CN109409219A - Indoor occupant locating and tracking algorithm based on depth convolutional network - Google Patents
Indoor occupant locating and tracking algorithm based on depth convolutional network Download PDFInfo
- Publication number
- CN109409219A CN109409219A CN201811092330.2A CN201811092330A CN109409219A CN 109409219 A CN109409219 A CN 109409219A CN 201811092330 A CN201811092330 A CN 201811092330A CN 109409219 A CN109409219 A CN 109409219A
- Authority
- CN
- China
- Prior art keywords
- network
- depth convolutional
- indoor occupant
- tracking algorithm
- autocoder
- 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
Links
- 238000012549 training Methods 0.000 claims abstract description 33
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 238000012512 characterization method Methods 0.000 claims abstract description 9
- 230000008859 change Effects 0.000 claims abstract description 6
- 230000013016 learning Effects 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 8
- 230000001815 facial effect Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000000034 method Methods 0.000 abstract description 12
- 239000010410 layer Substances 0.000 description 26
- 230000006870 function Effects 0.000 description 7
- 230000009466 transformation Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000002356 single layer Substances 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002087 whitening effect Effects 0.000 description 1
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Multimedia (AREA)
- Molecular Biology (AREA)
- Human Computer Interaction (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of indoor occupant locating and tracking algorithm based on depth convolutional network, in carrying out indoor occupant locating and tracking algorithm, including being input in network in real time by shooting video image by camera, ZCA albefaction carries out the data prediction for reducing the correlation of feature, based on the convolution, Chi Hua, the sparse autocoder building deep layer network characterization extractor of multilayer carries out the average pond, weight parameter using training set network parameter and the fully-connected network classifies to test image, classified by the Softmax classifier, the feature being collected into is accurately identified by the face recognition module, finally export the information of recognition of face.It is efficiently located in the case where the change of multi-target jamming of the present invention or target appearance.It is compared with the traditional method, this algorithm can faster adapt to environment.
Description
Technical field
The present invention relates to deep learnings, machine learning techniques field, more particularly to the indoor people based on depth convolutional network
Member's locating and tracking algorithm.
Background technique
With the development of science and technology, people are gradually increasing the Demand and service of positioning.Especially in complicated room
In interior environment, such as building, airport hall, mine and other environment, it is generally necessary to mobile worker be accurately positioned in room
Interior accurate location simultaneously realizes personal scheduling management, personal safety, Emergency Assistance.Most of location algorithms have the following problems:
Positioning accuracy is poor.It is inadequate all to there is indoor position accuracy in most of positioning systems, generally can only by RSSI location model
The precision for enough reaching 3 meters also can only achieve 1.5 meters, the requirement of indoor accurate position be much not achieved even with trigger;
Coverage area is small, during radio propagation due to the complexity of the decaying of signal and communication environments, all positioning moulds
Type all must satisfy certain orientation range, and more than after orientation range, model error be will increase, and precision is seriously affected;
Anti-interference ability is weak.Due to the relatively narrow of indoor environment and closing, wireless signal be cannot achieve in sighting distance in many cases
It directly transmits, wall, baffle, floor are very big to the propagation effect of wireless channel, and the noise of other electrical equipments also will affect nothing
The propagation of line signal, here it is so-called non-market values.
Summary of the invention
The present invention is directed to solve at least some of the technical problems in related technologies.For this purpose, of the invention
One purpose is to propose the indoor occupant locating and tracking algorithm based on depth convolutional network, multi-target jamming or target appearance
It is efficiently located in the case where change.It is compared with the traditional method, this algorithm can faster adapt to environment.
A kind of indoor occupant locating and tracking algorithm based on depth convolutional network according to an embodiment of the present invention is carrying out room
In interior personnel positioning track algorithm, including pass through shooting video image and be input in network in real time by camera, ZCA albefaction carries out
The data prediction is used to reduce the correlation of feature, deep based on the convolution, Chi Hua, the sparse autocoder building of multilayer
Layer network feature extractor carries out the average pond, utilizes the weight parameter of training set network parameter and the fully-connected network
Classify to test image, classified by the Softmax classifier, the feature being collected into is by the recognition of face mould
Block accurately identifies, and finally exports the information of recognition of face.
In some embodiments of the invention, indoor occupant locating and tracking algorithm, is specifically divided into following step:
S1: the monitor video under the conditions of in collection room generates training data and test data, initializes depth convolutional network
Structure;
S2: ZCA albefaction and normalization are introduced, raw video image is pre-processed;
S3: special come the deep layer for extracting video image using the sparse autocoder building depth network characterization extractor of multilayer
Sign;
S4: introducing Softmax regression model, and the feature vector obtained to deep layer network characterization extractor is classified;
S5: face recognition module facial characteristics for identification is introduced, the information of recognition of face is exported.
In other embodiments of the invention, depth convolutional network structure is initialized in the S1, including sparse
Quantity, the range of decrease of pond layer number and pond layer of autocoder.
ZCA albefaction is introduced in other embodiments of the invention, in the S2 and is normalized to Pretreatment Test data,
The normalized uses dimension normalization, gray scale normalization and histogram equalization.
In other embodiments of the invention, the sparse autocoder of multilayer includes the sparse automatic volume of noise reduction in the S3
Code device and sparse autocoder, described be the sparse autocoder of noise reduction is first layer autocoder, artificially by noise
Be added in training data, depth e-learning removes noise to obtain the not input by noise jamming, it is described it is sparse from
Dynamic encoder is second layer autocoder, and the sparse autocoder is the autocoder for not adding noise.
In other embodiments of the invention, Softmax regression model is obtained by supervised learning in the S4, described
It is stringent mutually exclusive between class categories needed for supervised learning requirement, to avoid falling into the situation of locally optimal solution, to obtain
Obtain globally optimal solution.
Face recognition module facial characteristics for identification, tool are introduced in other embodiments of the invention, in the S5
Body step includes:
S5.1: pre-processing target image so that image meets training requirement network;
S5.2: obtaining suitable data by carrying out stochastical sampling to the image in S5.1, by sparse self-encoding encoder without
Supervise the weight that pre-training obtains CNN initialization filter;
S5.3: it is obtained by convolution scheduled more between training set image in the filter and S5.1 obtained in S5.2
A characteristic pattern;
S5.4: Generalized image is obtained by maximizing the signature obtained in S5.3;
S5.5: by secondary convolution, the characteristic pattern exported in double sampling S5.6 obtains required characteristic pattern;
S5.5: all characteristic patterns in S5.5 are all converted into an individual column vector, are input into complete connection
Layer calculates the difference between recognition result and label, and network parameter is adjusted and more from top to bottom by back-propagation algorithm
Newly;
S5.7: the weight parameter using training set network parameter and fully-connected network classifies to test image, and leads to
Cross the recognition result that Softmax classifier obtains image.
Beneficial effect in the present invention is: the indoor occupant positioning that the invention proposes a kind of based on depth convolutional network with
Track algorithm reduces network training complexity, convolution sum Chi Huayong firstly, ZCA whitening pretreatment is used to reduce the correlation of feature
In the feature of extraction interesting target, and its location information is extracted, then, inhibits accurate to the target of human body according to non-maximum value
It outlines, the external appearance characteristic of moving target is extracted by using deep layer network characterization, tracked and calculated based on deep-controlled indoor positioning
Method can be efficiently located in the case where target is blocked the change of multi-target jamming or target appearance by background, with tradition side
Method is compared, this algorithm can faster adapt to environment.
Detailed description of the invention
Attached drawing is used to provide further understanding of the present invention, and constitutes part of specification, with reality of the invention
It applies example to be used to explain the present invention together, not be construed as limiting the invention.In the accompanying drawings:
Fig. 1 is the flow chart of the indoor occupant locating and tracking algorithm the present invention is based on depth convolutional network;
Fig. 2 is the stream of depth characteristic extractor in the indoor occupant locating and tracking algorithm the present invention is based on depth convolutional network
Cheng Tu;
Fig. 3 is the sparse automatic volume of single layer noise reduction in the indoor occupant locating and tracking algorithm the present invention is based on depth convolutional network
The training block diagram of code device;
Fig. 4 is the sparse automatic volume of double-layered noise reduction in the indoor occupant locating and tracking algorithm the present invention is based on depth convolutional network
The training block diagram of code device;
Fig. 5 is the training frame for having supervision to finely tune in the indoor occupant locating and tracking algorithm the present invention is based on depth convolutional network
Figure;
Fig. 6 is the process of face recognition algorithms in the indoor occupant locating and tracking algorithm the present invention is based on depth convolutional network
Figure.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.
Examples of the embodiments are shown in the accompanying drawings, and in which the same or similar labels are throughly indicated identical or classes
As element or element with the same or similar functions.The embodiments described below with reference to the accompanying drawings are exemplary, purport
It is being used to explain the present invention, and is being not considered as limiting the invention.
Embodiment 1:
As shown in figures 1 to 6, a kind of indoor occupant locating and tracking algorithm based on depth convolutional network is carrying out indoor occupant
In locating and tracking algorithm, including pass through shooting video image and be input in network in real time by camera, it is pre- that ZCA albefaction carries out data
The correlation for reducing feature is handled, is extracted based on convolution, Chi Hua, the sparse autocoder building deep layer network characterization of multilayer
Device carries out average pond, and the weight parameter using training set network parameter and fully-connected network classifies to test image, leads to
It crosses Softmax classifier to classify, the feature being collected into is accurately identified by face recognition module, finally exports recognition of face
Information.
Indoor occupant locating and tracking algorithm, is specifically divided into following step:
S1: the monitor video under the conditions of in collection room generates training data and test data, initializes depth convolutional network
Structure;
S2: ZCA albefaction and normalization are introduced, raw video image is pre-processed;
S3: special come the deep layer for extracting video image using the sparse autocoder building depth network characterization extractor of multilayer
Sign;
S4: introducing Softmax regression model, and the feature vector obtained to deep layer network characterization extractor is classified;
S5: face recognition module facial characteristics for identification is introduced, the information of recognition of face is exported.
In S1 initialize depth convolutional network structure, including the quantity of sparse autocoder, pond layer number and
The range of decrease of pond layer.
Implement 2:
As shown in figures 1 to 6, according to embodiment 1 the step of, further, in S2 introduces ZCA albefaction and is normalized to pre-
Experimental data is handled, normalized uses dimension normalization, gray scale normalization and histogram equalization, reduces overall network
Computation complexity, while garbage is reduced on the basis of retaining original information.Firstly, being gone by PCA transformation special unless each
Then correlation between sign makes to export feature with unit variance, then data rotation is gone back, obtain the processing of ZCA albefaction
As a result, reducing the redundancy of input, further, S2 further includes:
S2.1: it firstly, going the correlation between feature unless each by PCA transformation, converts as follows:
xrot,i=UTxi;
S2.2: make to export feature with unit variance, convert as follows:
S2.3: again going back data rotation, obtains the processing result of ZCA albefaction, reduces the redundancy of input, transformation is such as
Under:
xZCAWhite=UxPCAWhite。
Embodiment 3:
As shown in figures 1 to 6, according to embodiment 1,2 the step of, further, autocoder can be with table for capturing in S3
Show the primary clustering of input information, therefore it can reappear input as far as possible.The sparse autocoder of noise reduction is that first layer is automatic
Noise is artificially added in training data by encoder.It is dry not by noise to obtain that depth e-learning removes noise
The input disturbed, so the expression that there is stronger generalization ability to go study input signal for it.Second layer autocoder is not using
Add the sparse autocoder of noise.
Further, step 3 further comprises:
S3.1: given without label data, with unsupervised learning method learning characteristic, sample is preprocessed, convolution sum pond
The sparse autocoder of single layer noise reduction is inputted afterwards, and input X obtains coding h, convert as follows through unsupervised learning;
H=WTX;
Then it is rebuild and is inputted by decoder, calculate the reconstructed error between input and reconstructed results.Construct cost function L
(X, W) is converted as follows;
Using L-BFGS algorithm, the parameter of encoder and decoder is adjusted, makes cost function L (X;W) minimum, first layer
Parameter network is saved during this time.
S3.2: the next layer of training realizes that successively training starts the training of layer 2 network after the completion of the training of the 1st layer network,
By the coding 1 of the 1st layer network output as the 2nd layer of input signal, S3.1 is repeated, obtains the 2nd layer of expression of former input information
Coding 2, and save second layer network parameter at this time.
S3.3: being finely adjusted using the training set tested in selected java standard library, and coding 2 is inputted Softmax regression model,
It goes to train by the supervised training method of the multilayer neural network of standard, using end-to-end learning method, by there is label
Sample finely tunes whole system, and after completing aforesaid operations, obtained network can be used for extracting further feature.
Embodiment 4:
As shown in figures 1 to 6, according to embodiment 1-3 the step of, further, Softmax regression model is learned by supervision in S4
Acquistion is arrived.It is stringent mutually exclusive between class categories needed for its requirement, that is, two categories cannot be simultaneously by a sample
Occupy, weight attenuation term is added in cost function, this can punish excessive parameter value, and cost function is made to become stringent
Convex function, guarantee is uniquely solved, to avoid falling into the situation of locally optimal solution, to obtain globally optimal solution, Softmax
The cost function of regression model is as follows:
Embodiment 5:
As shown in figures 1 to 6, according to embodiment 1-4 the step of, further introduces face recognition module for identification in S5
Facial characteristics, specific steps include:
S5.1: pre-processing target image so that image meets training requirement network;
S5.2: obtaining suitable data by carrying out stochastical sampling to the image in S5.1, by sparse self-encoding encoder without
Supervise the weight that pre-training obtains CNN initialization filter;
S5.3: it is obtained by convolution scheduled more between training set image in the filter and S5.1 obtained in S5.2
A characteristic pattern;
S5.4: Generalized image is obtained by maximizing the signature obtained in S5.3;
S5.5: by secondary convolution, the characteristic pattern exported in double sampling S5.6 obtains required characteristic pattern;
S5.5: all characteristic patterns in S5.5 are all converted into an individual column vector, are input into complete connection
Layer calculates the difference between recognition result and label, and network parameter is adjusted and more from top to bottom by back-propagation algorithm
Newly;
S5.7: the weight parameter using training set network parameter and fully-connected network classifies to test image, and leads to
Cross the recognition result that Softmax classifier obtains image.
The invention proposes a kind of indoor occupant locating and tracking algorithm based on depth convolutional network, firstly, ZCA albefaction is pre-
The correlation for reducing feature is handled, reduces network training complexity, convolution sum pond is for extracting the spy of interesting target
Sign, and its location information is extracted, then, the target of human body is accurately outlined according to the inhibition of non-maximum value, by using deep layer net
The external appearance characteristic of network feature extraction moving target can be hidden in target by background based on deep-controlled indoor positioning track algorithm
Gear, efficiently locates in the case where the change of multi-target jamming or target appearance, is compared with the traditional method, this algorithm can be faster
Adapt to environment.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
Centainly refer to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be any
One or more embodiment or examples in can be combined in any suitable manner.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
Anyone skilled in the art in the technical scope disclosed by the present invention, according to the technique and scheme of the present invention and its
Inventive concept is subject to equivalent substitution or change, should be covered by the protection scope of the present invention.
Claims (7)
1. a kind of indoor occupant locating and tracking algorithm based on depth convolutional network, it is characterised in that: fixed carrying out indoor occupant
In the track algorithm of position, including pass through shooting video image and be input in network in real time by camera, ZCA albefaction carries out the data
The correlation for reducing feature is pre-processed, it is special based on the convolution, Chi Hua, the sparse autocoder building deep layer network of multilayer
It levies extractor and carries out the average pond, using the weight parameter of training set network parameter and the fully-connected network to test chart
As classifying, classified by the Softmax classifier, the feature being collected into accurately is known by the face recognition module
Not, the information of recognition of face is finally exported.
2. the indoor occupant locating and tracking algorithm according to claim 1 based on depth convolutional network, which is characterized in that room
Interior personnel positioning track algorithm, is specifically divided into following step:
S1: the monitor video under the conditions of in collection room generates training data and test data, initializes depth convolutional network knot
Structure;
S2: ZCA albefaction and normalization are introduced, raw video image is pre-processed;
S3: the further feature of video image is extracted using the sparse autocoder building depth network characterization extractor of multilayer;
S4: introducing Softmax regression model, and the feature vector obtained to deep layer network characterization extractor is classified;
S5: face recognition module facial characteristics for identification is introduced, the information of recognition of face is exported.
3. the indoor occupant locating and tracking algorithm according to claim 2 based on depth convolutional network, it is characterised in that: institute
Initialization depth convolutional network structure in S1 is stated, including the quantity, pond layer number and pond layer of sparse autocoder
The range of decrease.
4. the indoor occupant locating and tracking algorithm according to claim 2 based on depth convolutional network, it is characterised in that: institute
It states and introduces ZCA albefaction in S2 and be normalized to Pretreatment Test data, the normalized is returned using dimension normalization, gray scale
One change and histogram equalization.
5. the indoor occupant locating and tracking algorithm according to claim 2 based on depth convolutional network, it is characterised in that: institute
Stating the sparse autocoder of multilayer in S3 includes the sparse autocoder of noise reduction and sparse autocoder, and described is that noise reduction is sparse
Autocoder is first layer autocoder, and noise is artificially added in training data, and the removal of depth e-learning is made an uproar
For sound to obtain the not input by noise jamming, the sparse autocoder is second layer autocoder, described sparse
Autocoder is the autocoder for not adding noise.
6. the indoor occupant locating and tracking algorithm according to claim 2 based on depth convolutional network, it is characterised in that: institute
It states Softmax regression model in S4 to be obtained by supervised learning, the supervised learning requires stringent between required class categories
It is mutually exclusive, to avoid falling into the situation of locally optimal solution, to obtain globally optimal solution.
7. the indoor occupant locating and tracking algorithm according to claim 2 based on depth convolutional network, it is characterised in that: institute
Stating introducing face recognition module in S5, facial characteristics, specific steps include: for identification
S5.1: pre-processing target image so that image meets training requirement network;
S5.2: obtaining suitable data by carrying out stochastical sampling to the image in S5.1, unsupervised by sparse self-encoding encoder
Pre-training obtains the weight of CNN initialization filter;
S5.3: scheduled multiple spies in the filter and S5.1 obtained in S5.2 between training set image are obtained by convolution
Sign figure;
S5.4: Generalized image is obtained by maximizing the signature obtained in S5.3;
S5.5: by secondary convolution, the characteristic pattern exported in double sampling S5.6 obtains required characteristic pattern;
S5.5: being converted into an individual column vector for all characteristic patterns in S5.5, is input into complete articulamentum
The difference between recognition result and label is calculated, network parameter is adjusted and is updated from top to bottom by back-propagation algorithm;
S5.7: the weight parameter using training set network parameter and fully-connected network classifies to test image, and passes through
The recognition result of Softmax classifier acquisition image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811092330.2A CN109409219A (en) | 2018-09-19 | 2018-09-19 | Indoor occupant locating and tracking algorithm based on depth convolutional network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811092330.2A CN109409219A (en) | 2018-09-19 | 2018-09-19 | Indoor occupant locating and tracking algorithm based on depth convolutional network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109409219A true CN109409219A (en) | 2019-03-01 |
Family
ID=65465106
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811092330.2A Pending CN109409219A (en) | 2018-09-19 | 2018-09-19 | Indoor occupant locating and tracking algorithm based on depth convolutional network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109409219A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109905675A (en) * | 2019-03-13 | 2019-06-18 | 武汉大学 | A kind of mine personnel monitoring system based on computer vision and method |
CN110245603A (en) * | 2019-06-12 | 2019-09-17 | 成都信息工程大学 | A kind of group abnormality behavior real-time detection method |
CN110378292A (en) * | 2019-07-22 | 2019-10-25 | 广州络维建筑信息技术咨询有限公司 | Three dimension location system and method |
CN110443177A (en) * | 2019-07-29 | 2019-11-12 | 上海工程技术大学 | A kind of airport indoor locating system based on recognition of face |
CN110705621A (en) * | 2019-09-25 | 2020-01-17 | 北京影谱科技股份有限公司 | Food image identification method and system based on DCNN and food calorie calculation method |
CN111626154A (en) * | 2020-05-14 | 2020-09-04 | 闽江学院 | Face tracking method based on convolution variational encoder |
CN111753898A (en) * | 2020-06-23 | 2020-10-09 | 扬州大学 | Representation learning method based on superposition convolution sparse self-encoding machine |
CN111783889A (en) * | 2020-07-03 | 2020-10-16 | 北京字节跳动网络技术有限公司 | Image recognition method and device, electronic equipment and computer readable medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106778563A (en) * | 2016-12-02 | 2017-05-31 | 江苏大学 | A kind of quick any attitude facial expression recognizing method based on the coherent feature in space |
CN107506722A (en) * | 2017-08-18 | 2017-12-22 | 中国地质大学(武汉) | One kind is based on depth sparse convolution neutral net face emotion identification method |
-
2018
- 2018-09-19 CN CN201811092330.2A patent/CN109409219A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106778563A (en) * | 2016-12-02 | 2017-05-31 | 江苏大学 | A kind of quick any attitude facial expression recognizing method based on the coherent feature in space |
CN107506722A (en) * | 2017-08-18 | 2017-12-22 | 中国地质大学(武汉) | One kind is based on depth sparse convolution neutral net face emotion identification method |
Non-Patent Citations (1)
Title |
---|
HANHUI YUE 等: "Research and Implementation of Indoor Positioning Algorithm for Personnel Based on Deep Learning" * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109905675A (en) * | 2019-03-13 | 2019-06-18 | 武汉大学 | A kind of mine personnel monitoring system based on computer vision and method |
CN110245603A (en) * | 2019-06-12 | 2019-09-17 | 成都信息工程大学 | A kind of group abnormality behavior real-time detection method |
CN110378292A (en) * | 2019-07-22 | 2019-10-25 | 广州络维建筑信息技术咨询有限公司 | Three dimension location system and method |
CN110443177A (en) * | 2019-07-29 | 2019-11-12 | 上海工程技术大学 | A kind of airport indoor locating system based on recognition of face |
CN110705621A (en) * | 2019-09-25 | 2020-01-17 | 北京影谱科技股份有限公司 | Food image identification method and system based on DCNN and food calorie calculation method |
CN111626154A (en) * | 2020-05-14 | 2020-09-04 | 闽江学院 | Face tracking method based on convolution variational encoder |
CN111626154B (en) * | 2020-05-14 | 2023-04-07 | 闽江学院 | Face tracking method based on convolution variational encoder |
CN111753898A (en) * | 2020-06-23 | 2020-10-09 | 扬州大学 | Representation learning method based on superposition convolution sparse self-encoding machine |
CN111753898B (en) * | 2020-06-23 | 2023-09-22 | 扬州大学 | Representation learning method based on superposition convolution sparse self-encoder |
CN111783889A (en) * | 2020-07-03 | 2020-10-16 | 北京字节跳动网络技术有限公司 | Image recognition method and device, electronic equipment and computer readable medium |
CN111783889B (en) * | 2020-07-03 | 2022-03-01 | 北京字节跳动网络技术有限公司 | Image recognition method and device, electronic equipment and computer readable medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109409219A (en) | Indoor occupant locating and tracking algorithm based on depth convolutional network | |
CN111860282B (en) | Subway section passenger flow volume statistics and pedestrian retrograde motion detection method and system | |
Wang et al. | Fall detection based on dual-channel feature integration | |
CN113076809B (en) | Visual transducer-based high-altitude object detection method | |
CN109522793A (en) | More people's unusual checkings and recognition methods based on machine vision | |
CN110119703A (en) | The human motion recognition method of attention mechanism and space-time diagram convolutional neural networks is merged under a kind of security protection scene | |
CN110070074A (en) | A method of building pedestrian detection model | |
CN110490174A (en) | Multiple dimensioned pedestrian detection method based on Fusion Features | |
CN110135476A (en) | A kind of detection method of personal safety equipment, device, equipment and system | |
CN106778854A (en) | Activity recognition method based on track and convolutional neural networks feature extraction | |
CN108564049A (en) | A kind of fast face detection recognition method based on deep learning | |
CN107092894A (en) | A kind of motor behavior recognition methods based on LSTM models | |
CN110110649A (en) | Alternative method for detecting human face based on directional velocity | |
CN106907697A (en) | A kind of intelligent road-lamp of built-in collision sound detection function | |
CN109858407A (en) | A kind of video behavior recognition methods based on much information stream feature and asynchronous fusion | |
Chen et al. | Attention-based dual-stream vision transformer for radar gait recognition | |
CN113516005A (en) | Dance action evaluation system based on deep learning and attitude estimation | |
CN115527271A (en) | Elevator car passenger abnormal behavior detection system and method | |
CN109740418B (en) | Yoga action identification method based on multiple acceleration sensors | |
CN105160285A (en) | Method and system for recognizing human body tumble automatically based on stereoscopic vision | |
Luo et al. | A system for the detection of polyphonic sound on a university campus based on CapsNet-RNN | |
US20240176933A1 (en) | Digital twin based method for monitoring behavior of passenger on escalator | |
CN116758479B (en) | Coding deep learning-based intelligent agent activity recognition method and system | |
CN111291804A (en) | Multi-sensor time series analysis model based on attention mechanism | |
CN112307894A (en) | Pedestrian age identification method based on wrinkle features and posture features in community monitoring scene |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190301 |
|
WD01 | Invention patent application deemed withdrawn after publication |