CN110197142A - Object identification method, device, medium and terminal device under faint light condition - Google Patents
Object identification method, device, medium and terminal device under faint light condition Download PDFInfo
- Publication number
- CN110197142A CN110197142A CN201910406833.0A CN201910406833A CN110197142A CN 110197142 A CN110197142 A CN 110197142A CN 201910406833 A CN201910406833 A CN 201910406833A CN 110197142 A CN110197142 A CN 110197142A
- Authority
- CN
- China
- Prior art keywords
- object identification
- faint light
- light condition
- data
- image
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the object identification methods under a kind of faint light condition, comprising: carries out real-time detection to actual scene by the low-light level night vision device being arranged on AR wearable device, obtains live image;The live image is transferred on the AR wearable device by data line, identification object is carried out by the object identification model of foundation;The invention also discloses the object identification devices under a kind of faint light condition;The present invention will be become suitable for the visible images of eye-observation by low-light level night vision device by the distant target of faint natural lighting, then algorithm identification is carried out in the image of return, form recognition result, the technical issues of identifying can not be realized to object under conditions of faint light by solving the prior art, be identified under conditions of faint light to object to realize.
Description
Technical field
The present invention relates under object recognition technique field more particularly to a kind of faint light condition object identification method,
Device, storage medium and terminal device.
Background technique
Visual pattern identification technology is applied in various fields, under normal circumstances, when identifying, is needed bright and clear
Under the conditions of, it can just be accurately finished object identification;If light is weaker or night, there is following manner: 1, auxiliary by structure light
It helps, completes object identification;2, enhancing processing is carried out to image by software algorithm, then identified.
But it is bad by software progress reinforcing effect, it is limited very much;And assisted in identifying by laser point cloud structure light etc.
It can be because the defect of artificial disturbance or scanner itself to generate three-dimensional data often with influence of noise identification accurately
Property.
Therefore, a kind of method that object identification can be realized under the conditions of faint light is needed at present.
Summary of the invention
The present invention provides the object identification methods under a kind of faint light condition, can not be faint to solve the prior art
Object realizes the technical issues of identification under conditions of light, thus will be by the distant place mesh of faint natural lighting by low-light level night vision device
Mark becomes suitable for the visible images of eye-observation, and algorithm identification is then carried out in the image of return, forms recognition result, into
And it realizes and object is identified under conditions of faint light.
In order to solve the above-mentioned technical problem, the embodiment of the invention provides the object identification sides under a kind of faint light condition
Method, comprising:
Real-time detection is carried out to actual scene by the low-light level night vision device being arranged on AR wearable device, obtains scene photo
Picture;
The live image is transferred on the AR wearable device by data line, passes through the object identification of foundation
Model carries out identification object.
Preferably, the object identification model of the foundation, comprising:
It obtains input picture and extracts the depth characteristic of the input picture;
Structured modeling is carried out to the object in the input picture based on random field structural model, obtains the object
Structuring expression;
Structuring expression based on the object solves ladder using gradient back-propagation algorithm learning structure parameter
Degree, and learnt and trained using stochastic gradient descent algorithm, obtain the object identification model.
Preferably, the depth characteristic for extracting the input picture specifically includes: utilizing convolutional neural networks
The convolutional layer and pond layer of model, extract the depth characteristic of the input picture.
Preferably, described that the object progress structuring in the input picture is built based on random field structural model
Mould obtains the structuring expression of the object, specifically includes:
Component convolution operation is carried out to the depth characteristic of the input picture, it is each to obtain object described in the input picture
The apparent expression of a component;
The operation of structure pondization is carried out to the apparent expression of the object all parts, determines the optimal of each component of the object
Position;
Based on the optimal location of each component of the object, random field structural model is made inferences using mean field algorithm,
Obtain the structuring expression of the object.
Preferably, before the object identification model of the foundation, further includes: from webpage obtain image and
The corresponding text data of image;By the way that the label of object text data corresponding with image is matched, filtering and object
The corresponding image of the unmatched text data of label, obtains data set;Object identification model is trained using the data set.
Preferably, it is described obtain data set after, further includes: by the data set by copying as trained number
According to collection and test data set;The data that the training data is concentrated are transferred to the object identification model and carry out repetition training,
Feature integration data structure is extracted, until reaching deconditioning after trained threshold value and training accuracy;By the test data set
Interior data are transferred in the object identification model and are tested repeatedly, optimize the object identification model, until reaching survey
Stop test after trying threshold value and test accuracy.
