safety, and excessive reliance on manpower. Through this, it provides an intelligent video detection system that significantly improves on-site safety in various areas, such as crime prevention, traffic safety, and industrial accident prevention, and enables managers to respond quickly to dangerous situations.

[0171] The above detailed description of the present invention describes only specific embodiments thereof. However, it should be understood that the present invention is not limited to the specific forms mentioned in the detailed description, but rather should be understood to include all variations, equivalents, and substitutions within the spirit and scope of the invention as defined by the appended claims.

[0172] In other words, the present invention is not limited to the specific embodiments and descriptions described above, and any person skilled in the art to which the present invention pertains can make various modifications without departing from the essence of the invention as claimed in the claims, and such modifications fall within the scope of protection of the present invention. Explanation of the symbols

[0173] 100: Intelligent Video Detection System 110: Filming Department 120: Danger state detection unit 120A: Cluster Risk Detection Unit A1: Personnel filtering module A2: Personnel Count Calculation Module A3: Personnel Increase / Decrease Tracking Module A4: Overcrowding detection module 120B: Road hazard detection unit B1: Vehicle filtering module B2: Abnormal driving vehicle detection module B3: Traffic Accident Vehicle Detection Module B4: Pedestrian filtering module B5: Abnormal gait detection module 120C: Labor Hazard Detection Unit C1: Weather Information Detection Module C2: Suitability determination module C3: Worker Detection Module C4: Hazardous Work Status Determination Module 130: Safety Warning Section 131: Field Warning Output Module 132: Control Center Output Module 133: Wireless transmission module 134: Emergency Reporting Module

Claims

Claim 1 A shooting unit (110) installed adjacent to a location to be monitored and capturing the location to be monitored in real time and transmitting it to a danger state detection unit (120); a danger state detection unit (120) equipped with a cluster risk detection unit (120A) that detects the risk level of a cluster of people within a preset zone based on image data transmitted from the shooting unit (110), a road risk detection unit (120B) that detects the risk level of a road within a preset zone, and a work risk detection unit (120C) that detects the risk level of a work state within a preset zone; The system includes a safety warning unit (130) installed at a location to be monitored and a central control center, which outputs a preset warning signal when a dangerous state is detected by a danger state detection unit (120); and the cluster danger detection unit (120A) of the danger state detection unit (120) comprises: a personnel filtering module (A1) that filters only images corresponding to people from images within a preset area; a personnel count calculation module (A2) that calculates the number of people from images filtered by the personnel filtering module; and a personnel count increase / decrease tracking module (A3) that tracks the number of people changing in real time from the personnel count calculation module (A2) and monitors the increase / decrease status of the number of people over time. It includes a person overcrowding detection module (A4) that predicts the time when the number of people exceeds a reference value based on data detected through the person increase / decrease tracking module (A3), determines a dangerous situation due to the person overcrowding state, and transmits the information to the safety warning unit (130) in real time if a dangerous situation occurs; and the road danger detection unit (120B) of the danger state detection unit (120) comprises: a vehicle filtering module (B1) that filters only images corresponding to means of transportation including automobiles from images within a preset area; and an abnormal driving vehicle detection module (B2) that detects a means of transportation driving abnormally from the images filtered by the vehicle filtering module (B1) and transmits the information to the safety warning unit (130) in real time.A traffic accident vehicle detection module (B3) that detects a means of transport involved in a traffic accident in an image filtered by the vehicle filtering module (B1) and transmits the information in real time to a safety warning unit (130); and a pedestrian filtering module (B4) that filters only images corresponding to animals and people, excluding means of transport, from an image within a preset area. The system includes an abnormal pedestrian detection module (B5) that determines whether an animal or person detected in an image filtered by the pedestrian filtering module (B4) enters the road and transmits the information in real time to the safety warning unit (130); and the work risk detection unit (120C) of the risk state detection unit (120) comprises: a weather information detection module (C1) that detects weather information, temperature information, humidity information, and illuminance information in real time within a preset area; a suitability determination module (C2) that determines whether the conditions are suitable for outdoor work by comparing data detected from the weather information detection module (C1) after setting conditions suitable for outdoor work within the preset area using a manager or AI; and a worker detection module (C3) that detects whether a worker exists within the preset area and detects whether a worker is performing work within the preset area. The system includes a dangerous work state judgment module (C4) that transmits information in real time to a safety warning unit (130) when a worker is performing work within a preset area despite the conditions not being suitable for outdoor work, as determined by the suitability judgment module (C2); wherein the safety warning unit (130) comprises: a field warning output module (131) that is installed adjacent to the location to be monitored, operates by receiving an operation control signal from the safety warning unit (130), and outputs preset warning signals, evacuation signals, guidance broadcasts, and recommendation broadcasts; and a control center output module (132) that is installed in the central control center, operates by receiving an operation control signal from the safety warning unit (130), and outputs preset warning signals, rapid response guidance signals, and response recommendation broadcasts to the manager.An intelligent video detection system characterized by comprising: a wireless transmission module (133) installed in a central control center, receiving an operation control signal from a safety warning unit (130), and transmitting an emergency situation notification signal to a mobile phone held by an administrator; and an emergency reporting module (134) installed in a central control center, receiving an operation control signal from a safety warning unit (130), and reporting to 119 or 112.; Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete

Images