WO2024101726A1 - Plateforme embarquée de reconnaissance d'images et de gestion de sécurité intégrée comprenant un système de reconnaissance d'images basée sur l'intelligence artificielle - Google Patents

Plateforme embarquée de reconnaissance d'images et de gestion de sécurité intégrée comprenant un système de reconnaissance d'images basée sur l'intelligence artificielle Download PDF

Info

Publication number
WO2024101726A1
WO2024101726A1 PCT/KR2023/016700 KR2023016700W WO2024101726A1 WO 2024101726 A1 WO2024101726 A1 WO 2024101726A1 KR 2023016700 W KR2023016700 W KR 2023016700W WO 2024101726 A1 WO2024101726 A1 WO 2024101726A1
Authority
WO
WIPO (PCT)
Prior art keywords
unit
image recognition
cnn
recognition
video
Prior art date
Application number
PCT/KR2023/016700
Other languages
English (en)
Korean (ko)
Inventor
강일형
최평호
김도근
노동원
유재곤
장용준
이정욱
Original Assignee
주식회사 영신
주식회사 경우시스테크
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 주식회사 영신, 주식회사 경우시스테크 filed Critical 주식회사 영신
Publication of WO2024101726A1 publication Critical patent/WO2024101726A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/08—Construction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/292—Multi-camera tracking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00—Burglar, theft or intruder alarms
    • G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
    • G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
    • G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
    • G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/02—Alarms for ensuring the safety of persons
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N7/00—Television systems
    • H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast

Definitions

  • the present invention relates to an embedded image recognition safety integrated control platform including an artificial intelligence-based image recognition system. More specifically, the present invention relates to an embedded image recognition safety integrated control platform including an artificial intelligence-based image recognition system, and more specifically, an image data receiver that receives image capture data around a construction or industrial site from a camera module; A CNN-based object recognition unit that recognizes a human object in the video shooting data received from the video data receiver, boxes the recognized human object in response to the data of the recognized human object, and detects the CNN-based object recognition unit.
  • An image recognition system including an event information generator that generates a warning light or a warning alarm when a human object detected from is exposed to a preset risk radius and situation including around heavy equipment;
  • a false positive/no positive classification unit that receives event information detected from the CNN-based object recognition unit and classifies it as a true positive, false positive, or false positive, and a deep learning-based object recognition learning algorithm that classifies the false positive or false positive received from the false positive/unpositive classification unit.
  • a re-learning unit that re-trains in real time through YOLO (You Only Look Once), a learning model update unit that uploads the re-learning data set generated from the re-learning unit to the CNN-based object recognition unit in real time online, and the image recognition unit. It relates to an embedded video recognition safety integrated control platform that includes an artificial intelligence-based video recognition system consisting of an AI video recognition platform that includes an LTE communication unit for communication between the system and external devices, including external servers or administrator terminals.
  • the average industrial accident death rate in OECD member countries is 2.43 per 100,000 people, with Korea (3.61) being the fifth highest after Canada (5.84), Turkey (5.17, as of 2016), Chile (4.04), and Luxembourg (3.69).
  • Korea the number of deaths from industrial accidents in the '3050 Club', with a population of more than 50 million and per capita income exceeding $30,000, is Korea (3.61), Japan (1.50), the United States (3.36), the United Kingdom (0.88), and France (2.18). ), Germany (1.03), and Italy (2.10), Korea ranks first.
  • Korea's industrial accident death rate is higher than other countries.
  • the government is also implementing various policies to prevent industrial disasters and safety accidents that continuously occur.
  • the Occupational Safety and Health Act (Industrial Safety Act) Amendment Act of 2022 which significantly strengthens safety regulations at industrial sites, including preventing 'outsourcing of risks'. It went into effect on January 27, 2018, and requires management to be sentenced to prison when a worker dies or has an accident, and a punitive damages system has also been introduced to enforce the obligation of employers and corporations to ensure safety and health due to intentional or gross negligence.
  • the Serious Accident Punishment Act is being implemented on January 27, 2022, which stipulates that if a serious disaster occurs or damages are caused in violation of the Act, the person is liable for compensation up to 5 times the amount of damage.
  • the tag-based technology is a method in which the vehicle and worker have a sensor tag that can directly transmit and receive radio waves to measure the distance between heavy equipment and workers. It is divided into one-way tag and two-way tag methods depending on the radio wave transmission and reception method of the tag, and two-way tag method.
  • Mainly used sensors include RF and UWB, and a method of measuring the radio wave strength (RSSI) and radio time of arrival (TOA) received by the sensor and converting them to distance is used.
  • RSSI radio wave strength
  • TOA radio time of arrival
  • RF approach warning products are mainly mainstream.
  • UWB products with better performance in terms of measurement distance accuracy and uniformity have been developed and are being commercialized.
  • the one-way tag method includes a PASSIVE type product that uses an RFID tag.
  • An RFID reader is installed on the vehicle side, and the pedestrian tag has a directional PASSIVE type RFID tag.
  • This type of product is a two-way tag type product. Compared to other products, there was a problem in that the reliability of the product was low due to the large distance recognition error.
  • the non-tag-based approach warning uses simple sensors such as non-tag-based cameras and ultrasonic waves, and due to the limitations of the non-tag-based sensor's own characteristics and limited functionality in terms of operation, radar and optical technology using wireless radio signals are used. LiDAR products using are being developed, and recently, with the development of AI technology, object image recognition products using cameras are also being developed.
  • Korea Registered Patent No. 10-1808587 shows that it can rotate 360 degrees. and a video input unit that includes a PTZ (Pan/Tilt/Zoom) camera or a fixed camera with built-in top, bottom, left, right and zoom functions, and the video captured from the video input unit is processed according to a preset abnormal situation algorithm. Invasion, crowding, wandering, abandonment, emergence, entry, Waltham.
  • PTZ Pan/Tilt/Zoom
  • An abnormal situation detection unit that detects whether an abnormal situation is selected from people count, collapse, reverse driving, and number recognition, and an image pre-processing process and an object extraction image generation process when an abnormal situation is detected by the abnormal situation detection unit.
  • Object recognition is performed through an object analysis process, and the object analysis process performs edge pattern extraction using the Haar algorithm, HOG algorithm, or SURF algorithm depending on the abnormal situation for the object extracted in the object extraction image generation process.
