CN109685066A - A kind of mine object detection and recognition method based on depth convolutional neural networks - Google Patents

A kind of mine object detection and recognition method based on depth convolutional neural networks Download PDF

Info

Publication number
CN109685066A
CN109685066A CN201811582240.1A CN201811582240A CN109685066A CN 109685066 A CN109685066 A CN 109685066A CN 201811582240 A CN201811582240 A CN 201811582240A CN 109685066 A CN109685066 A CN 109685066A
Authority
CN
China
Prior art keywords
target
mine
frame
serial
training
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
Application number
CN201811582240.1A
Other languages
Chinese (zh)
Other versions
CN109685066B (en
Inventor
张帆
崔东林
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.)
China University of Mining and Technology Beijing CUMTB
Original Assignee
China University of Mining and Technology Beijing CUMTB
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 China University of Mining and Technology Beijing CUMTB filed Critical China University of Mining and Technology Beijing CUMTB
Priority to CN201811582240.1A priority Critical patent/CN109685066B/en
Publication of CN109685066A publication Critical patent/CN109685066A/en
Application granted granted Critical
Publication of CN109685066B publication Critical patent/CN109685066B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/14—Image acquisition
    • G06V30/148—Segmentation of character regions
    • G06V30/153—Segmentation of character regions using recognition of characters or words
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield 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
    • G06N3/04—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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • G06V10/267—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 by performing operations on regions, e.g. growing, shrinking or watersheds

Landscapes

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

Abstract

The invention belongs to target detections and intelligent recognition field, disclose a kind of mine object detection and recognition method of depth convolutional neural networks.The method key step includes: the first step, makes target data set, constructs mode input using a frame of mine video image acquisition equipment capture underground coal mine target original image, accordingly production training verifying collection and test set；Second step, training objective detect network model, are verified using training and collect off-line training network model, until model accuracy rate with higher；Third step is detected using the Serial No. that trained target detection network model carries target in test set, and obtains the four-dimensional coordinate of the Serial No.；4th step intercepts the Serial No. region in picture, cutting operation, and is sequentially sent to be identified in LeNet-5 network, and the identity of mobile target is determined according to recognition result.This method can effectively improve the speed and precision of mine object detection and recognition.