Preferably, the trained threshold value is 200,000 times, and the training accuracy is 90%;The test threshold
It is 200,000 times, the test accuracy is 90%.
The embodiment of the invention also provides the object identification devices under a kind of faint light condition, comprising:
Acquisition module visits actual scene for the low-light level night vision device by being arranged on AR wearable device in real time
It surveys, obtains live image;
Transmission module, for the live image to be transferred to the AR wearable device by data line;
Object identification module carries out identification object to the live image for the object identification model by establishing.
The embodiment of the invention also provides a kind of computer readable storage medium, the computer readable storage medium includes
The computer program of storage;Wherein, where the computer program controls the computer readable storage medium at runtime
Equipment executes the object identification method under faint light condition as described in any one of the above embodiments.
The embodiment of the invention also provides a kind of terminal device, including processor, memory and it is stored in the storage
In device and it is configured as the computer program executed by the processor, the processor is real when executing the computer program
Object identification method under existing faint light condition as described in any one of the above embodiments.
Compared with the prior art, the embodiment of the present invention has the following beneficial effects:
The present invention by low-light level night vision device by by the distant target of faint natural lighting become suitable for eye-observation can
Then light-exposed image carries out algorithm identification in the image of return, form recognition result, and solving the prior art can not be in faint light
Under conditions of to object realize identify the technical issues of, thus realize object is identified under conditions of faint light.
Detailed description of the invention
Fig. 1: for the object identification method flow diagram under the faint light condition in the embodiment of the present invention.
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.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Please refer to Fig. 1, the preferred embodiment of the present invention provides the object identification method under a kind of faint light condition, packet
It includes:
Real-time detection is carried out to actual scene by the low-light level night vision device being arranged on AR wearable device, obtains scene photo
Picture;
The live image is transferred on the AR wearable device by data line, passes through the object identification of foundation
Model carries out identification object.
In the present embodiment, the object identification model of the foundation, comprising: obtain input picture and extract the input figure
The depth characteristic of picture;Structured modeling is carried out to the object in the input picture based on random field structural model, is obtained described
The structuring of object is expressed;Structuring expression based on the object, is joined using gradient back-propagation algorithm learning structure
Number solves gradient, and is learnt and trained using stochastic gradient descent algorithm, obtains the object identification model.
In the present embodiment, the depth characteristic for extracting the input picture specifically includes: utilizing convolutional neural networks
The convolutional layer and pond layer of model, extract the depth characteristic of the input picture.
In the present embodiment, described that the object progress structuring in the input picture is built based on random field structural model
Mould, obtains the structuring expression of the object, specifically includes: carrying out component convolution behaviour to the depth characteristic of the input picture
Make, obtains the apparent expression of object all parts described in the input picture;Apparent expression to the object all parts
The operation of structure pondization is carried out, determines the optimal location of each component of the object;Based on the optimal location of each component of the object, benefit
Random field structural model is made inferences with mean field algorithm, obtains the structuring expression of the object.
In the present embodiment, before the object identification model of the foundation, further includes: from webpage obtain image and
The corresponding text data of image;By the way that the label of object text data corresponding with image is matched, filtering and object
The corresponding image of the unmatched text data of label, obtains data set;Object identification model is trained using the data set.
In the present embodiment, it is described obtain data set after, further includes: by the data set by copying as trained number
According to collection and test data set;The data that the training data is concentrated are transferred to the object identification model and carry out repetition training,
Feature integration data structure is extracted, until reaching deconditioning after trained threshold value and training accuracy;By the test data set
Interior data are transferred in the object identification model and are tested repeatedly, optimize the object identification model, until reaching survey
Stop test after trying threshold value and test accuracy.
In the present embodiment, the trained threshold value is 200,000 times, and the training accuracy is 90%;The test threshold
It is 200,000 times, the test accuracy is 90%.
Correspondingly, the preferred embodiment of the present invention additionally provides the object identification device under a kind of faint light condition, comprising:
Acquisition module visits actual scene for the low-light level night vision device by being arranged on AR wearable device in real time
It surveys, obtains live image;
Transmission module, for the live image to be transferred to the AR wearable device by data line;
Object identification module carries out identification object to the live image for the object identification model by establishing.
The embodiment of the invention also provides a kind of computer readable storage medium, the computer readable storage medium includes
The computer program of storage;Wherein, where the computer program controls the computer readable storage medium at runtime
Equipment executes the charging online test method of UPS battery described in any of the above-described embodiment.
The embodiment of the invention also provides a kind of terminal device, the terminal device includes processor, memory and deposits
The computer program executed by the processor is stored up in the memory and is configured as, the processor is executing the meter
The charging online test method of UPS battery described in any of the above-described embodiment is realized when calculation machine program.
Preferably, the computer program can be divided into one or more module/units (such as computer program, meter
Calculation machine program), one or more of module/units are stored in the memory, and are executed by the processor, with
Complete the present invention.One or more of module/units can be and can complete the series of computation machine program of specific function and refer to
Section is enabled, the instruction segment is for describing implementation procedure of the computer program in the terminal device.
The processor can be central processing unit (Central Processing Unit, CPU), can also be it
His general processor, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit
(Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc., general processor can be microprocessor or the processor is also possible to any conventional processing
Device, the processor are the control centres of the terminal device, utilize each of terminal device described in various interfaces and connection
A part.
The memory mainly includes program storage area and data storage area, wherein program storage area can store operation dress
It sets, application program needed at least one function etc., data storage area can store related data etc..In addition, the memory can
To be high-speed random access memory, nonvolatile memory, such as plug-in type hard disk, intelligent memory card (Smart can also be
Media Card, SMC), secure digital (Secure Digital, SD) card and flash card (Flash Card) etc. or described deposit
Reservoir is also possible to other volatile solid-state parts.
It should be noted that above-mentioned terminal device may include, but it is not limited only to, processor, memory, those skilled in the art
Member is appreciated that above-mentioned terminal device is only example, does not constitute the restriction to terminal device, may include more or less
Component, perhaps combine certain components or different components.
The present invention by low-light level night vision device by by the distant target of faint natural lighting become suitable for eye-observation can
Then light-exposed image carries out algorithm identification in the image of return, form recognition result, and solving the prior art can not be in faint light
Under conditions of to object realize identify the technical issues of, thus realize object is identified under conditions of faint light.
Particular embodiments described above has carried out further the purpose of the present invention, technical scheme and beneficial effects
It is described in detail, it should be understood that the above is only a specific embodiment of the present invention, the protection being not intended to limit the present invention
Range.It particularly points out, to those skilled in the art, all within the spirits and principles of the present invention, that is done any repairs
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (10)
1. the object identification method under a kind of faint light condition characterized by comprising
Real-time detection is carried out to actual scene by the low-light level night vision device being arranged on AR wearable device, obtains live image;
The live image is transferred on the AR wearable device by data line, passes through the object identification model of foundation
Carry out identification object.
2. the object identification method under faint light condition as described in claim 1, which is characterized in that the object of the foundation
Identification model, comprising:
It obtains input picture and extracts the depth characteristic of the input picture;
Structured modeling is carried out to the object in the input picture based on random field structural model, obtains the structure of the object
Change expression;
Structuring expression based on the object solves gradient using gradient back-propagation algorithm learning structure parameter, and
Learnt using stochastic gradient descent algorithm and trained, obtains the object identification model.
3. the object identification method under faint light condition as claimed in claim 2, which is characterized in that the extraction is described defeated
The depth characteristic for entering image specifically includes: using the convolutional layer and pond layer of convolutional neural networks model, extracting the input figure
The depth characteristic of picture.
4. the object identification method under faint light condition as claimed in claim 2, which is characterized in that described to be based on random field
Structural model carries out structured modeling to the object in the input picture, obtains the structuring expression of the object, specific to wrap
It includes:
Component convolution operation is carried out to the depth characteristic of the input picture, obtains each portion of object described in the input picture
The apparent expression of part;
The operation of structure pondization is carried out to the apparent expression of the object all parts, determines the optimal position of each component of the object
It sets;
Based on the optimal location of each component of the object, random field structural model is made inferences using mean field algorithm, is obtained
The structuring of the object is expressed.
5. the object identification method under faint light condition as claimed in claim 2, which is characterized in that in the object of the foundation
Before body identification model, further includes: obtain image and the corresponding text data of image from webpage;By by the label of object
Text data corresponding with image is matched, and is filtered image corresponding with the unmatched text data of the label of object, is obtained
Data set;Object identification model is trained using the data set.
6. the object identification method under faint light condition as claimed in claim 5, which is characterized in that obtain data described
After collection, further includes: by the data set by copying as training dataset and test data set;The training data is concentrated
Data be transferred to the object identification model and carry out repetition training, feature integration data structure is extracted, until reaching trained threshold
Deconditioning after value and training accuracy;Data in the test data set are transferred in the object identification model and are carried out
It tests repeatedly, optimizes the object identification model, until stopping test after reaching test threshold and test accuracy.
7. the object identification method under faint light condition as claimed in claim 6, which is characterized in that the trained threshold value is
200000 times, the training accuracy is 90%;The test threshold is 200,000 times, and the test accuracy is 90%.
8. the object identification device under a kind of faint light condition characterized by comprising
Acquisition module carries out real-time detection to actual scene for the low-light level night vision device by being arranged on AR wearable device, obtains
To live image;
Transmission module, for the live image to be transferred to the AR wearable device by data line;
Object identification module carries out identification object to the live image for the object identification model by establishing.
9. a kind of computer readable storage medium, which is characterized in that the computer readable storage medium includes the calculating of storage
Machine program;Wherein, the equipment where the computer program controls the computer readable storage medium at runtime executes such as
Object identification method under the described in any item faint light conditions of claim 1~7.
10. a kind of terminal device, which is characterized in that including processor, memory and store in the memory and matched
It is set to the computer program executed by the processor, the processor is realized when executing the computer program as right is wanted
Seek the object identification method under 1~7 described in any item faint light conditions.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910406833.0A CN110197142A (en) | 2019-05-16 | 2019-05-16 | Object identification method, device, medium and terminal device under faint light condition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910406833.0A CN110197142A (en) | 2019-05-16 | 2019-05-16 | Object identification method, device, medium and terminal device under faint light condition |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110197142A true CN110197142A (en) | 2019-09-03 |
Family
ID=67752781
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910406833.0A Pending CN110197142A (en) | 2019-05-16 | 2019-05-16 | Object identification method, device, medium and terminal device under faint light condition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110197142A (en) |
Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1638039A1 (en) * | 2004-09-21 | 2006-03-22 | Valeo Vision | Procedure and device for night vision on the road allowing increased readability of images |
US7164117B2 (en) * | 1992-05-05 | 2007-01-16 | Automotive Technologies International, Inc. | Vehicular restraint system control system and method using multiple optical imagers |
CN103946732A (en) * | 2011-09-26 | 2014-07-23 | 微软公司 | Video display modification based on sensor input for a see-through near-to-eye display |
CN106570522A (en) * | 2016-10-24 | 2017-04-19 | 中国科学院自动化研究所 | Object recognition model establishment method and object recognition method |
CN206311847U (en) * | 2016-12-23 | 2017-07-07 | 王国清 | A kind of Night vision helmet of use AR technologies |
CN207067540U (en) * | 2017-10-27 | 2018-03-02 | 广东军丰特种装备科技发展有限公司 | Night vision device simulation trainer |
CN108647668A (en) * | 2018-05-21 | 2018-10-12 | 北京亮亮视野科技有限公司 | The construction method of multiple dimensioned lightweight Face datection model and the method for detecting human face based on the model |
CN109213862A (en) * | 2018-08-21 | 2019-01-15 | 北京京东尚科信息技术有限公司 | Object identification method and device, computer readable storage medium |
CN109309757A (en) * | 2018-08-24 | 2019-02-05 | 百度在线网络技术(北京)有限公司 | Memorandum based reminding method and terminal |
CN109376159A (en) * | 2018-12-05 | 2019-02-22 | 广州中浩控制技术有限公司 | A kind of batch record automatic generation method |
CN109413631A (en) * | 2018-12-13 | 2019-03-01 | 谷东科技有限公司 | A kind of AR glasses system and its implementation based on wireless blue tooth technology |
CN109446182A (en) * | 2018-11-05 | 2019-03-08 | 广州中浩控制技术有限公司 | A kind of compound acquisition method |
CN109492209A (en) * | 2018-10-25 | 2019-03-19 | 广州中浩控制技术有限公司 | A kind of table signature automatic generation method |
CN109508953A (en) * | 2018-10-25 | 2019-03-22 | 广州中浩控制技术有限公司 | A kind of form template generation method |
CN109522536A (en) * | 2018-11-05 | 2019-03-26 | 广州中浩控制技术有限公司 | A kind of table Auto-writing method |
US20190098227A1 (en) * | 2017-09-26 | 2019-03-28 | Samsung Electronics Co., Ltd. | Apparatus and method for displaying ar object |
-
2019
- 2019-05-16 CN CN201910406833.0A patent/CN110197142A/en active Pending
Patent Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7164117B2 (en) * | 1992-05-05 | 2007-01-16 | Automotive Technologies International, Inc. | Vehicular restraint system control system and method using multiple optical imagers |
EP1638039A1 (en) * | 2004-09-21 | 2006-03-22 | Valeo Vision | Procedure and device for night vision on the road allowing increased readability of images |
CN103946732A (en) * | 2011-09-26 | 2014-07-23 | 微软公司 | Video display modification based on sensor input for a see-through near-to-eye display |
CN106570522A (en) * | 2016-10-24 | 2017-04-19 | 中国科学院自动化研究所 | Object recognition model establishment method and object recognition method |
CN206311847U (en) * | 2016-12-23 | 2017-07-07 | 王国清 | A kind of Night vision helmet of use AR technologies |
US20190098227A1 (en) * | 2017-09-26 | 2019-03-28 | Samsung Electronics Co., Ltd. | Apparatus and method for displaying ar object |
CN207067540U (en) * | 2017-10-27 | 2018-03-02 | 广东军丰特种装备科技发展有限公司 | Night vision device simulation trainer |
CN108647668A (en) * | 2018-05-21 | 2018-10-12 | 北京亮亮视野科技有限公司 | The construction method of multiple dimensioned lightweight Face datection model and the method for detecting human face based on the model |
CN109213862A (en) * | 2018-08-21 | 2019-01-15 | 北京京东尚科信息技术有限公司 | Object identification method and device, computer readable storage medium |
CN109309757A (en) * | 2018-08-24 | 2019-02-05 | 百度在线网络技术(北京)有限公司 | Memorandum based reminding method and terminal |
CN109492209A (en) * | 2018-10-25 | 2019-03-19 | 广州中浩控制技术有限公司 | A kind of table signature automatic generation method |
CN109508953A (en) * | 2018-10-25 | 2019-03-22 | 广州中浩控制技术有限公司 | A kind of form template generation method |
CN109446182A (en) * | 2018-11-05 | 2019-03-08 | 广州中浩控制技术有限公司 | A kind of compound acquisition method |
CN109522536A (en) * | 2018-11-05 | 2019-03-26 | 广州中浩控制技术有限公司 | A kind of table Auto-writing method |
CN109376159A (en) * | 2018-12-05 | 2019-02-22 | 广州中浩控制技术有限公司 | A kind of batch record automatic generation method |
CN109413631A (en) * | 2018-12-13 | 2019-03-01 | 谷东科技有限公司 | A kind of AR glasses system and its implementation based on wireless blue tooth technology |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6926335B2 (en) | Variable rotation object detection in deep learning | |
CN107392218A (en) | A kind of car damage identification method based on image, device and electronic equipment | |
CN109658418A (en) | Learning method, device and the electronic equipment of scene structure | |
CN107690660A (en) | Image-recognizing method and device | |
CN109583345A (en) | Roads recognition method, device, computer installation and computer readable storage medium | |
CN111310518B (en) | Picture feature extraction method, target re-identification method, device and electronic equipment | |
CN106408037A (en) | Image recognition method and apparatus | |
CN111444873A (en) | Method and device for detecting authenticity of person in video, electronic device and storage medium | |
CN113869429A (en) | Model training method and image processing method | |
CN103577875A (en) | CAD (computer-aided design) people counting method based on FAST (features from accelerated segment test) | |
CN106778910A (en) | Deep learning system and method based on local training | |
CN110705573A (en) | Automatic modeling method and device of target detection model | |
CN112802076A (en) | Reflection image generation model and training method of reflection removal model | |
CN111126264A (en) | Image processing method, device, equipment and storage medium | |
CN114596584A (en) | Intelligent detection and identification method for marine organisms | |
CN108764248B (en) | Image feature point extraction method and device | |
CN114168768A (en) | Image retrieval method and related equipment | |
CN111652242B (en) | Image processing method, device, electronic equipment and storage medium | |
CN116258756B (en) | Self-supervision monocular depth estimation method and system | |
CN110197142A (en) | Object identification method, device, medium and terminal device under faint light condition | |
CN112329736B (en) | Face recognition method and financial system | |
JP2024514175A (en) | Bird detection and species determination | |
CN105513050B (en) | A kind of target image extracting method and device | |
CN112766481B (en) | Training method and device for neural network model and image detection method | |
CN110490950B (en) | Image sample generation method and device, computer equipment and storage medium |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190903 |
|
RJ01 | Rejection of invention patent application after publication |