  • an object recognition unit that performs object recognition and discrimination through pattern matching with learned data accumulated and stored through a deep learning algorithm, and an object recognition unit that analyzes changes in coordinates of objects recognized by the object recognition unit and identifies them within the captured image.
  • An object tracking unit that predicts the movement path or direction of the object or performs tracking so that the object is located at the center of the captured image, and an image input unit that displays and monitors the image captured from the image input unit.
  • An intelligent integrated surveillance and control system using object recognition, tracking monitoring, and abnormal situation detection technology has been developed, which consists of an integrated control unit that sets and controls an abnormal situation detection unit, an object recognition unit, and an object tracking unit. there is.
  • Korean Patent No. 10-2185859 (registration date: November 26, 2020) describes an object tracking device using deep learning that recognizes human objects from video data and tracks the human objects on a frame-by-frame basis.
  • a video data receiving unit that receives the video data;
  • a pre-processing unit that resizes the received image data and reduces the influence of light;
  • An object recognition unit that recognizes human objects in pre-processed image data through deep learning-based object recognition learning and boxes the recognized human objects in response to the data of the recognized human objects;
  • the data match between the boxed human object in the first frame portion and the boxed human object in the second frame portion following the first frame is calculated, and a box showing a matching degree higher than the set match value is selected as the same human object.
  • Korea Patent No. 10-2206662 (registration date: January 18, 2021) includes multiple cameras installed in each area of the port gate, yard/block, block entrance, ARMGC, and QC at the port container terminal; And, by receiving the plurality of camera image data, the deep learning module detects objects in the camera image and recognizes characters according to the learning data, and performs lane recognition, vehicle number recognition, container number (ISO code) character recognition, and container damage recognition. , Recognition of vehicles entering block entrances, vehicles and workers entering dangerous areas, vehicles traveling in reverse, detection of whether yard workers are wearing safety gear/safety vests, detection of loading and unloading equipment locations, extraction of learned objects, marking with events (text) and square boxes.
  • TLEM FPGA-based embedded vision system
  • ISO vehicle number and container number
  • Korea Patent No. 10-2263512 (registration date: June 4, 2021) provides that in the IoT integrated intelligent video analysis platform system that integrates and analyzes video data and non-video data, video data that acquires at least one video data acquisition department; a non-image data acquisition unit that acquires at least one non-image data; an image data processing unit that analyzes the image data; a non-image data processing unit that analyzes the non-image data;
  • the video data processing unit or the non-video data processing unit determines that an abnormal situation is present based on the video data or the non-video data
  • the video data processing unit includes an integrated data determination unit that finally determines the abnormal situation, wherein the video data processing unit Recognizing an object from image data, estimating the state of the object, estimating the authenticity of the object, or estimating an action event of the object, and the non-image data processing unit analyzes the non-image data.
  • the processing unit includes an object processing unit that processes a function of recognizing an object from the acquired image data; It further includes a user learning setting unit that provides functions related to machine learning of image data to the user, wherein the object processing unit extracts an object from the image data, and includes an object authenticity identification unit that determines whether or not it is forged; an object state recognition unit that estimates the state of the object from the image data; It further includes an object action recognition unit that estimates an action event of an object from the image data, wherein the object authenticity identification unit extracts an image from the image data, analyzes the colors of pixels constituting the extracted image, and analyzes the analyzed colors.
  • the probability that the object is genuine is derived through an authenticity judgment algorithm, and clustering is performed using a similarity-based clustering algorithm between data to extract the color ratio through the K-mean Clustering algorithm (Minimize the variance between clusters, identify the color ratio within the item from the clustered colors, extract the color ratio through OpenCV, learn the color scale for the authentic product, and identify the authentic product according to the resulting color ratio. It is possible to distinguish the difference between fake and fake images, generates a replica image using the DCGAN (Deep Convolution Generative Adversarial Network) algorithm, and uses the difference in surface material to provide feedback and learning between the genuine and fake models to read the genuine article.
  • An IoT integrated intelligent video analysis platform system capable of smart object recognition has been developed, which is characterized by deriving the probability that an object is genuine by applying a counterfeit product reading algorithm corresponding to the learning model.
  • the present inventors have overcome the limitations of the existing image recognition proximity warning system and improved the visibility of construction equipment operators to accurately monitor blind spots around the equipment in real time, making it suitable for industrial sites.