Description

A kind of mine object detection and recognition method based on depth convolutional neural networks
Technical field
The invention belongs to target detections and intelligent recognition field, relate in particular to a kind of based on depth convolutional neural networks Mine object detection and recognition method.
Background technique
Intelligentized mining is the development trend of mine safety, high-efficiency intensifying production, and research underground operators, locomotive are set The mobile target such as standby and robot is accurately detected to be identified with real-time tracking, to the exploitation of guarantee mine intelligent and safe and improves coal mine calamity Evil intelligent early-warning is of great significance.The recognition methods of the Moving target detections such as existing mine personnel and underground locomotive mainly uses quiet The Radio Frequency Identification Technology of state, but aforesaid way cannot achieve and the multidimensional information of target is detected and tracked and identified, and especially exist It is difficult to realize that mine movable target is carried out real-time tracking and accurately identified in underground spray dust, low light environment.In recent years, for The theoretical research of deep learning has caused many scholars at home and abroad to pay close attention to, and depth convolutional neural networks (DCNN) are considered a kind of It is suitble to the important deep learning method of target detection and classification task, and there is accuracy of identification high, strong antijamming capability and can Remote the features such as obtaining target image, the application in the fields such as intelligent monitoring, moving object detection and identification, vision guided navigation As research hotspot.Therefore, in order to overcome the shortcomings of existing mine target identification technology, the present invention proposes to roll up using based on depth The mine object detection and recognition method of product neural network, realization are accurately detected and are tracked and identified to mine target.
Summary of the invention
The present invention proposes a kind of mine object detection and recognition method.It is mainly used for the class for solving to accurately identify mine target Not, precise positioning target position in the picture, size, realize the detection of mine personnel, locomotive and robot movable target with The vision guided navigation and positioning function of the coal petrography of identification and mine working face identification and environment scene.
The present invention is a kind of mine object detection and recognition method based on depth convolutional neural networks, including model training Method, target position detection method and target category recognition methods.
The mine object detection and recognition method, implementation step include:
Step 1, the data set for making target sample, it is former using mine video image acquisition equipment capture underground coal mine target One frame of beginning image or environment scene constructs mode input, makes the number of targets containing Serial No. and its location information accordingly It is divided into training verifying collection and test set according to collection, and by the data set；
The network model that step 2, training objective detect is used for target using the training verifying collection off-line training in step 1 The network model of detection, until the model reaches the accuracy rate of Serial No. in detection target image；
Step 3 examines the target that Serial No. is carried in test set using trained target detection network model It surveys, and obtains the four-dimensional coordinate (cx, cy, w, h) of the Serial No., wherein cx, cy respectively indicate candidate frame center in the picture Transverse and longitudinal coordinate, w, h indicate candidate frame width and height；
Step 4, the Serial No. four-dimensional coordinate detected according to step 3 carry out the Serial No. region in image special Sign is extracted and Character segmentation operation, then sequentially inputs in LeNet-5 network single character and identifies, and will identify that Character arranged in sequence finally obtains the content information of the Serial No., determines the identity of target accordingly；
Step 2 further comprises following sub-step:
2.1 initialization: using the main net network layers portion of ImageNet pre-training model VGG16 netinit SSD network Point, it uses zero-mean, variance to initialize the extension layer part of SSD network for 0.01 Gaussian Profile, sets confidence threshold and instruction Practice threshold value；
It extracts 2.2 candidate regions: inputting mobile target training sample, using convolution kernel from Conv4_3, Conv7, Conv8_ 2, a series of candidate frames are extracted on the characteristic pattern of six different scales of Conv9_2, Conv10_2 and Conv11_2；To the true of target Real frame markup information is pre-processed, and is mapped that on corresponding candidate frame；It is chosen just according to the confidence threshold of setting Negative sample is matched according to matching strategy with the true frame of target；
2.3 error calculations: calculate the true frame that the positive negative sample chosen is matched in step 2.2 position offset and The error penalty values of classification confidence level；
2.4 right value updates: the weight of network model is updated using the deep learning back-propagation algorithm declined based on gradient；
2.5 iteration convergences: the entire mobile target training set of traversal repeats step 2.2~step 2.4, iteration is simultaneously counted Error amount of the network model on target verification collection is calculated, until the error amount reaches trained threshold value；And
Step 3 further comprises following sub-step:
3.1 are input to a width image to be detected in the network model, automatically extract target image characteristics；
3.2 according to the target image characteristics extracted in step 3.1, choose Conv4_3, Conv7, Conv8_2, Conv9_2, The characteristic pattern of six different scales of Conv10_2 and Conv11_2, from a series of prediction blocks for wherein extracting highest scoring；
3.3 setting target size threshold values and confidence threshold, by non-maxima suppression method, from the system in step 3.2 Column prediction block removes that Duplication is higher and the lower prediction block of confidence level, retains last target prediction frame；
Step 4 further comprises following sub-step:
4.1, which choose a sample from MNIST data set, inputs LeNet-5 network, rolls up to the feature vector of sample Product calculating, nonlinear change, down-sampling and full connection calculate, and export a prediction result；
4.2, by stochastic gradient descent algorithm optimization object function, obtain the updated value of network parameter；
4.3 repeat step 4.1~step 4.2, and iteration updates the weight of network, loss function are minimized, until net Until the error of network is reduced to preset trained threshold value；
4.4, according to the position coordinates in target detection frame, are cut using region to be identified of the openCV to target image It takes, gray processing and binary conversion treatment, and carries out Character segmentation using the mode of pixel projection；
4.5 identify the target signature after segmentation, and the characteristic character that will identify that is merged into Serial No., to time It selects target to be identified, exports the feature classification of test image；
4.6 finally retrieve from target identities identification database and confirm target identity information.
The mine object detection and recognition method, wherein the Serial No., refer to underground coal mine target carry can be only The numeric identifier of one identification target identity information；The location information of the Serial No., digital sequence in feeling the pulse with the finger-tip mark original image Arrange the four-dimensional coordinate information of region.
The mine object detecting and tracking recognition methods, wherein the mine video image acquisition equipment includes mining Intrinsic safety type visual sensor, Mine-used I. S camera and Mine-used I. S video camera.
The mine object detecting and tracking recognition methods, for realizing to underground operators, locomotive equipment and movement The coal petrography identification of object detection and recognition and mine working face of robot and the vision guided navigation of environment scene and positioning function Energy.
The beneficial effects of the present invention are:
The invention use depth convolutional neural networks method, to mine target image by network model training output and The corresponding target signature classification of self-position generates the region candidate frame of different scale on multilayer feature figure, avoids selection A large amount of calculating of property searching method；By expanding local receptor field and reducing the method for convolution filter size, obtain more aobvious The Analysis On Multi-scale Features for writing distinction and expressiveness, effectively increase the accuracy rate of target detection, to realize different to mine big Accurate, the quick detection and identification of Small object.The present invention is to guarantee personnel in the pit's hedging, collision prevention of vehicle and intelligent and safe exploitation tool It is significant.
