CN115082403A - Belt deviation detection algorithm based on semantic segmentation - Google Patents

Belt deviation detection algorithm based on semantic segmentation Download PDF

Info

Publication number
CN115082403A
CN115082403A CN202210712523.3A CN202210712523A CN115082403A CN 115082403 A CN115082403 A CN 115082403A CN 202210712523 A CN202210712523 A CN 202210712523A CN 115082403 A CN115082403 A CN 115082403A
Authority
CN
China
Prior art keywords
belt
semantic segmentation
detection algorithm
neural network
deviation detection
Prior art date
Legal status (The legal status 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 status listed.)
Pending
Application number
CN202210712523.3A
Other languages
Chinese (zh)
Inventor
陈叶亮
徐晨鑫
雷凌
朱恩东
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Beixin Intelligent Technology Co ltd
Original Assignee
Nanjing Beixin Intelligent Technology Co ltd
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 Nanjing Beixin Intelligent Technology Co ltd filed Critical Nanjing Beixin Intelligent Technology Co ltd
Priority to CN202210712523.3A priority Critical patent/CN115082403A/en
Publication of CN115082403A publication Critical patent/CN115082403A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30164—Workpiece; Machine component
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30232—Surveillance
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Quality & Reliability (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a belt deviation detection algorithm based on semantic segmentation, which comprises the following steps: reading or drawing an offset limiting area; reading the image frame of the network camera; image pre-processing (resizing, transposor data, etc.); a predictive neural network; predicting the position of the belt; extracting the position and shape characteristics of the belt; calculating the relative position of the belt edge and the offset limiting area; if the belt deviates, if so, counting the deviation, and alarming for more than 15 times; and (no) returning to the step of reading the image frame of the network camera for circulation. The belt deviation detection algorithm based on semantic segmentation only needs to adopt a pan-tilt camera to carry out real-time video acquisition, has low hardware overhead, high stability and low environmental requirement, can adjust a detection area in real time compared with the existing belt deviation detection method, and is convenient to operate; the edge position of the belt is displayed in real time, so that a client can feel visually; and when the deviation occurs, the system can also inform staff in time and take corresponding treatment measures for the belt.