  • the present invention is intended to solve the above-described conventional problems, and includes an image data receiving unit that receives image shooting data around a construction or industrial site from a camera module; A CNN-based object recognition unit that recognizes human objects in the video shooting data received from the video data receiver, boxes the recognized human objects in response to the data of the recognized human objects, and detects them; and a CNN-based object recognition unit.
  • An image recognition system including an event information generator that generates a warning light or a warning alarm when a human object detected from is exposed to a preset risk radius and situation including around heavy equipment;
  • a false positive/no positive classification unit that receives event information detected from the CNN-based object recognition unit and classifies it as a true positive, false positive, or false positive, and a deep learning-based object recognition learning algorithm that classifies the false positive or false positive received from the false positive/unpositive classification unit.
  • a re-learning unit that re-trains in real time through YOLO (You Only Look Once), a learning model update unit that uploads the re-learning data set generated from the re-learning unit to the CNN-based object recognition unit in real time online, and the image recognition unit. It is a technical technology to provide an embedded video recognition safety integrated control platform that includes an artificial intelligence-based video recognition system consisting of an AI video recognition platform that includes an LTE communication unit for communication between the system and external devices, including external servers or administrator terminals. Make it an assignment.
  • the present invention includes an image data reception unit that receives image capture data around a construction or industrial site from a camera module; A CNN-based object recognition unit that recognizes human objects in the video shooting data received from the video data receiver, boxes the recognized human objects in response to the data of the recognized human objects, and detects them; and a CNN-based object recognition unit.
  • An image recognition system including an event information generator that generates a warning light or a warning alarm when a human object detected from is exposed to a preset risk radius and situation including around heavy equipment;
  • a false positive/no positive classification unit that receives event information detected from the CNN-based object recognition unit and classifies it as a true positive, false positive, or false positive, and a deep learning-based object recognition learning algorithm that classifies the false positive or false positive received from the false positive/unpositive classification unit.
  • a re-learning unit that re-trains in real time through YOLO (You Only Look Once), a learning model update unit that uploads the re-learning data set generated from the re-learning unit to the CNN-based object recognition unit in real time online, and the image recognition unit.
  • An embedded video recognition safety integrated control platform including an artificial intelligence-based video recognition system consisting of an AI video recognition platform including an LTE communication unit for communication between the system and external devices, including external servers or administrator terminals, is a technical solution. do.
  • a preset risk radius and situation including the surroundings of the heavy equipment is set as a virtual boundary or virtual area in the video shooting data received from the image data receiver, and appearance, entry, and exit within the virtual boundary or virtual area are determined. , including falling.
  • the re-learning data set is created by labeling video recording data according to the false positive or false positive classification using an auto-labeling tool, then saving it together with the video recording data and uploading it online in real time.
  • the external device that communicates with the LTE communication unit further includes an equipment fleet management system that includes equipment operation time, movement location tracking, and equipment down-time.
  • the equipment fleet management system is linked to the CNN-based object recognition unit to monitor the surroundings of the heavy equipment. Updates are reflected in preset risk radius and situations, including:
  • the external device communicating with the LTE communication unit further includes a field risk map display unit that collects event information detected from the CNN-based object recognition unit and displays the risk zone on a map.
  • the external device communicating with the LTE communication unit further includes a risk event monitoring display unit for each equipment and worker to track equipment work paths and worker movements based on event information detected from the CNN-based object recognition unit.
  • the LTE communication unit is equipped with GPS to locate and track equipment or workers.
  • the external server or administrator terminal receives and uploads event information, including event images or clip videos, in real time when a worker approaches within the danger radius detected by the CNN-based object recognition unit.
  • the embedded image recognition safety integrated control platform including the artificial intelligence-based image recognition system has an object detection accuracy of 88% or more (human, based on F1 Score), a maximum recognition distance of 7m or more, a range of 360 degrees, high temperature reliability of 60°C, and risk factors.
  • the recognition speed is less than 0.5s.
  • the embedded image recognition safety integrated control platform including the artificial intelligence-based image recognition system of the present invention includes an image data reception unit that receives image shooting data around a construction or industrial site from a camera module; A CNN-based object recognition unit that recognizes human objects in the video shooting data received from the video data receiver, boxes the recognized human objects in response to the data of the recognized human objects, and detects them; and a CNN-based object recognition unit.