Detailed description of the invention
Fig. 1 is the mine object detection and recognition flow diagram according to the embodiment of the present invention
Fig. 2 is the network architecture figure according to the embodiment of the present invention
Fig. 3 is the network model training flow chart according to the embodiment of the present invention
Fig. 4 is the mine target position overhaul flow chart according to the embodiment of the present invention
Fig. 5 is according to the embodiment of the present invention to mine target category identification process figure
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawing to tool of the invention Body embodiment is described in detail.
Fig. 1 is mine object detection and recognition method overall procedure block diagram.Realization process are as follows: adopted using mine video image Collection equipment captures a frame of underground coal mine target original image or environment scene to construct mode input, and production contains number accordingly The target data set of sequence and its location information, and the data set is divided into training verifying collection and test set；It is verified using training Collect off-line training be used for target detection network model, until the model reach detection target image in Serial No. it is accurate Rate；The target for carrying Serial No. in test set is detected using trained target detection network model, and is somebody's turn to do The four-dimensional coordinate (cx, cy, w, h) of Serial No., wherein cx, cy respectively indicate the transverse and longitudinal coordinate of candidate frame center in the picture, The width and height of w, h expression candidate frame；According to the Serial No. four-dimensional coordinate detected, the Serial No. region in image is carried out Feature extraction and Character segmentation operation, then single character is sequentially input in LeNet-5 network and is identified, and will identify that Character arranged in sequence, finally obtain the content information of the Serial No., determine the identity of target accordingly.
The network architecture figure of Fig. 2 to realize the present invention.Main net network layers are using VGG16 network as basic network knot Structure is mainly used for convolution feature extraction, extracts picture feature；Extension layer is then the convolutional layer that some bulks are gradually reduced, It is mainly used for multilayer feature fusion, extracts the candidate frame under different scale.Choose Conv4_3, Conv7, Conv8_2, Conv9_ 2, six different scale characteristic patterns such as Conv10_2 and Conv11_2, generate a series of candidate frames on each of which characteristic face, Then candidate frame is matched with the true callout box of target according to matching strategy again, is handled finally by NMS strategy each Redundancy frame near true frame, filters extra non-textual information, to obtain final testing result.
Fig. 3 is the network model training flow chart of the embodiment of the present invention.Include the following steps:
Step 1. initialization: using the main net network layers of ImageNet pre-training model VGG16 netinit SSD network Part, use equal value zero, variance for 0.01 Gaussian Profile initialize SSD network extension layer part, setting confidence threshold and Training threshold value；
Extract step 2. candidate region: inputting mobile target training sample, using convolution kernel from Conv4_3, Conv7, A series of candidate frames are extracted on the characteristic pattern of six different scales of Conv8_2, Conv9_2, Conv10_2 and Conv11_2；To mesh The true frame markup information of target is pre-processed, and is mapped that on corresponding candidate frame；According to the confidence threshold of setting come Positive negative sample is chosen, is matched according to matching strategy with the true frame of target；
Step 3. error calculation: the position offset of candidate frame and the error damage of classification confidence level are calculated using loss function Mistake value；
DefinitionIt is matched for i-th of candidate frame of mine target with j-th of true frame of p classification, if matching It is successful then be 1, be otherwise 0, according to above-mentioned matching strategy, centainly have
Objective function is made of the Classification Loss of candidate frame and positioning loss, and total loss function may be expressed as:
In formula, x is the three-dimensional matrice of digital picture, and c is the confidence level of sample, and l is candidate frame, and g is true frame；N be with The candidate frame number that true frame matches；λ is regular terms weight；LconfFor category task loss function, the classification of candidate frame is indicated Error；LlosFor position offset loss function, the position deviation of candidate frame and true frame is indicated.
The category task loss function L of targetconfIt is defined as follows:
In formula,J-th true frame matching of i-th of candidate frame to classification as p is represented,For in i-th of candidate frame The confidence level of classification p,For the confidence level of the background classes in i-th of target candidate frame.
The position offset loss function L of target candidate framelosIt indicates are as follows:
In formula, (cx, cy) is the centre coordinate of compensated candidate frame, and (w, h) is the width and height of candidate frame,Expression pair The correction amount of candidate frame position,Indicate position offset of the true frame relative to candidate frame.
Step 4. right value update: SSD network model is updated using the deep learning back-propagation algorithm declined based on gradient Weight；
Step 5. iteration convergence: traversing entire target training set, repeats step 2~step 4, and iteration simultaneously calculates SSD Error amount of the model on mobile target verification collection, until the error amount reaches minimum.
Fig. 4 is mine target position overhaul flow chart.Include the following steps:
One test set picture is input in network model by step 1., automatically extracts Target Photo feature；
Step 2. chooses Conv4_3, Conv7, Conv8_2, Conv9_ according to the Target Photo feature extracted in step 1 2, the characteristic pattern of six different scales such as Conv10_2 and Conv11_2, from a series of predictions for wherein extracting highest scoring Frame；
Step 3. sets target size threshold value and confidence threshold, by non-maxima suppression (NMS) method, from step 3.2 In a series of removal of prediction blocks Duplication is higher and the lower prediction block of confidence level, retain last target prediction frame.
Fig. 5 is mine target category identification process figure.Include the following steps:
Step 1. intercepts new picture using openCV according to the position coordinates in target detection frame, to the area of interception Domain carries out gray processing and binary conversion treatment, and carries out Character segmentation using the mode of pixel projection；
Step 2. identifies the target signature after segmentation, and the characteristic character that will identify that is merged into Serial No., Candidate target is identified, the feature classification of test image is exported；
Target identity information is finally retrieved from target identities identification database and confirmed to step 3..
Obviously, those skilled in the art should be understood that recognition methods involved by the present invention and above-described embodiment, remove conduct The positioning of mine target is applied to outside underground coal mine environment, by being also applied for the non-coal mines such as nonmetallic and metal after being suitably modified Mobile monitoring, tracking and positioning and the mobile operating equipment of downhole intelligent working face track and identify, be accurately positioned and can Vision guided navigation and positioning function depending on changing the coal petrography identification and environment scene of monitoring and mine working face.In this way the present invention not Non-coal mine, intelligent work face mobile monitor and internet of things equipment in addition to underground coal mine Moving objects location is limited precisely to know Not with positioning etc. fields of communication technology.
The above content is the further descriptions for combining specific preferred embodiment mode to be the present invention, cannot recognize Determine a specific embodiment of the invention and be only limitted to this, for those of ordinary skill in the art to which the present invention belongs, not Under the premise of being detached from mentality of designing of the present invention, several simple replacements and change can be also carried out, all shall be regarded as belonging to the present invention Protection scope involved in the claims submitted.

Claims (4)

1. a kind of mine object detection and recognition method based on depth convolutional neural networks, it is characterised in that including walking as follows It is rapid:
Step 1, the data set for making target sample capture underground coal mine target original graph using mine video image acquisition equipment As or a frame of environment scene construct mode input, make the target data containing Serial No. and its location information accordingly Collection, and the data set is divided into training verifying collection and test set；
The network model that step 2, training objective detect is used for target detection using the training verifying collection off-line training in step 1 Network model, until the model reach detection target image in Serial No. accuracy rate；
Step 3 detects the target that Serial No. is carried in test set using trained target detection network model, and Obtain the four-dimensional coordinate (cx, cy, w, h) of the Serial No., wherein cx, cy respectively indicate the cross of candidate frame center in the picture Ordinate, w, h indicate the width and height of candidate frame；
Step 4, the Serial No. four-dimensional coordinate detected according to step 3 carry out feature to the Serial No. region in image and mention It takes and is operated with Character segmentation, the character that then single character is sequentially input in LeNet-5 network and is identified, and will identify that Arranged in sequence finally obtains the content information of the Serial No., determines the identity of mobile target accordingly；
It is further characterized in that step 2 further comprises following sub-step:
2.1 initialization: it using the main net network layers part of ImageNet pre-training model VGG16 netinit SSD network, adopts The extension layer part of SSD network, setting confidence threshold and training threshold are initialized with zero-mean, the Gaussian Profile that variance is 0.01 Value；
Extract 2.2 candidate regions: inputting mobile target training sample, using convolution kernel from Conv4_3, Conv7, Conv8_2, A series of candidate frames are extracted on the characteristic pattern of six different scales of Conv9_2, Conv10_2 and Conv11_2；To the true of target Frame markup information is pre-processed, and is mapped that on corresponding candidate frame, is chosen according to the confidence threshold of setting positive and negative Sample is matched according to matching strategy with the true frame of target；
2.3 error calculations: calculating the position offset of candidate frame and the error penalty values of classification confidence level using loss function, public Formula is
In formula, x is the three-dimensional matrice of digital picture, and c is the confidence level of sample, and l is candidate frame, and g is true frame；N be with it is true The candidate frame number that frame matches；λ is regular terms weight；LconfFor category task loss function, the error in classification of candidate frame is indicated； LlosFor position offset loss function, the position deviation of candidate frame and true frame is indicated；
2.4 right value updates: the weight of network model is updated using the deep learning back-propagation algorithm declined based on gradient；
2.5 iteration convergences: the entire mobile target training set of traversal repeats step 2.2~step 2.4, iteration simultaneously calculates net Error amount of the network model on target verification collection, until the error amount reaches trained threshold value；And
Step 3 further comprises following sub-step:
3.1 are input to a width image to be detected in the network model, automatically extract target image characteristics；
3.2 according to the target image characteristics extracted in step 3.1, choose Conv4_3, Conv7, Conv8_2, Conv9_2, The characteristic pattern of six different scales of Conv10_2 and Conv11_2, from a series of prediction blocks for wherein extracting highest scoring；
3.3 setting target size threshold values and confidence threshold, by non-maxima suppression method, from a series of pre- in step 3.2 Surveying frame removal, Duplication is higher and the lower prediction block of confidence level, retains last target prediction frame；
It is further characterized in that step 4 further comprises following sub-step:
4.1, which choose a sample from MNIST data set, inputs LeNet-5 network, carries out convolution meter to the feature vector of sample Calculation, nonlinear change, down-sampling and full connection calculate, and export a prediction result；
4.2, by stochastic gradient descent algorithm optimization object function, obtain the updated value of network parameter；
4.3 repeat step 4.1~step 4.2, and iteration updates the weight of network, loss function are minimized, until network Until error is reduced to preset trained threshold value；
4.4 according to the position coordinates in target detection frame, are intercepted using to be identified region of the openCV to target image, ash Degreeization and binary conversion treatment, and Character segmentation is carried out using the mode of pixel projection；
4.5 identify the target signature after segmentation, and the characteristic character that will identify that is merged into Serial No., to candidate mesh Mark is identified, the feature classification of test image is exported；
4.6 finally retrieve from target identities identification database and confirm target identity information.
2. the method according to claim 1, wherein the Serial No., refers to the energy that underground coal mine target carries The numeric identifier of unique identification target identity information；The location information of the Serial No., number in feeling the pulse with the finger-tip mark original image The four-dimensional coordinate information of sequence region.
3. the method according to claim 1, wherein the mine video image acquisition equipment includes mine intrinsic safety Type visual sensor, Mine-used I. S camera and Mine-used I. S video camera.
4. the method according to claim 1, wherein for realizing to underground operators, locomotive equipment and shifting The coal petrography identification of object detection and recognition and mine working face of mobile robot and the vision guided navigation of environment scene and positioning function Energy.
CN201811582240.1A 2018-12-24 2018-12-24 Mine target detection and identification method based on deep convolutional neural network Active CN109685066B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811582240.1A CN109685066B (en) 2018-12-24 2018-12-24 Mine target detection and identification method based on deep convolutional neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811582240.1A CN109685066B (en) 2018-12-24 2018-12-24 Mine target detection and identification method based on deep convolutional neural network

Publications (2)

Publication Number Publication Date
CN109685066A true CN109685066A (en) 2019-04-26
CN109685066B CN109685066B (en) 2021-03-09

Family

ID=66189048

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811582240.1A Active CN109685066B (en) 2018-12-24 2018-12-24 Mine target detection and identification method based on deep convolutional neural network

Country Status (1)

Country Link
CN (1) CN109685066B (en)

Cited By (36)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110119701A (en) * 2019-04-30 2019-08-13 东莞恒创智能科技有限公司 The coal mine fully-mechanized mining working unsafe acts recognition methods of view-based access control model relationship detection
CN110135289A (en) * 2019-04-28 2019-08-16 北京天地玛珂电液控制系统有限公司 A kind of underground coal mine intelligent use cloud service platform based on deep learning
CN110197152A (en) * 2019-05-28 2019-09-03 南京邮电大学 A kind of road target recognition methods for automated driving system
CN110197158A (en) * 2019-05-31 2019-09-03 广西南宁市博睿通软件技术有限公司 A kind of security cloud system and its application
CN110215232A (en) * 2019-04-30 2019-09-10 南方医科大学南方医院 Ultrasonic patch analysis method in coronary artery based on algorithm of target detection
CN110276352A (en) * 2019-06-28 2019-09-24 拉扎斯网络科技（上海）有限公司 Index identification method, device, electronic equipment and computer readable storage medium
CN110363122A (en) * 2019-07-03 2019-10-22 昆明理工大学 A kind of cross-domain object detection method based on multilayer feature alignment
CN110478911A (en) * 2019-08-13 2019-11-22 苏州钛智智能科技有限公司 The unmanned method of intelligent game vehicle and intelligent vehicle, equipment based on machine learning
CN110532937A (en) * 2019-08-26 2019-12-03 北京航空航天大学 Method for distinguishing is known to targeting accuracy with before disaggregated model progress train based on identification model
CN110569843A (en) * 2019-09-09 2019-12-13 中国矿业大学(北京) Intelligent detection and identification method for mine target
CN110689021A (en) * 2019-10-17 2020-01-14 哈尔滨理工大学 Real-time target detection method in low-visibility environment based on deep learning
CN110807384A (en) * 2019-10-24 2020-02-18 华东计算技术研究所（中国电子科技集团公司第三十二研究所） Small target detection method and system under low visibility
CN110852896A (en) * 2019-12-22 2020-02-28 上海眼控科技股份有限公司 Date validity judgment method, date validity judgment device, date validity judgment equipment and storage medium
CN110930008A (en) * 2019-11-15 2020-03-27 中国矿业大学 Mine disaster event detection method based on convolutional neural network
CN111105590A (en) * 2019-12-27 2020-05-05 深圳前海微众银行股份有限公司 Alarm method and device
CN111161227A (en) * 2019-12-20 2020-05-15 成都数之联科技有限公司 Target positioning method and system based on deep neural network
CN111208759A (en) * 2019-12-30 2020-05-29 中国矿业大学(北京) Digital twin intelligent monitoring system for unmanned fully mechanized coal mining face of mine
CN111738257A (en) * 2020-06-18 2020-10-02 福建师范大学 Different network model cascade target accurate reasoning system and method based on deep learning
CN111932562A (en) * 2020-09-22 2020-11-13 平安科技（深圳）有限公司 Image identification method and device based on CT sequence, electronic equipment and medium
CN111931857A (en) * 2020-08-14 2020-11-13 桂林电子科技大学 MSCFF-based low-illumination target detection method
WO2020253499A1 (en) * 2019-06-17 2020-12-24 平安科技（深圳）有限公司 Video object acceleration monitoring method and apparatus, and server and storage medium
CN112396648A (en) * 2019-08-12 2021-02-23 北京全路通信信号研究设计院集团有限公司 Target identification method and system capable of positioning mass center of target object
CN112508935A (en) * 2020-12-22 2021-03-16 郑州金惠计算机系统工程有限公司 Product packaging detection method and system based on deep learning and product packaging sorting system
CN112686888A (en) * 2021-01-27 2021-04-20 上海电气集团股份有限公司 Method, system, equipment and medium for detecting cracks of concrete sleeper
CN112800952A (en) * 2021-01-27 2021-05-14 山东大学 Marine organism identification method and system based on improved SSD algorithm
CN113408361A (en) * 2021-05-25 2021-09-17 中国矿业大学 Deep learning-based mining conveyor belt bulk material detection method and system
WO2021226977A1 (en) * 2020-05-15 2021-11-18 安徽中科智能感知产业技术研究院有限责任公司 Method and platform for dynamically monitoring typical ground features in mining on the basis of multi-source remote sensing data fusion and deep neural network
CN113887380A (en) * 2021-09-28 2022-01-04 淮北矿业股份有限公司淮北选煤厂 Intelligent sample preparation system for coal samples
CN114267155A (en) * 2021-11-05 2022-04-01 国能大渡河革什扎水电开发有限公司 Geological disaster monitoring and early warning system based on video recognition technology
CN114429607A (en) * 2022-01-24 2022-05-03 中南大学 Transformer-based semi-supervised video object segmentation method
CN115330702A (en) * 2022-08-01 2022-11-11 无锡雪浪数制科技有限公司 Beverage bottle filling defect identification method based on deep vision
CN115583510A (en) * 2022-09-29 2023-01-10 华能伊敏煤电有限责任公司 Automatic soil discharging control method and system based on laser scanner
CN116030418A (en) * 2023-02-14 2023-04-28 北京建工集团有限责任公司 Automobile lifting line state monitoring system and method
CN111626279B (en) * 2019-10-15 2023-06-02 西安网算数据科技有限公司 Negative sample labeling training method and highly-automatic bill identification method
CN116524017A (en) * 2023-03-13 2023-08-01 明创慧远科技集团有限公司 Underground detection, identification and positioning system for mine
CN116634638A (en) * 2023-05-16 2023-08-22 珠海光通智装科技有限公司 Light control strategy generation method, light control method and related device

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103488986A (en) * 2013-09-18 2014-01-01 西安理工大学 Method for segmenting and extracting characters in self-adaptation mode
US20170206431A1 (en) * 2016-01-20 2017-07-20 Microsoft Technology Licensing, Llc Object detection and classification in images
US20180158189A1 (en) * 2016-12-07 2018-06-07 Samsung Electronics Co., Ltd. System and method for a deep learning machine for object detection
CN108182413A (en) * 2017-12-29 2018-06-19 中国矿业大学(北京) A kind of mine movable object detecting and tracking recognition methods
CN108564097A (en) * 2017-12-05 2018-09-21 华南理工大学 A kind of multiscale target detection method based on depth convolutional neural networks
CN109002824A (en) * 2018-06-27 2018-12-14 淮阴工学院 A kind of architectural drawing label information detection method based on OpenCV

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103488986A (en) * 2013-09-18 2014-01-01 西安理工大学 Method for segmenting and extracting characters in self-adaptation mode
US20170206431A1 (en) * 2016-01-20 2017-07-20 Microsoft Technology Licensing, Llc Object detection and classification in images
US20180158189A1 (en) * 2016-12-07 2018-06-07 Samsung Electronics Co., Ltd. System and method for a deep learning machine for object detection
CN108564097A (en) * 2017-12-05 2018-09-21 华南理工大学 A kind of multiscale target detection method based on depth convolutional neural networks
CN108182413A (en) * 2017-12-29 2018-06-19 中国矿业大学(北京) A kind of mine movable object detecting and tracking recognition methods
CN109002824A (en) * 2018-06-27 2018-12-14 淮阴工学院 A kind of architectural drawing label information detection method based on OpenCV

Cited By (53)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110135289A (en) * 2019-04-28 2019-08-16 北京天地玛珂电液控制系统有限公司 A kind of underground coal mine intelligent use cloud service platform based on deep learning
CN110215232A (en) * 2019-04-30 2019-09-10 南方医科大学南方医院 Ultrasonic patch analysis method in coronary artery based on algorithm of target detection
CN110119701A (en) * 2019-04-30 2019-08-13 东莞恒创智能科技有限公司 The coal mine fully-mechanized mining working unsafe acts recognition methods of view-based access control model relationship detection
CN110119701B (en) * 2019-04-30 2023-04-07 东莞恒创智能科技有限公司 Visual relationship detection-based coal mine fully mechanized coal mining face unsafe behavior identification method
CN110197152B (en) * 2019-05-28 2022-08-26 南京邮电大学 Road target identification method for automatic driving system
CN110197152A (en) * 2019-05-28 2019-09-03 南京邮电大学 A kind of road target recognition methods for automated driving system
CN110197158A (en) * 2019-05-31 2019-09-03 广西南宁市博睿通软件技术有限公司 A kind of security cloud system and its application
US11816570B2 (en) 2019-06-17 2023-11-14 Ping An Technology (Shenzhen) Co., Ltd. Method for accelerated detection of object in videos, server, and non-transitory computer readable storage medium
WO2020253499A1 (en) * 2019-06-17 2020-12-24 平安科技（深圳）有限公司 Video object acceleration monitoring method and apparatus, and server and storage medium
CN110276352A (en) * 2019-06-28 2019-09-24 拉扎斯网络科技（上海）有限公司 Index identification method, device, electronic equipment and computer readable storage medium
CN110363122A (en) * 2019-07-03 2019-10-22 昆明理工大学 A kind of cross-domain object detection method based on multilayer feature alignment
CN110363122B (en) * 2019-07-03 2022-10-11 昆明理工大学 Cross-domain target detection method based on multi-layer feature alignment
CN112396648A (en) * 2019-08-12 2021-02-23 北京全路通信信号研究设计院集团有限公司 Target identification method and system capable of positioning mass center of target object
CN112396648B (en) * 2019-08-12 2024-04-30 北京全路通信信号研究设计院集团有限公司 Target identification method and system capable of positioning mass center of target object
CN110478911A (en) * 2019-08-13 2019-11-22 苏州钛智智能科技有限公司 The unmanned method of intelligent game vehicle and intelligent vehicle, equipment based on machine learning
CN110532937A (en) * 2019-08-26 2019-12-03 北京航空航天大学 Method for distinguishing is known to targeting accuracy with before disaggregated model progress train based on identification model
CN110532937B (en) * 2019-08-26 2022-03-08 北京航空航天大学 Method for accurately identifying forward targets of train based on identification model and classification model
CN110569843A (en) * 2019-09-09 2019-12-13 中国矿业大学(北京) Intelligent detection and identification method for mine target
CN111626279B (en) * 2019-10-15 2023-06-02 西安网算数据科技有限公司 Negative sample labeling training method and highly-automatic bill identification method
CN110689021A (en) * 2019-10-17 2020-01-14 哈尔滨理工大学 Real-time target detection method in low-visibility environment based on deep learning
CN110807384A (en) * 2019-10-24 2020-02-18 华东计算技术研究所（中国电子科技集团公司第三十二研究所） Small target detection method and system under low visibility
CN110930008B (en) * 2019-11-15 2022-03-15 中国矿业大学 Mine disaster event detection method based on convolutional neural network
CN110930008A (en) * 2019-11-15 2020-03-27 中国矿业大学 Mine disaster event detection method based on convolutional neural network
CN111161227A (en) * 2019-12-20 2020-05-15 成都数之联科技有限公司 Target positioning method and system based on deep neural network
CN110852896A (en) * 2019-12-22 2020-02-28 上海眼控科技股份有限公司 Date validity judgment method, date validity judgment device, date validity judgment equipment and storage medium
CN111105590A (en) * 2019-12-27 2020-05-05 深圳前海微众银行股份有限公司 Alarm method and device
CN111208759A (en) * 2019-12-30 2020-05-29 中国矿业大学(北京) Digital twin intelligent monitoring system for unmanned fully mechanized coal mining face of mine
CN111208759B (en) * 2019-12-30 2021-02-02 中国矿业大学(北京) Digital twin intelligent monitoring system for unmanned fully mechanized coal mining face of mine
WO2021226977A1 (en) * 2020-05-15 2021-11-18 安徽中科智能感知产业技术研究院有限责任公司 Method and platform for dynamically monitoring typical ground features in mining on the basis of multi-source remote sensing data fusion and deep neural network
CN111738257B (en) * 2020-06-18 2023-05-23 福建师范大学 Deep learning-based accurate reasoning system and method for cascading targets of different network models
CN111738257A (en) * 2020-06-18 2020-10-02 福建师范大学 Different network model cascade target accurate reasoning system and method based on deep learning
CN111931857A (en) * 2020-08-14 2020-11-13 桂林电子科技大学 MSCFF-based low-illumination target detection method
CN111931857B (en) * 2020-08-14 2022-09-02 桂林电子科技大学 MSCFF-based low-illumination target detection method
CN111932562A (en) * 2020-09-22 2020-11-13 平安科技（深圳）有限公司 Image identification method and device based on CT sequence, electronic equipment and medium
CN111932562B (en) * 2020-09-22 2021-01-19 平安科技（深圳）有限公司 Image identification method and device based on CT sequence, electronic equipment and medium
CN112508935A (en) * 2020-12-22 2021-03-16 郑州金惠计算机系统工程有限公司 Product packaging detection method and system based on deep learning and product packaging sorting system
CN112800952A (en) * 2021-01-27 2021-05-14 山东大学 Marine organism identification method and system based on improved SSD algorithm
CN112800952B (en) * 2021-01-27 2022-09-16 山东大学 Marine organism identification method and system based on improved SSD algorithm
CN112686888A (en) * 2021-01-27 2021-04-20 上海电气集团股份有限公司 Method, system, equipment and medium for detecting cracks of concrete sleeper
CN113408361B (en) * 2021-05-25 2023-09-19 中国矿业大学 Mining conveyor belt massive material detection method and system based on deep learning
CN113408361A (en) * 2021-05-25 2021-09-17 中国矿业大学 Deep learning-based mining conveyor belt bulk material detection method and system
CN113887380A (en) * 2021-09-28 2022-01-04 淮北矿业股份有限公司淮北选煤厂 Intelligent sample preparation system for coal samples
CN114267155A (en) * 2021-11-05 2022-04-01 国能大渡河革什扎水电开发有限公司 Geological disaster monitoring and early warning system based on video recognition technology
CN114429607A (en) * 2022-01-24 2022-05-03 中南大学 Transformer-based semi-supervised video object segmentation method
CN114429607B (en) * 2022-01-24 2024-03-29 中南大学 Transformer-based semi-supervised video object segmentation method
CN115330702A (en) * 2022-08-01 2022-11-11 无锡雪浪数制科技有限公司 Beverage bottle filling defect identification method based on deep vision
CN115583510A (en) * 2022-09-29 2023-01-10 华能伊敏煤电有限责任公司 Automatic soil discharging control method and system based on laser scanner
CN115583510B (en) * 2022-09-29 2023-10-27 华能伊敏煤电有限责任公司 Automatic soil discharging control method and system based on laser scanner
CN116030418A (en) * 2023-02-14 2023-04-28 北京建工集团有限责任公司 Automobile lifting line state monitoring system and method
CN116030418B (en) * 2023-02-14 2023-09-12 北京建工集团有限责任公司 Automobile lifting line state monitoring system and method
CN116524017B (en) * 2023-03-13 2023-09-19 明创慧远科技集团有限公司 Underground detection, identification and positioning system for mine
CN116524017A (en) * 2023-03-13 2023-08-01 明创慧远科技集团有限公司 Underground detection, identification and positioning system for mine
CN116634638A (en) * 2023-05-16 2023-08-22 珠海光通智装科技有限公司 Light control strategy generation method, light control method and related device

Also Published As

Publication number Publication date
CN109685066B (en) 2021-03-09

Similar Documents

Publication Publication Date Title
CN109685066A (en) A kind of mine object detection and recognition method based on depth convolutional neural networks
CN108182413A (en) A kind of mine movable object detecting and tracking recognition methods
CN107818571B (en) Ship automatic tracking method and system based on deep learning network and average drifting
CN106127204B (en) A kind of multi-direction meter reading Region detection algorithms of full convolutional neural networks
WO2020253308A1 (en) Human-machine interaction behavior security monitoring and forewarning method for underground belt transportation-related personnel
CN110232350A (en) A kind of real-time water surface multiple mobile object detecting and tracking method based on on-line study
CN104931960B (en) Trend message and radar target situation information full flight path segment data correlating method
CN108710868A (en) A kind of human body critical point detection system and method based under complex scene
CN107817798A (en) A kind of farm machinery barrier-avoiding method based on deep learning system
CN109919974A (en) Online multi-object tracking method based on the more candidate associations of R-FCN frame
CN110569843B (en) Intelligent detection and identification method for mine target
CN110110649A (en) Alternative method for detecting human face based on directional velocity
CN104134077B (en) A kind of gait recognition method unrelated with visual angle based on the determination theories of learning
CN108154159B (en) A kind of method for tracking target with automatic recovery ability based on Multistage Detector
CN107133569A (en) The many granularity mask methods of monitor video based on extensive Multi-label learning
CN104835178A (en) Low SNR(Signal to Noise Ratio) motion small target tracking and identification method
CN113052876B (en) Video relay tracking method and system based on deep learning
CN110008913A (en) The pedestrian's recognition methods again merged based on Attitude estimation with viewpoint mechanism
CN110084165A (en) The intelligent recognition and method for early warning of anomalous event under the open scene of power domain based on edge calculations
CN110135237B (en) Gesture recognition method
CN103593679A (en) Visual human-hand tracking method based on online machine learning
CN106709938A (en) Multi-target tracking method based on improved TLD (tracking-learning-detected)
CN107067410A (en) A kind of manifold regularization correlation filtering method for tracking target based on augmented sample
CN110533661A (en) Adaptive real-time closed-loop detection method based on characteristics of image cascade
CN115311241A (en) Coal mine down-hole person detection method based on image fusion and feature enhancement

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