Description

Belt deviation detection algorithm based on semantic segmentation
Technical Field
The invention relates to the technical field of computer vision detection, in particular to a belt deviation detection algorithm based on semantic segmentation.
Background
The existing belt deviation detection mainly comprises the modes of fixing a traditional mechanical lifting type pressure turbine, detecting by a fluorescence sensor and an infrared sensor, detecting by a traditional visual algorithm and the like, wherein one mechanical lifting type pressure turbine is fixed, the detection mode of fixing the traditional mechanical lifting type pressure turbine increases the friction among transmission parts, and meanwhile, the belt abrasion is increased when the left and right deviation is limited, and the belt transmission efficiency is reduced; the fluorescence detection reduces the possibility that the conveying belt is abraded by the outside, but the detection effect is influenced by uneven materials, object deviation and even sand and dust shielding during some conveying, so that omission or misinformation occurs; the three infrared sensors are used for detecting, and the detection devices are arranged at the edges of the two sides of the belt, so that the belt deviation is detected by using the infrared sensors, but the method is also influenced by the installation position, external force and the like, and is inconvenient to adjust in real time according to the field condition; four traditional visual algorithms: the requirement on the resolution of the camera is very high, the requirement on the illumination condition is strict, different production environments need to be specifically adjusted, and the deployment is complicated.
In summary, most of the conventional belt deviation detection methods use additional contact devices or sensors, so that non-contact unattended detection cannot be achieved, and the adjustment of the devices is complicated. Few schemes through traditional visual algorithms have harsh requirements on equipment and production environments and are not favorable for deployment.
We propose a belt deviation detection algorithm based on semantic segmentation in order to solve the problems set forth above.
Disclosure of Invention
The invention aims to provide a belt deviation detection algorithm based on semantic segmentation so as to solve the problems in the market in the prior art.
In order to achieve the purpose, the invention provides the following technical scheme: a belt deviation detection algorithm based on semantic segmentation comprises the following steps: reading or drawing an offset limiting area; reading the image frame of the network camera; image preprocessing (resizing, transtensor data, etc.); a predictive neural network; predicting the position of the belt; extracting the position and shape characteristics of the belt; calculating the relative position of the belt edge and the offset limiting area; whether the belt is deviated or not; (yes) offset count, over 15 alarms; and (no) returning to the step of reading the image frame of the network camera for circulation.
Preferably, the platform for reading or drawing the offset limiting area is an X86 platform PC, and the monitoring camera is used for reading the image frame of the web camera.
Preferably, the belt position prediction is performed by the following process: preparing training data; building a training neural network; preprocessing data; training a network model; data iteration and loss calculation; and obtaining a network model.
Preferably, the neural network adopts the mobilene as a semantic segmentation model of the backbone, the mobilene and the Unet, the neural network extracts the belt characteristics after predicting the position of the belt, and finally, whether the belt deviates or not is judged by calculating the position of the edge of the belt.
Preferably, the Unet model uses a lightweight faster MobileNet network.
Preferably, the Unet semantic segmentation adopts a symmetrical structure; firstly, performing feature extraction through convolution pooling, and then performing reconstruction through up-sampling; feature extraction is mainly carried out by adopting a convolution kernel of 3x3 and a relu activation function, and downsampling is carried out by using 2x2 maximum pooling for compression; when in upsampling, adjacent interpolation is used, and transposition convolution is not used; jump connection is carried out after each upsampling, concat operation is used, the features are spliced together in the channel dimension to form thicker features, and the global features and the local features are combined instead of simple addition; subsequently performing channel dimension compression using the wraps; when jumping and connecting, the big picture is cut down to supplement the whole situation and assist the position correction.
Preferably, the training neural network is built by adopting a Pythrch and Opencv, the Pythrch is an open-source neural network framework and is specially designed for GPU accelerated Deep Neural Network (DNN) programming, and the Opencv is used for an open-source function library in the aspects of image processing, analysis and machine vision.
Preferably, the image preprocessing object is in the form of a picture, the picture preprocessing mode is divided into picture resolution compression and binarization, the input size of the Unet neural network is 512X512X1, the image is firstly compressed into 512X512 resolution, the picture training model is a single-channel model, and default three channels need to be binarized into a single-channel picture; and (4) converting the picture into the Tensor, converting the processed picture into a Tensor format and sending the Tensor format into a neural network.
Preferably, the relative position of the belt edge and the deviation limiting area is calculated to obtain belt MASK data through prediction, the format of the MASK data is converted and a binary image is extracted, and the binary image and the limiting area binary image are subjected to AND operation to obtain the belt deviation amount through calculation.
Compared with the prior art, the invention has the beneficial effects that: the belt deviation detection algorithm based on semantic segmentation carries out image preprocessing aiming at image video streams collected by a monitoring camera, and identifies and extracts the position of a belt in real time through a visual algorithm based on semantic segmentation to obtain information such as the shape, the position and the edge of the belt in a monitoring range. The allowable belt offset is adjusted by manually limiting left and right thresholds of the edge, the belt offset is out of the threshold range to trigger alarm and upload to a server, safety personnel are informed to solve hidden dangers, only a pan-tilt camera is needed to carry out real-time video acquisition, hardware cost is low, stability is high, environmental requirements are low, compared with the existing belt offset detection method, a detection area can be adjusted in real time, and operation is convenient; the edge position of the belt is displayed in real time, so that a client can feel visually; and when the deviation occurs, the system can also inform staff in time and take corresponding treatment measures for the belt.
Drawings
FIG. 1 is a schematic flow chart of a belt deviation detection algorithm based on semantic segmentation according to the present invention;
FIG. 2 is a schematic diagram of a belt deviation detection algorithm mobilene network structure based on semantic segmentation according to the present invention;
FIG. 3 is a schematic diagram of a belt deviation detection algorithm Unet network structure based on semantic segmentation.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example one
A belt deviation detection algorithm based on semantic segmentation comprises the following steps:
training a neural network model: preparing training data, namely preparing a plurality of groups of training parameters in advance;
the building of the training neural network is carried out by adopting Pythrch and Opencv, wherein the Pythrch is an open-source neural network framework and is specially designed for GPU accelerated Deep Neural Network (DNN) programming. Torch is a classical tensor (tensor) library that operates on multidimensional matrix data, and has wide application in machine learning and other mathematically intensive applications. Opencv is an open source function library used for image processing, analysis and machine vision, all codes of the library are optimized, the calculation efficiency is very high, the design is concentrated on the open source library used for a real-time system, a neural network adopts a mobilene as a backbone, a semantic segmentation model of the mobilene + Unet, the neural network extracts belt features after predicting the position of a belt, and finally, whether the belt deviates or not is judged by calculating the position of the edge of the belt, the Unet model adopts a lightweight faster MobileNet network, and Unet semantic segmentation adopts a symmetrical structure; firstly, performing feature extraction through convolution pooling, and then performing reconstruction through up-sampling; feature extraction is mainly carried out by adopting a convolution kernel of 3x3 and a relu activation function, and downsampling is carried out by using 2x2 maximum pooling for compression; when in upsampling, adjacent interpolation is used, and transposition convolution is not used; jump connection is carried out after each upsampling, concat operation is used, the features are spliced together in the channel dimension to form thicker features, and the global features and the local features are combined instead of simple addition; subsequently performing channel dimensional compression using the wraps; when jumping connection is carried out, the large graph is cut down, global supplement is carried out, and position correction is assisted;
preprocessing data; training a network model; data iteration and loss calculation; and obtaining a network model.
Predicting by a neural network model: reading or drawing an offset limiting area; reading the image frame of the network camera;
image preprocessing (size adjustment, transtensor data and the like), wherein the image preprocessing mode comprises image resolution compression and binarization, the input size of an Unet neural network is 512X512X1, the image is firstly compressed into 512X512 resolution, a picture training model is a single-channel model, and default three channels need to be binarized into a single-channel picture; the picture is transferred to the Tensor, the processed picture is transferred to the Tensor format and sent to the neural network; a predictive neural network; predicting the belt position, and processing through a network model; extracting the position and shape characteristics of the belt; calculating the relative position of the belt edge and the deviation limiting area, obtaining belt MASK data through prediction, converting the format of the MASK data, extracting a binary image, performing AND operation on the binary image and the limiting area binary image, and obtaining the belt deviation through calculation. The alarm is triggered or the deviation is counted for a certain time, and the alarm is accumulated for more than 15 times.
Example two
A belt deviation detection algorithm based on semantic segmentation comprises the following steps:
training a neural network model: preparing training data, namely preparing a plurality of groups of training parameters in advance;
the building of the training neural network is carried out by adopting Pythrch and Opencv, wherein the Pythrch is an open-source neural network framework and is specially designed for GPU accelerated Deep Neural Network (DNN) programming. Torch is a classical tensor (tensor) library that operates on multidimensional matrix data, and has wide application in machine learning and other mathematically intensive applications. Opencv is an open source function library used for image processing, analysis and machine vision, all codes of the library are optimized, the calculation efficiency is very high, the design is concentrated on the open source library used for a real-time system, a neural network adopts a mobilene as a backbone, a semantic segmentation model of the mobilene + Unet, the neural network extracts belt features after predicting the position of a belt, and finally, whether the belt deviates or not is judged by calculating the position of the edge of the belt, the Unet model adopts a lightweight faster MobileNet network, and Unet semantic segmentation adopts a symmetrical structure; firstly, performing feature extraction through convolution pooling, and then performing reconstruction through up-sampling; feature extraction is mainly carried out by adopting a convolution kernel of 3x3 and a relu activation function, and downsampling is carried out by using maximum pooling of 2x2 for compression; when in upsampling, adjacent interpolation is used, and transposition convolution is not used; jump connection is carried out after each upsampling, concat operation is used, the features are spliced together in the channel dimension to form thicker features, and the global features and the local features are combined instead of simple addition; subsequently performing channel dimension compression using the wraps; when jumping connection is carried out, the large graph is cut down, global supplement is carried out, and position correction is assisted;
preprocessing data; training a network model; data iteration and loss calculation; and obtaining a network model.
Predicting by a neural network model: reading or drawing an offset limiting area; reading the image frame of the network camera;
image preprocessing (size adjustment, transtensor data and the like), wherein the image preprocessing mode comprises image resolution compression and binarization, the input size of an Unet neural network is 512X512X1, the image is firstly compressed into 512X512 resolution, a picture training model is a single-channel model, and default three channels need to be binarized into a single-channel picture; the picture is transferred to the Tensor, the processed picture is transferred to the Tensor format and sent to the neural network; a predictive neural network; predicting the belt position, and processing through a network model; extracting the position and shape characteristics of the belt; calculating the relative position of the belt edge and the deviation limiting area, obtaining belt MASK data through prediction, converting the format of the MASK data and extracting a binary image, carrying out ' AND ' operation ' on the binary image and the limiting area binary image, obtaining the belt deviation through calculation to judge whether the belt deviates, and if not, repeating the process from reading the network camera image frame.
And those not described in detail in this specification are well within the skill of those in the art.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications may be made to the embodiments described in the foregoing embodiments, or equivalents may be substituted for elements thereof.

Claims (9)

1. A belt deviation detection algorithm based on semantic segmentation comprises the following steps: reading or drawing an offset limiting area; reading the image frame of the network camera; image preprocessing (resizing, transtensor data, etc.); a predictive neural network; predicting the position of the belt; extracting the position and shape characteristics of the belt; calculating the relative position of the belt edge and the offset limiting area; whether the belt is deviated or not; (yes) offset count, over 15 alarms; and (no) returning to the step of reading the image frame of the network camera for circulation.
2. The belt deviation detection algorithm based on semantic segmentation as claimed in claim 1, wherein: the platform for reading or drawing the offset limiting area is an X86 platform PC, and the monitoring camera is used for reading the image frame of the network camera.
3. The belt deviation detection algorithm based on semantic segmentation as claimed in claim 1, wherein: the belt position prediction is carried out by adopting the following process: preparing training data; building a training neural network; preprocessing data; training a network model; data iteration and loss calculation; and obtaining a network model.
4. A semantic segmentation based belt deviation detection algorithm according to claim 3, characterized in that: the neural network adopts the mobilene as a semantic segmentation model of backbone, the mobilene and Unet, the neural network extracts belt characteristics after predicting the position of the belt, and finally judges whether the belt deviates or not by calculating the position of the edge of the belt.
5. The belt deviation detection algorithm based on semantic segmentation according to claim 4, characterized in that: the Unet model uses a lightweight faster MobileNet network.
6. The belt deviation detection algorithm based on semantic segmentation according to claim 4, characterized in that: the Unet semantic segmentation adopts a symmetrical structure; firstly, performing feature extraction through convolution pooling, and then performing reconstruction through up-sampling; feature extraction is mainly carried out by adopting a convolution kernel of 3x3 and a relu activation function, and downsampling is carried out by using 2x2 maximum pooling for compression; when in upsampling, adjacent interpolation is used, and transposition convolution is not used; jump connection is carried out after each upsampling, concat operation is used, the features are spliced together in the channel dimension to form thicker features, and the global features and the local features are combined instead of simple addition; subsequently performing channel dimension compression using the wraps; when jumping and connecting, the big picture is cut down to supplement the whole situation and assist the position correction.
7. A semantic segmentation based belt deviation detection algorithm according to claim 3, characterized in that: the training neural network is built by adopting a Pythrch and Opencv, the Pythrch is an open-source neural network framework and specially aims at GPU accelerated Deep Neural Network (DNN) programming, and the Opencv is used for an open-source function library in the aspects of image processing, analysis and machine vision.
8. The belt deviation detection algorithm based on semantic segmentation as claimed in claim 1, wherein: the method comprises the following steps that an object of image preprocessing is in a picture form, the picture preprocessing mode is divided into picture resolution compression and binarization, the input size of a Unet neural network is 512X512X1, firstly, the image is compressed into 512X512 resolution, a picture training model is a single-channel model, and default three channels need to be binarized into a single-channel picture; and (4) converting the picture into the Tensor, converting the processed picture into a Tensor format and sending the Tensor format into a neural network.
9. The belt deviation detection algorithm based on semantic segmentation as claimed in claim 1, wherein: and calculating the relative position of the belt edge and the deviation limiting area to obtain belt MASK data through prediction, converting the MASK data format and extracting a binary image, and performing AND operation on the binary image and the limiting area binary image to obtain the belt deviation.
CN202210712523.3A 2022-06-22 2022-06-22 Belt deviation detection algorithm based on semantic segmentation Pending CN115082403A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210712523.3A CN115082403A (en) 2022-06-22 2022-06-22 Belt deviation detection algorithm based on semantic segmentation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210712523.3A CN115082403A (en) 2022-06-22 2022-06-22 Belt deviation detection algorithm based on semantic segmentation

Publications (1)

Publication Number Publication Date
CN115082403A true CN115082403A (en) 2022-09-20

Family

ID=83254498

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210712523.3A Pending CN115082403A (en) 2022-06-22 2022-06-22 Belt deviation detection algorithm based on semantic segmentation

Country Status (1)

Country Link
CN (1) CN115082403A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115272979A (en) * 2022-09-22 2022-11-01 南京北新智能科技有限公司 Semantic segmentation-based coal flow proportion detection method
CN115359055A (en) * 2022-10-19 2022-11-18 煤炭科学技术研究院有限公司 Conveyor belt edge detection method, conveyor belt edge detection device, electronic equipment and storage medium
CN117911501A (en) * 2024-03-20 2024-04-19 陕西中铁华博实业发展有限公司 High-precision positioning method for metal processing drilling

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115272979A (en) * 2022-09-22 2022-11-01 南京北新智能科技有限公司 Semantic segmentation-based coal flow proportion detection method
CN115359055A (en) * 2022-10-19 2022-11-18 煤炭科学技术研究院有限公司 Conveyor belt edge detection method, conveyor belt edge detection device, electronic equipment and storage medium
CN117911501A (en) * 2024-03-20 2024-04-19 陕西中铁华博实业发展有限公司 High-precision positioning method for metal processing drilling
CN117911501B (en) * 2024-03-20 2024-06-04 陕西中铁华博实业发展有限公司 High-precision positioning method for metal processing drilling

Similar Documents

Publication Publication Date Title
CN115082403A (en) Belt deviation detection algorithm based on semantic segmentation
CN109615019B (en) Abnormal behavior detection method based on space-time automatic encoder
CN112287816B (en) Dangerous work area accident automatic detection and alarm method based on deep learning
CN107679471B (en) Indoor personnel air post detection method based on video monitoring platform
KR102194499B1 (en) Apparatus for CCTV Video Analytics Based on Object-Image Recognition DCNN and Driving Method Thereof
EP2172903B1 (en) Video motion detection
KR102528547B1 (en) Apparatus and method for early detecting earthquakes
CN113052029A (en) Abnormal behavior supervision method and device based on action recognition and storage medium
CN102663375B (en) Active target identification method based on digital watermark technology in H.264
Dhole et al. Anomaly detection using convolutional spatiotemporal autoencoder
CN110598606A (en) Indoor falling behavior detection method with visual privacy protection advantage
Vijayan et al. A fully residual convolutional neural network for background subtraction
CN117576632B (en) Multi-mode AI large model-based power grid monitoring fire early warning system and method
CN116092198B (en) Mining safety helmet identification detection method, device, equipment and medium
CN114155472A (en) Method, device and equipment for detecting abnormal state of factory scene empty face protection equipment
CN113507605A (en) Dangerous area monitoring video compression storage method and system based on artificial intelligence
CN112633179A (en) Farmer market aisle object occupying channel detection method based on video analysis
CN117253191A (en) Safety helmet wearing detection method based on DETR model
CN111127355A (en) Method for finely complementing defective light flow graph and application thereof
US20220392225A1 (en) Concept for Detecting an Anomaly in Input Data
CN115027904A (en) Belt tearing detection method based on example segmentation
CN114463681A (en) Fire detection method based on video monitoring platform
CN113893517A (en) Rope skipping true and false judgment method and system based on difference frame method
JP6124739B2 (en) Image sensor
CN112307936A (en) Passenger flow volume analysis method, system and device based on head and shoulder detection

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination