CN107886081A - Two-way U Net deep neural network mine down-holes hazardous act is intelligently classified discrimination method - Google Patents
Two-way U Net deep neural network mine down-holes hazardous act is intelligently classified discrimination method Download PDFInfo
- Publication number
- CN107886081A CN107886081A CN201711185883.8A CN201711185883A CN107886081A CN 107886081 A CN107886081 A CN 107886081A CN 201711185883 A CN201711185883 A CN 201711185883A CN 107886081 A CN107886081 A CN 107886081A
- Authority
- CN
- China
- Prior art keywords
- way
- hazardous act
- image
- network
- net
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
Abstract
The invention discloses a kind of two-way U Net deep neural network mine down-holes hazardous act to be intelligently classified discrimination method, including the step such as data acquisition, image preprocessing, hazardous act sample make, one-way network designs and parameter training, structure two-way U Net network structures, image input module design, two-way U Net network characterizations extraction module design, warning module design.The present invention is using the image comprising hazardous act as input data, hazardous act therein is identified by U Net networks, classified and the method for early warning, the timely hazardous act for finding to occur during downhole production simultaneously sends early warning, so as to farthest avoid the generation of production accident.
Description
Technical field
The invention belongs to technical field of video image processing, and in particular to one kind is classified by U-Net deep neural networks
Realize that the hazardous act being likely to occur in process of production to miner carries out INTELLIGENT IDENTIFICATION, classification and the method for early warning.
Background technology
The main eye-observation of traditional video monitoring, there are some researches show professional monitoring personnel is only monitoring 2 monitoring
In the case of device, 95% behavior will be missed after 22 minutes.Therefore its can not the substantial amounts of camera of effective monitoring, video monitoring system
System thus lose that its due prevention is dangerous and the ability of active intervention, become one kind and can only provide evidence obtaining record is provided afterwards
The instrument of picture.Generally speaking, there is following problem in traditional video surveillance:
(1) weakness of mankind itself-be difficult to is focused on for a long time;
(2) it can only afterwards be checked, early warning and management in advance, in thing can not be realized;
(3) generation of hazardous act can not be effectively prevented in advance;
(4) wrong report occurs in system, but ignores real alert;
(5) retrieval difficulty-needs take a significant amount of time the picture that can just find needs.
In this case, intelligent Video Surveillance Technology is paid close attention to by national governments and scholar, is regarded with traditional
Frequency monitor mode is compared, intelligent monitoring technology can in time, useful information is automatically extracted from original video information, be used for
The transmission, preservation and retrieval of video are realized, so as to easily complete the task that manpower is difficult to complete, greatly improves monitoring effect
Rate.Importantly, intelligent Video Surveillance Technology by analyzing captured video, be capable of detecting when hazardous activity,
Event or behavior pattern, and timely early warning, so as to effectively prevent the generation of production accident.
But in Intelligent Recognition field, lack at present by nerual network technique to the identification of underground unsafe acts and
The research of early warning.
The content of the invention
In order to solve the above-mentioned technical problem, the invention provides one kind to pass through two-way U-Net deep neural networks dynamic point
Level early warning technology is recognized to the unsafe acts being likely to occur during downhole production, classified and the method for early warning, in time
It was found that and early warning these unsafe acts, so as to farthest avoid the generation of production accident.
The technical solution adopted in the present invention is:A kind of two-way U-Net deep neural networks mine down-hole hazardous act intelligence
Discrimination method can be classified, it is characterised in that comprise the following steps:
Step 1:Data acquisition;
From underground according to intervals interception image;
Step 2:Image preprocessing, including color space conversion, removal noise and image enhaucament;
Step 3:Make dangerous play sample;
Vector quantization is carried out to the border of monitor video sectional drawing, image contour is irised out and assigns class name numbering, and in image
Pixel is checked, to determine if to represent a kind of hazardous act, by all pixel groups that can represent a kind of hazardous act
Into image supply U-Net Network Recognitions as sample.
Step 4:One-way network structure design and parameter training;
One-way network structure include input layer, 10 convolutional layers, 6 Batchnorm layers, 10 ELU layers, 2 pond layers,
2 up-sampling layers, 2 Dropout layers, 1 Softmax layer and output layer, and relevant parameter is set.The danger presorted is moved
Make sample and add network as input data, network parameter is trained, preserve the network structure after the completion of result and training
Each layer weight;
Step 5:Build two-way U-Net structures;
Two-way U-Net structures are that two-way network forms a structure for alternately receiving input, are inputted by image, two-way U-
Net feature extractions and the step of early warning three are formed;
Step 6:Image inputs;
One queue structure for alternately receiving input is formed using two-way network, image input uses first in first out strategy;
Specific implementation comprises the following steps:
Step 6.1:Before inputting for the first time, it is 0 to set two-way to indicate position;
Step 6.2:Whether check image sample set is empty, if it is, terminating;Otherwise, image inputs, and checks two-channel structure
Input flag, selective value is 0 all the way renewal input;
Step 6.3:Update dual input and represent position, corresponding branch road input sign position will be recently entered and be set to 1, will be another
The input sign position of individual branch road is arranged to 0;
Step 6.4:Flow goes to step 6.2.
Step 7:Two-way U-Net feature extractions;
The one-way network trained in step 4 is replicated, forms two-way network structure, its each layer weight is by step 4
One-way network structure copy, output layer is trained;
Step 8:Early warning;
Underground hazardous act is identified respectively for two-way network, after two branch roads identify hazardous act simultaneously, leads to
Quaternary tree search method is crossed, classification, quantity and the locus that hazardous act occurs quickly are identified, and be classified pre-
Alert and preservation sectional drawing, early warning release after two-way does not identify hazardous act simultaneously.
The quaternary tree search method, specific implementation include following sub-step:
Step 8.1:Image-region is divided into upper left, upper right, lower-left and the quadrant of bottom right four;
Step 8.2:Since upper left, the pixel of hazardous act in quadrant, numbering 1-5, corresponding 5 kinds of dangerous rows are retrieved
For background and not dangerous behavior numbering are 0;
Step 8.3:If the dangerous behavior pixel of some pixel or classification are more than or equal to 2, step 8.1 is transferred to;
Step 8.4:If a kind of hazardous act in quadrant be present, class variable deposit hazardous act class number;
Step 8.5:Count four quadrant hazardous act classifications, intensive variables;
Step 8.6:When dangerous behavior quadrant for 0 and less than 2 when, two level early warning;When dangerous behavior
When quadrant is more than 2, one-level early warning;When different classes of hazardous act be present, one-level early warning.
From the perspective of network structure, obtain in the patent on nerual network technique of mandate and used single channel knot more
Structure.By contrast, the present invention can be obviously improved recognition efficiency using two-channel structure.
From the perspective of network parameter, compared to existing patent, the data volume used in the present invention is relatively fewer, because
And amount of calculation is also smaller, recognition efficiency can be equally lifted.
From the point of view of early warning degree of accuracy angle, early warning can be just only sent when two branch roads monitor hazardous act simultaneously,
Therefore the degree of accuracy of early warning is improved.
Brief description of the drawings
Fig. 1 is the general flow chart of present example;
Fig. 2 is the entire flow figure of present example;
Fig. 3 is the network structure of present example;
Fig. 4 is sample schematic diagram of the five class hazardous acts as experiment in present example;
Fig. 5 be present example in imagery zone be divided into upper left, upper right, lower-left and bottom right four it is equal quadrant signal
Figure;
Fig. 6 is that the present invention is will to be respectively divided into four quadrants there are still all quadrants of hazardous act in embodiment to illustrate
Figure.
Embodiment
Understand for the ease of those of ordinary skill in the art and implement the present invention, below in conjunction with the accompanying drawings and embodiment is to this hair
It is bright to be described in further detail, it will be appreciated that implementation example described herein is merely to illustrate and explain the present invention, not
For limiting the present invention.
See Fig. 1, Fig. 2 and Fig. 3, a kind of two-way U-Net deep neural networks mine down-hole danger row provided by the invention
Discrimination method is classified for intelligence, is comprised the following steps:
1. data acquisition:The image for including hazardous act is obtained from video, hazardous act is divided into 5 classes here:(1)
Laugh slept when quarrelling and fighting noisily (2) work (3) underground smoke (4) climb onto a slow-going train (5) embrace lamp heating, background and not dangerous behavior are numbered in addition
For 0;
2. image preprocessing:Including color space conversion, noise, image enhaucament etc. are removed, the purpose is to cut down image
Extraneous features, while strengthen in image and useful feature is identified to hazardous act.The purpose of image preprocessing is for next step
Network training is prepared.
3. sample hazardous act makes:Vector quantization is carried out to the border of imaged object (monitor video sectional drawing), used
Photoshop irises out image contour and assigns class name numbering, and the pixel in image is checked, to determine if to represent one
Class hazardous act, the image that all pixels that can represent a kind of hazardous act are formed supply U-Net Network Recognitions as sample;
Choose experiment sample collection:Sample of the five class hazardous acts as experiment in the 1st step is chosen, as shown in figure 4, wherein
(1) laughed quarrel and fight noisily (including 1-1 and 1-2), is slept (including 2-1 and 2-2) when (2) work, and (3) underground is smoked, and (4) climb onto a slow-going train, (5)
Embrace lamp heating (including 5-1 and 5-2);
Sample total is 3000, wherein 2400 are made training set, 600 are made checking collection.
4. one-way network structure design and training:Including input layer, convolutional layer, ELU layers, Batchnorm layers, maximum pond
Layer, up-sampling layer, Dropout layers, Softmax layers and output layer, and relevant parameter is set.
Network is added using the dangerous play sample presorted as input data, network parameter is trained, preserves knot
The weight of each layer of network structure after the completion of fruit and training.Rate of accuracy reached is to 92%.
5. build two-way network structure:Two-way network structure is that two-way network forms a structure for alternately receiving input,
Inputted by image, two-way U-Net feature extractions and early warning three parts are formed.
6. image inputs:The present invention forms a queue structure for alternately receiving input, image input using two-way network
Using first in first out (FIFO) strategy:Comprise the following steps that:
6.1:Before inputting for the first time, it is 0 to set two-way to indicate position
6.2:Whether check image sample set is empty, if it is, terminating;Otherwise, JPG images input, and check two-channel structure
Input flag, selective value is 0 all the way renewal input,
6.3:Dual input sign position is updated, corresponding branch road input sign position will be recently entered and be set to 1, another is propped up
The input sign position on road is arranged to 0;Two-way network merely enters a width JPG images every time, after two-way has image input, subsequently
The input of sequential JPG images is always to replace the branch road of input timing at most, so ensures that the image of two branch roads meets sequential
Continuity, it is easy to carry out constantly identification and early warning, while each two-way to personnel in the pit's dangerous play by continuous image
Structure only needs to re-start identification all the way, greatly reduces operand, improves the real-time of network.
6.4:Flow goes to step 6.2.
7. two-way U-Net feature extractions;
The one-way network trained in step 4 is replicated, forms two-way network structure, its each layer weight is by step 4
One-way network structure copy, output layer is trained;
8. early warning:Underground hazardous act is identified respectively for two-way network, when two branch roads identify dangerous row simultaneously
To be rear, system carries out early warning, and preserves sectional drawing, and early warning releases after two-way does not identify hazardous act simultaneously.Above-mentioned knot
Structure, the accuracy of underground hazardous act identification is ensured, reduced the rate of false alarm of early warning.
Comprise the following steps that:
8.1:Imagery zone is divided into upper left, upper right, the equal quadrant in lower-left and bottom right four;See Fig. 5;
8.2:Since upper left, retrieve in each quadrant all pixels (it is 2-1 or 2-2 then to be numbered if sleep behavior,
0) background and not dangerous behavior numbering are;
Be safe from danger behavior in left upper quadrant, and all pixels numbering is 0, dangerous behavior in its excess-three quadrant
(sleep), hazardous act numbering is 2-1 wherein in left lower quadrant, and hazardous act numbering is 2-2 in the quadrant of two, right side.
8.3:The classification of sleep behavior pixel in all quadrants is judged, if sleep behavior pixel class or deposited in some quadrant
It is more than or equal to 2 in the quadrant number of sleep behavior, is transferred to 8.1;
Be safe from danger behavior in left upper quadrant, dangerous behavior (sleep) in its excess-three quadrant.
See Fig. 6, four quadrants are respectively divided into by there are still all quadrants of hazardous act, carry out step 8.2 again,
By that analogy, until only a kind of hazardous act or no longer dangerous behavior in all quadrants.