  • An image recognition system including an event information generator that generates a warning light or a warning alarm when a human object detected from is exposed to a preset risk radius and situation including around heavy equipment; A false positive/no positive classification unit that receives event information detected from the CNN-based object recognition unit and classifies it as a true positive, false positive, or false positive, and a deep learning-based object recognition learning algorithm that classifies the false positive or false positive received from the false positive/unpositive classification unit.
  • a re-learning unit that re-trains in real time through YOLO (You Only Look Once), a learning model update unit that uploads the re-learning data set generated from the re-learning unit to the CNN-based object recognition unit in real time online, and the image recognition unit.
  • AI image recognition platform that includes an LTE communication unit for communication between the system and external devices, including external servers or administrator terminals, to ensure object recognition accuracy in various environments in industrial or construction sites, and to improve and sustain risk event reliability. It has a technological effect in preventing collisions and collisions through collection and re-learning of possible learning data.
  • Figure 1 is an example of preventing construction equipment collisions and constriction accidents.
  • Figure 2 is a schematic diagram of the entire embedded image recognition safety integrated control platform including the artificial intelligence-based image recognition system of the present invention.
  • FIG. 3 is a detailed diagram of the entire embedded image recognition safety integrated control platform including the artificial intelligence-based image recognition system of the present invention.
  • Figure 4 is an overall flow chart of the embedded image recognition safety integrated control platform including the artificial intelligence-based image recognition system of the present invention.
  • Figure 5 is a process diagram of the embedded image recognition safety integrated control platform re-learning unit including the artificial intelligence-based image recognition system of the present invention.
  • Figure 6 is a configuration diagram of the re-learning unit of the embedded image recognition safety integrated control platform including the artificial intelligence-based image recognition system of the present invention.
  • Figure 7 is a case diagram of classification of true positives, false positives, or missing positives according to the present invention.
  • Figure 8 is an example of a re-learning dataset using the video recording data auto-labeling tool.
  • Figure 9 is an example of an equipment fleet management system added to the present invention.
  • Figure 10 is a case of risk event monitoring by equipment and operator.
  • Figure 11 is a test and performance specification diagram of the image recognition safety integrated control platform of the present invention
  • the embedded image recognition safety integrated control platform including the artificial intelligence-based image recognition system of the present invention collects video shooting data from a camera module around a construction or industrial site.
  • a video data receiving unit 101 that receives video data;
  • a CNN-based object recognition unit that recognizes human objects in the video shooting data received from the video data receiver, boxes the recognized human objects in response to the data of the recognized human objects, and detects them; and a CNN-based object recognition unit.
  • an image recognition system 102 including an event information generator that generates a warning light or a warning alarm when a human object detected from is exposed to a preset risk radius and situation including around heavy equipment;
  • a false positive/no positive classification unit that receives event information detected from the CNN-based object recognition unit and classifies it as a true positive, false positive, or false positive, and a deep learning-based object recognition learning algorithm that classifies the false positive or false positive received from the false positive/unpositive classification unit.
  • a re-learning unit that re-trains in real time through YOLO (You Only Look Once), a learning model update unit that uploads the re-learning data set generated from the re-learning unit to the CNN-based object recognition unit in real time online, and the image recognition unit.
  • It consists of an AI image recognition platform (103) that includes an LTE communication unit for communication between the system and external devices, including external servers or administrator terminals.
  • the CNN-based object recognition unit sets a virtual border or virtual area in the video shooting data received from the video data receiver 101 to determine the preset risk radius and situation including the surroundings of the heavy equipment. It may include appearance, entry, and collapse within, and the setting of the virtual boundary or virtual area can be changed or set at any time in the CNN-based object recognition unit.
  • the image data receiver In the video shooting data received from 101
  • 2 human objects are recognized by the CNN-based object recognition unit
  • the recognized human objects are boxed and detected in response to the data of the recognized human objects
  • 3 the detected human objects are detected.
  • the CNN Detected event information is received from the base object recognition unit and classified into true positives, false positives, or false positives, and the classification of false positives and false positives is re-learned in real time through YOLO (You Only Look Once), a deep learning-based object recognition learning algorithm.
  • YOLO You Only Look Once
  • the re-learning data set labels the video recording data according to the false positive or false positive classification using an auto-labeling tool, is then stored together with the video recording data, and is uploaded online in real time.
  • a false positive/no positive classification unit that receives event information detected from the CNN-based object recognition unit and classifies it as a true positive, false positive, or false positive, and the false positive/unpositive classification unit.
  • a re-learning unit that re-trains the false positive or no-positive classification received from the unit in real time through YOLO (You Only Look Once), a deep learning-based object recognition learning algorithm, and the re-learning data set generated by the re-learning unit is classified as a false positive.
  • YOLO You Only Look Once
  • YOLO is one of the deep learning-based supervised learning algorithms for object detection. It is an abbreviation for You Only Look Once, which literally means seeing only once, and determines the classification and location of objects through a single regression of the image.
  • YOLO is based on the CNN structure, and the network architecture is based on the GoogLeNet model and consists of 24 Convolutional layers and 2 Fully Connected layers.
  • FIG. 9 shows the structure of YOLO.
  • the way to process images with YOLO is to resize the input image, run a convolutional network on the image, and threshold the resulting detections based on the confidence of the model.
  • the bounding box consists of five elements: x, y, w, h, and Confidence Score.
  • (x, y) are the coordinates of the center of the bounding box relative to the boundary of the grid cell.
  • (w, h) refers to the width and height of the bounding box.
  • Confidence Score represents the IOU between the predicted bounding box and all correct bounding boxes.
  • Each grid cell also predicts the C conditional class probability, Pr(Classi
  • the score encodes the probability that the class appears in the box and how well the predicted box fits the object.
  • YOLO's system models detection as a regression problem. Divide the image into an S ⁇ S grid and predict for each grid cell a B bounding box, confidence for that box, and C class probability. These predictions are encoded in the S ⁇ S ⁇ (B ⁇ 5 + C) tensor.
  • the present invention uses YOLOv4, the fourth version of YOLO, which has the advantage of being fast, capable of real-time detection, and greatly improved accuracy, thus providing satisfactory performance.
  • the external device communicating with the LTE communication unit further includes an equipment fleet management system including equipment operation time, movement location tracking, and equipment down-time, as shown in [ Figure 10].
  • an equipment fleet management system including equipment operation time, movement location tracking, and equipment down-time, as shown in [ Figure 10].
  • the external device communicating with the LTE communication unit includes a device for each equipment and worker to track the equipment work path and worker movement based on the event information detected from the CNN-based object recognition unit.
  • a risk event monitoring display unit may be further included.
  • the LTE communication unit is equipped with GPS to locate and track equipment or workers.
  • the external server or administrator terminal may be configured to receive and upload event information, including event images or clip videos, in real time when a worker approaches within the danger radius detected by the CNN-based object recognition unit.
  • the embedded image recognition safety integrated control platform including the artificial intelligence-based image recognition system according to the present invention has an object detection accuracy of 88% or more (human, based on F1 Score), maximum Recognition distance is 7m or more, range is 360 degrees, high temperature reliability is 60°C, and risk factor recognition speed is 0.5s or less.
  • the embedded image recognition safety integrated control platform including the artificial intelligence-based image recognition system of the present invention includes an image data reception unit that receives image shooting data around a construction or industrial site from a camera module; A CNN-based object recognition unit that recognizes human objects in the video shooting data received from the video data receiver, boxes the recognized human objects in response to the data of the recognized human objects, and detects them; and a CNN-based object recognition unit.
  • An image recognition system including an event information generator that generates a warning light or a warning alarm when a human object detected from is exposed to a preset risk radius and situation including around heavy equipment; A false positive/no positive classification unit that receives event information detected from the CNN-based object recognition unit and classifies it as a true positive, false positive, or false positive, and a deep learning-based object recognition learning algorithm that classifies the false positive or false positive received from the false positive/unpositive classification unit.
  • a re-learning unit that re-trains in real time through YOLO (You Only Look Once), a learning model update unit that uploads the re-learning data set generated from the re-learning unit to the CNN-based object recognition unit in real time online, and the image recognition unit.
  • AI image recognition platform that includes an LTE communication unit for communication between the system and external devices, including external servers or administrator terminals, to ensure object recognition accuracy in various environments in industrial or construction sites, and to improve and sustain risk event reliability. It has a technological effect in preventing collisions and collisions through collection and re-learning of available learning data, so it has potential for industrial use.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Tourism & Hospitality (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • General Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Primary Health Care (AREA)
  • Marketing (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Emergency Management (AREA)
  • Signal Processing (AREA)
  • Image Analysis (AREA)
  • Child & Adolescent Psychology (AREA)
  • Radar, Positioning & Navigation (AREA)

Abstract

La présente invention concerne une plateforme embarquée de reconnaissance d'images et de gestion de sécurité intégrée comprenant un système de reconnaissance d'images basée sur l'intelligence artificielle. Plus précisément, la plateforme comprend : une unité de réception de données d'image qui reçoit des données d'image capturées des alentours d'un chantier de construction ou d'un site industriel en provenance d'un module de caméra ; un système de reconnaissance d'images comprenant une unité de reconnaissance d'objets basée sur CNN qui reconnaît un objet humain dans les données d'image capturées reçues en provenance de l'unité de réception de données d'image et détecte l'objet humain reconnu par encadrement de celui-ci en correspondance avec des données de l'objet humain reconnu, et une unité de génération d'informations d'événement qui génère un feu d'avertissement ou une alarme d'avertissement lorsque l'objet humain détecté par l'unité de reconnaissance d'objets basée sur CNN est exposé à une situation dangereuse et dans un rayon de danger prédéfini comprenant les alentours d'un équipement lourd ; et une plateforme de reconnaissance d'images par intelligence artificielle (IA) comprenant une unité de classification en fausse détection/détection manquée qui reçoit les informations d'événement détectées par l'unité de reconnaissance d'objets basée sur CNN et classifie les informations d'événement en détection correcte, fausse détection ou détection manquée, une unité de ré-entraînement pour ré-entraîner la classification en fausse détection ou en détection manquée, reçue de l'unité de classification en fausse détection/détection manquée, en temps réel au moyen de You Only Look Once (YOLO), qui est un algorithme d'entraînement à la reconnaissance d'objets basé sur l'apprentissage profond, une unité de mise à jour de modèle d'entraînement qui téléverse un ensemble de données de ré-entraînement, généré par l'unité de ré-entraînement, vers l'unité de reconnaissance d'objets basée sur CNN en ligne en temps réel, et une unité de communication LTE pour une communication entre le système de reconnaissance d'images et un dispositif externe comprenant un serveur externe ou un terminal de gestionnaire.
PCT/KR2023/016700 2022-11-08 2023-10-26 Plateforme embarquée de reconnaissance d'images et de gestion de sécurité intégrée comprenant un système de reconnaissance d'images basée sur l'intelligence artificielle WO2024101726A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR1020220147778A KR102541212B1 (ko) 2022-11-08 2022-11-08 인공지능 기반 영상인식 시스템을 포함하는 임베디드 영상인식 안전통합관제 플랫폼
KR10-2022-0147778 2022-11-08

Publications (1)

Publication Number Publication Date
WO2024101726A1 true WO2024101726A1 (fr) 2024-05-16

Family

ID=86762411

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2023/016700 WO2024101726A1 (fr) 2022-11-08 2023-10-26 Plateforme embarquée de reconnaissance d'images et de gestion de sécurité intégrée comprenant un système de reconnaissance d'images basée sur l'intelligence artificielle

Country Status (2)

Country Link
KR (1) KR102541212B1 (fr)
WO (1) WO2024101726A1 (fr)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102541212B1 (ko) * 2022-11-08 2023-06-13 주식회사 영신 인공지능 기반 영상인식 시스템을 포함하는 임베디드 영상인식 안전통합관제 플랫폼

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101910542B1 (ko) * 2017-06-21 2018-10-22 에스케이 텔레콤주식회사 객체 검출을 위한 영상분석 서버장치 및 방법
KR102296274B1 (ko) * 2020-10-26 2021-09-01 주식회사 보고넷 사용자 학습 기반 딥러닝 객체 인식 서비스 제공 방법
KR102333143B1 (ko) * 2021-03-15 2021-12-02 다인스 주식회사 무인 계수 서비스 제공 시스템
KR102541212B1 (ko) * 2022-11-08 2023-06-13 주식회사 영신 인공지능 기반 영상인식 시스템을 포함하는 임베디드 영상인식 안전통합관제 플랫폼

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101808587B1 (ko) 2017-08-03 2017-12-13 주식회사 두원전자통신 객체인식과 추적감시 및 이상상황 감지기술을 이용한 지능형 통합감시관제시스템
KR102185859B1 (ko) 2020-03-31 2020-12-02 (주)케이넷 이엔지 객체 추적 장치 및 방법
KR102206662B1 (ko) 2020-08-14 2021-01-22 아이티플래닛 주식회사 항만 컨테이너 터미널에서 차량 출입 관리와 객체를 인식하는 비전 카메라 시스템 및 방법
KR102263512B1 (ko) 2020-09-14 2021-06-10 가온플랫폼 주식회사 스마트 객체인식이 가능한 IoT 통합 지능형 영상분석 플랫폼 시스템

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101910542B1 (ko) * 2017-06-21 2018-10-22 에스케이 텔레콤주식회사 객체 검출을 위한 영상분석 서버장치 및 방법
KR102296274B1 (ko) * 2020-10-26 2021-09-01 주식회사 보고넷 사용자 학습 기반 딥러닝 객체 인식 서비스 제공 방법
KR102333143B1 (ko) * 2021-03-15 2021-12-02 다인스 주식회사 무인 계수 서비스 제공 시스템
KR102541212B1 (ko) * 2022-11-08 2023-06-13 주식회사 영신 인공지능 기반 영상인식 시스템을 포함하는 임베디드 영상인식 안전통합관제 플랫폼

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
AHN SEO-HYUN, KIM SUNG-KEUN; LEE KWAN-HOON: "Development of a Fleet Management System for Cooperation Among Construction Equipment", JOURNAL OF THE KOREAN SOCIETY OF CIVIL ENGINEERS, vol. 36, no. 3, 1 June 2016 (2016-06-01), KR , pages 573 - 586, XP093170793, ISSN: 1015-6348, DOI: 10.12652/Ksce.2016.36.3.0573 *
SHIM SEUNGBO , SANG-IL CHOI: "Development on Identification Algorithm of Risk Situation around Construction Vehicle using YOLO-v3", JOURNAL OF THE KOREA ACADEMIA-INDUSTRIAL COOPERATION SOCIETY, vol. 20, no. 7, 31 July 2019 (2019-07-31), pages 622 - 629, XP093170790, ISSN: 1975-4701, DOI: 10.5762/KAIS.2019.20.7.622 *

Also Published As

Publication number Publication date
KR102541212B1 (ko) 2023-06-13

Similar Documents

Publication Publication Date Title
CN111898514B (zh) 一种基于目标检测与动作识别的多目标视觉监管方法
CN110188724B (zh) 基于深度学习的安全帽定位与颜色识别的方法及系统
Fang et al. Detecting non-hardhat-use by a deep learning method from far-field surveillance videos
KR102206662B1 (ko) 항만 컨테이너 터미널에서 차량 출입 관리와 객체를 인식하는 비전 카메라 시스템 및 방법
Park et al. Construction worker detection in video frames for initializing vision trackers
WO2024101726A1 (fr) Plateforme embarquée de reconnaissance d'images et de gestion de sécurité intégrée comprenant un système de reconnaissance d'images basée sur l'intelligence artificielle
KR102206658B1 (ko) 항만 컨테이너 터미널에서 미들웨어를 통해 터미널 운영 시스템과 연동되는 비전 카메라 시스템 및 그 방법
CN110602449A (zh) 一种基于视觉的大场景下施工安全智能监控系统方法
KR101877294B1 (ko) 객체, 영역 및 객체가 유발하는 이벤트의 유기적 관계를 기반으로 한 복수 개 기본행동패턴 정의를 통한 복합 상황 설정 및 자동 상황 인지가 가능한 지능형 방범 ｃｃｔｖ 시스템
CN106204640A (zh) 一种运动目标检测系统及方法
CN112489272A (zh) 一种专用于园区的智能安防控制系统
US7295106B1 (en) Systems and methods for classifying objects within a monitored zone using multiple surveillance devices
CN108376246A (zh) 一种多人脸识别及跟踪系统及方法
CN110619277A (zh) 一种多社区智慧布控方法以及系统
KR101989376B1 (ko) 궤도 회로 통합 감시 장치
KR102434154B1 (ko) 영상감시시스템에서의 고속 이동물체의 위치 및 모션 캡쳐 방법
CN112282819B (zh) 基于视觉的综采工作面人员目标安全监控方法及系统
US11308792B2 (en) Security systems integration
WO2022055023A1 (fr) Système de plateforme d'analyse d'image intelligent intégré ido capable de reconnaître des objets intelligents
CN109001210B (zh) 一种人防门密封胶条老化龟裂检测系统及方法
CN114022767A (zh) 一种电梯楼层定位方法、装置、终端设备及存储介质
CN117787844A (zh) 一种基于rfid的金融押运过程实时监控方法及系统
KR20230043662A (ko) 입국장 내 우범 여행자 추적 시스템 및 방법
KR101213255B1 (ko) 영상처리와 무선 센서네트워크를 이용한 상황인지 시스템 및 그 방법
JPH09130781A (ja) 広域監視装置

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2024525215

Country of ref document: JP

Kind code of ref document: A

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23888988

Country of ref document: EP

Kind code of ref document: A1