8.4:If sleep behavior pixel class is 1 in quadrant, that is, a kind of sleep behavior is only existed, class variable is designated as
1 and it is stored in hazardous act class number;
8.5:Count four quadrant hazardous act classifications, intensive variables;
8.6:When dangerous behavior in the image of two branch roads, but the hazardous act in two width images exists only in
Two level early warning when in 1 quadrant;When the hazardous act in two width images is present in 2 and during with upper quadrant, one-level early warning;When
When different classes of hazardous act be present in two width images, one-level early warning;When only above-mentioned behavior or two occurs in a branch road
When above-mentioned behavior is not present in individual branch road, early warning is released.
It should be appreciated that the part that this specification does not elaborate belongs to prior art.
It should be appreciated that the above-mentioned description for preferred embodiment is more detailed, therefore can not be considered to this
The limitation of invention patent protection scope, one of ordinary skill in the art are not departing from power of the present invention under the enlightenment of the present invention
Profit is required under protected ambit, can also be made replacement or deformation, be each fallen within protection scope of the present invention, this hair
It is bright scope is claimed to be determined by the appended claims.
Claims (6)
1. a kind of two-way U-Net deep neural network mine down-holes hazardous act is intelligently classified discrimination method, it is characterised in that bag
Include following steps:
Step 1:Data acquisition;
From underground according to intervals interception image;
Step 2:Image preprocessing, including color space conversion, removal noise and image enhaucament;
Step 3:Make hazardous act sample;
Vector quantization is carried out to the border of monitor video sectional drawing, image contour is irised out and assigns class name numbering, and to the pixel in image
Checked, to determine if to represent a kind of hazardous act, all pixels that can represent a kind of hazardous act are formed
Image supplies U-Net Network Recognitions as sample;
Step 4:One-way network structure design and parameter training;
One-way network structure include input layer, 10 convolutional layers, 6 Batchnorm layers, 10 ELU layers, 2 pond layers, 2
Layer, 2 Dropout layers, 1 Softmax layer and output layer are up-sampled, and relevant parameter is set;The dangerous play that will be presorted
Sample adds network as input data, and network parameter is trained, the network structure after the completion of preserving result and training
The weight of each layer;
Step 5:Build two-way U-Net structures;
Two-way U-Net structures are that two-way network forms a structure for alternately receiving input, and specific building process includes step 6-
Step 8;
Step 6:Image inputs;
One queue structure for alternately receiving input is formed using two-way network, image input uses first in first out strategy;
Step 7:Two-way U-Net feature extractions;
The one-way network trained in step 4 is replicated, formed two-way network structure, its each layer weight by step 4 list
Road network structure is copied, and output layer is trained;
Step 8:Early warning;
Underground hazardous act is identified respectively for two-way network, after two branch roads identify hazardous act simultaneously, passes through four
Fork tree search method, to hazardous act occur classification, quantity and locus quickly identified, and carry out grading forewarning system with
Sectional drawing is preserved, early warning releases after two-way does not identify hazardous act simultaneously.
2. two-way U-Net deep neural network mine down-holes hazardous act according to claim 1 is intelligently classified identification side
Method, it is characterised in that:Image preprocessing described in step 2 includes color space conversion, removes noise and image enhaucament.
3. two-way U-Net deep neural network mine down-holes hazardous act according to claim 1 is intelligently classified identification side
Method, it is characterised in that:In step 3, vector quantization is carried out to the border of monitor video sectional drawing, image contour is irised out and assigns class name volume
Number, and the pixel in image is checked, to determine if to represent a kind of hazardous act, a kind of danger can be represented by all
The image of the pixel composition of dangerous behavior supplies U-Net Network Recognitions as sample.
4. the two-way U-Net deep neural network mine down-holes hazardous act intelligence according to claim 1-3 any one
It is classified discrimination method, it is characterised in that:In step 6, a queue structure for alternately receiving input, figure are formed using two-way network
As input uses first in first out strategy;Specific implementation comprises the following steps:
Step 6.1:Before inputting for the first time, it is 0 to set two-way to indicate position;
Step 6.2:Whether check image sample set is empty, if it is, terminating;Otherwise, image inputs, and checks the defeated of two-channel structure
Inlet identity position, for 0 all the way, renewal inputs selective value;
Step 6.3:Update dual input and represent position, corresponding branch road input sign position will be recently entered and be set to 1, another is propped up
The input sign position on road is arranged to 0;
Step 6.4:Flow goes to step 6.2.
5. the two-way U-Net deep neural network mine down-holes hazardous act intelligence according to claim 1-3 any one
It is classified discrimination method, it is characterised in that:In step 8, underground hazardous act is identified respectively for two-way network, when two branch roads
After identifying hazardous act simultaneously, by quaternary tree search method, classification, quantity and the locus that hazardous act occurs are entered
The quick identification of row, and carry out grading forewarning system and preserve sectional drawing, early warning releases after two-way does not identify hazardous act simultaneously.
6. two-way U-Net deep neural network mine down-holes hazardous act according to claim 5 is intelligently classified identification side
Method, it is characterised in that:The quaternary tree search method, specific implementation include following sub-step:
Step 8.1:Image-region is divided into upper left, upper right, lower-left and the quadrant of bottom right four;
Step 8.2:Since upper left, the pixel of hazardous act in quadrant, numbering 1-5, corresponding 5 kinds of hazardous acts, the back of the body are retrieved
Scape and not dangerous behavior numbering are 0;
Step 8.3:If the dangerous behavior pixel of some pixel or classification are more than or equal to 2, step 8.1 is transferred to;
Step 8.4:If a kind of hazardous act in quadrant be present, class variable deposit hazardous act class number;
Step 8.5:Count four quadrant hazardous act classifications, intensive variables;
Step 8.6:When dangerous behavior quadrant for 0 and less than 2 when, two level early warning;When the quadrant of dangerous behavior
During more than 2, one-level early warning;When different classes of hazardous act be present, one-level early warning.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711185883.8A CN107886081B (en) | 2017-11-23 | 2017-11-23 | Intelligent underground dangerous behavior grading identification method for double-path U-Net deep neural network mine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711185883.8A CN107886081B (en) | 2017-11-23 | 2017-11-23 | Intelligent underground dangerous behavior grading identification method for double-path U-Net deep neural network mine |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107886081A true CN107886081A (en) | 2018-04-06 |
CN107886081B CN107886081B (en) | 2021-02-02 |
Family
ID=61774857
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711185883.8A Active CN107886081B (en) | 2017-11-23 | 2017-11-23 | Intelligent underground dangerous behavior grading identification method for double-path U-Net deep neural network mine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107886081B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109191471A (en) * | 2018-08-28 | 2019-01-11 | 杭州电子科技大学 | Based on the pancreatic cell image partition method for improving U-Net network |
CN109685811A (en) * | 2018-12-24 | 2019-04-26 | 北京大学第三医院 | PET/CT hypermetabolism lymph node dividing method based on dual path U-net convolutional neural networks |
CN111079821A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Derailment automatic braking pull ring falling fault image identification method |
CN111435422A (en) * | 2019-01-11 | 2020-07-21 | 商汤集团有限公司 | Motion recognition method, control method and device, electronic device and storage medium |
CN113240294A (en) * | 2021-05-19 | 2021-08-10 | 武汉拾易鑫科技有限公司 | Mine safety information intelligent management system based on big data analysis |
CN117054396A (en) * | 2023-10-11 | 2023-11-14 | 天津大学 | Raman spectrum detection method and device based on double-path multiplicative neural network |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1144434A (en) * | 1994-07-29 | 1997-03-05 | Dva公司 | Video decompression |
CN102009879A (en) * | 2010-11-18 | 2011-04-13 | 无锡中星微电子有限公司 | Elevator automatic keying control system and method, face model training system and method |
CN202634608U (en) * | 2012-06-13 | 2012-12-26 | 黄炳 | Video monitoring device for coal mine |
US20130301918A1 (en) * | 2012-05-08 | 2013-11-14 | Videostir Ltd. | System, platform, application and method for automated video foreground and/or background replacement |
CN106686472A (en) * | 2016-12-29 | 2017-05-17 | 华中科技大学 | High-frame-rate video generation method and system based on depth learning |
-
2017
- 2017-11-23 CN CN201711185883.8A patent/CN107886081B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1144434A (en) * | 1994-07-29 | 1997-03-05 | Dva公司 | Video decompression |
CN102009879A (en) * | 2010-11-18 | 2011-04-13 | 无锡中星微电子有限公司 | Elevator automatic keying control system and method, face model training system and method |
US20130301918A1 (en) * | 2012-05-08 | 2013-11-14 | Videostir Ltd. | System, platform, application and method for automated video foreground and/or background replacement |
CN202634608U (en) * | 2012-06-13 | 2012-12-26 | 黄炳 | Video monitoring device for coal mine |
CN106686472A (en) * | 2016-12-29 | 2017-05-17 | 华中科技大学 | High-frame-rate video generation method and system based on depth learning |
Non-Patent Citations (1)
Title |
---|
钱彬等: "子块鉴别分析的路面裂缝检测", 《中国图象图形学报》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109191471A (en) * | 2018-08-28 | 2019-01-11 | 杭州电子科技大学 | Based on the pancreatic cell image partition method for improving U-Net network |
CN109685811A (en) * | 2018-12-24 | 2019-04-26 | 北京大学第三医院 | PET/CT hypermetabolism lymph node dividing method based on dual path U-net convolutional neural networks |
CN109685811B (en) * | 2018-12-24 | 2019-12-13 | 北京大学第三医院 | PET/CT high-metabolism lymph node segmentation method based on dual-path U-net convolutional neural network |
CN111435422A (en) * | 2019-01-11 | 2020-07-21 | 商汤集团有限公司 | Motion recognition method, control method and device, electronic device and storage medium |
CN111435422B (en) * | 2019-01-11 | 2024-03-08 | 商汤集团有限公司 | Action recognition method, control method and device, electronic equipment and storage medium |
CN111079821A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Derailment automatic braking pull ring falling fault image identification method |
CN113240294A (en) * | 2021-05-19 | 2021-08-10 | 武汉拾易鑫科技有限公司 | Mine safety information intelligent management system based on big data analysis |
CN117054396A (en) * | 2023-10-11 | 2023-11-14 | 天津大学 | Raman spectrum detection method and device based on double-path multiplicative neural network |
CN117054396B (en) * | 2023-10-11 | 2024-01-05 | 天津大学 | Raman spectrum detection method and device based on double-path multiplicative neural network |
Also Published As
Publication number | Publication date |
---|---|
CN107886081B (en) | 2021-02-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107886081A (en) | Two-way U Net deep neural network mine down-holes hazardous act is intelligently classified discrimination method | |
CN106412501B (en) | A kind of the construction safety behavior intelligent monitor system and its monitoring method of video | |
CN108319926A (en) | A kind of the safety cap wearing detecting system and detection method of building-site | |
CN109829429A (en) | Security protection sensitive articles detection method under monitoring scene based on YOLOv3 | |
CN106816038A (en) | A kind of inland waters abnormal behaviour ship automatic identification system and method | |
Hossain et al. | Assessing the efficacy of camera trapping as a tool for increasing detection rates of wildlife crime in tropical protected areas | |
CN107452104A (en) | A kind of control method for vehicle and system of the vehicle bayonet socket based on intelligent monitoring | |
CN107808139A (en) | A kind of real-time monitoring threat analysis method and system based on deep learning | |
CN111626162B (en) | Water rescue system based on space-time big data analysis and drowning warning condition prediction method | |
CN106713341A (en) | Network security early-warning method and system based on big data | |
CN106355162A (en) | Method for detecting intrusion on basis of video monitoring | |
CN110380897A (en) | Network security situation awareness model and method based on improved BP | |
CN110516138A (en) | A kind of food safety affair early warning system threatening information bank based on multi-source self refresh | |
CN109543631A (en) | A kind of fire image detection alarm method based on machine learning | |
CN112434828B (en) | Intelligent safety protection identification method in 5T operation and maintenance | |
CN112434827B (en) | Safety protection recognition unit in 5T operation and maintenance | |
CN108256447A (en) | A kind of unmanned plane video analysis method based on deep neural network | |
CN104392464A (en) | Human intrusion detection method based on color video image | |
CN112766091B (en) | Video unsafe behavior recognition system and method based on human skeleton key points | |
Sugumar et al. | An improved real time image detection system for elephant intrusion along the forest border areas | |
CN106127814A (en) | A kind of wisdom gold eyeball identification gathering of people is fought alarm method and device | |
CN106875692A (en) | Vehicle integration early warning system and its method based on big data | |
CN108108695A (en) | Fire defector recognition methods based on Infrared video image | |
CN111539325A (en) | Forest fire detection method based on deep learning | |
CN109831634A (en) | The density information of target object determines method and device |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |