CN107341538A - A kind of statistical magnitude method of view-based access control model - Google Patents
A kind of statistical magnitude method of view-based access control model Download PDFInfo
- Publication number
- CN107341538A CN107341538A CN201710525089.7A CN201710525089A CN107341538A CN 107341538 A CN107341538 A CN 107341538A CN 201710525089 A CN201710525089 A CN 201710525089A CN 107341538 A CN107341538 A CN 107341538A
- Authority
- CN
- China
- Prior art keywords
- msub
- target object
- neutral net
- target
- mrow
- 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
- G06M—COUNTING MECHANISMS; COUNTING OF OBJECTS NOT OTHERWISE PROVIDED FOR
- G06M1/00—Design features of general application
- G06M1/27—Design features of general application for representing the result of count in the form of electric signals, e.g. by sensing markings on the counter drum
- G06M1/272—Design features of general application for representing the result of count in the form of electric signals, e.g. by sensing markings on the counter drum using photoelectric means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K17/00—Methods or arrangements for effecting co-operative working between equipments covered by two or more of main groups G06K1/00 - G06K15/00, e.g. automatic card files incorporating conveying and reading operations
- G06K17/0022—Methods or arrangements for effecting co-operative working between equipments covered by two or more of main groups G06K1/00 - G06K15/00, e.g. automatic card files incorporating conveying and reading operations arrangements or provisious for transferring data to distant stations, e.g. from a sensing device
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of statistical magnitude method of view-based access control model, comprise the following steps：1) depth neuroid training is carried out on target object, builds neutral net target detection model；2) target object in label detection region, number of lines of going forward side by side statistics are detected by RFID system；3) triggered after RFID system detects the RFID information of target object and obtain the image of the target object in label detection region, and will be identified in the neutral net target detection model that trains of image input and statistics numbers；4) statistics numbers of RFID system and neutral net target detection model, the final number for obtaining target object in label detection region are combined.Compared with prior art, the present invention has the advantages that to count accurate, self-studying mode, embedded system.
Description
Technical field
The present invention relates to logistics image recognition to count field, more particularly, to a kind of statistical magnitude method of view-based access control model.
Background technology
Extended Technology of the image recognition technology as machine vision, has been obtained for developing rapidly in the nearest more than ten years,
And powerful application value is shown in every field, it is widely used in industry, product introduction, medical research, military affairs, education etc.
Field, turn into a study hotspot of technical field of machine vision.And image recognition technology be machine vision key technology it
One, it causes a machine to that taken image is learnt and identified.Image identification system is made using the method for deep learning
Machine is learnt for specific object and the data outside sample are identified.When a certain amount of specific to computer input
After the view data of object, learnt by the neutral net preset, so as to which computer acquisition is certain for specific image
Recognition capability.In addition, image identification system allows for detecting that certain objects enter specific region in real time, then basis
The good model of training in advance and data are identified, and identify in the image of output the position of certain objects in the picture and
Number, here it is the task that image identification system to be completed.
Image recognition technology is roughly divided into three kinds both at home and abroad at present：Based on the front and rear identification side different according to frame number contrast images
Method, the recognition methods based on default article size contour shape and the artificial intelligence recognition methods based on machine deep learning.
Recognition methods based on default article size contour shape is image method more conventional at present, but can not overcome due to
Feature Fuzzy, change the real profile None- identifieds caused by reason such as lens location, hot spot and then the problem of identification is wrong occur.
Simultaneously early stage based on the front and rear recognition methods different according to frame number contrast images be mainly by means of front and rear frame in video flowing figure
As change carries out object identification, precision is low easily by external interference, poor universality and can not identify specific object.
Depth learning technology possess high universalizable, precision are high, support 30 degree to 150 degree between arbitrarily lens shooting angle,
10cm to 100cm shooting distances are supported, while the ginseng such as adjustable study subject object, object size, object color can also be provided
Number.Therefore deep learning module is integrated in the technology of machine vision, significantly facilitated after learning model is provided with addition
The renewal and management of identification object.
In addition, in scene of the non-contact RF ID technical products for identifying object, the identification of RFID read heads is RFID chip
In digital coding, rather than specific object.Under some specific scenes, identified digital coding can fail, so as to
Generate missing inspection.Image recognition technology is an effective means of supplementing out economy, by comparing quantity and the image that RFID read heads obtain
Identify the uniformity of the quantity obtained, it is ensured that zero defect requirement in logistic statistics.
The content of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide one kind statistics is accurate, learns by oneself
Habit pattern, embedded system view-based access control model statistical magnitude method.
The purpose of the present invention can be achieved through the following technical solutions：
A kind of statistical magnitude method of view-based access control model, to check target object in label detection region in self-help settlement
Number, comprise the following steps：
1) depth neuroid training is carried out on target object, builds neutral net target detection model；
2) target object in label detection region, number of lines of going forward side by side statistics are detected by RFID system；
3) triggered after RFID system detects the RFID information of target object and obtain the target in label detection region
The image of object, and will be identified in the neutral net target detection model that trains of image input and statistics numbers；
4) statistics numbers of RFID system and neutral net target detection model are combined, it is final to obtain in label detection region
The number of target object：
If the counting of RFID system is identical with the counting of neutral net target detection model, using the counting as final
Count；
If the counting of RFID system is different from the counting of neutral net target detection model, alarmed, request is carried out
Artificial counting.
Described step 1) specifically includes following steps：
11) data prepare and marked：Several images comprising target object are obtained as training sample, are located in advance by data
Reason and data enhancing conversion ensure sufficient training data, and enter rower to the target object in training sample by annotation tool
Note；
12) data are trained：The detection information of correct target object is obtained by annotation tool, neutral net is detected
Between classification, position and the size data of target object and the classification of correct target object, physical location and actual size
Error is as object function, by iteration so that object function completes training when reaching minimum value；
13) data model test and evaluation：It is iterated, accumulated and self study using test set, it is ensured that neutral net mesh
Mark the accuracy rate of detection model.
In described step 11), target object is included from several are obtained by way of network crawl and reality shooting
Image is as training sample.
In described step 12), described object function is：
Wherein, xi、yiThe position of i-th of target object is detected for neutral net,For i-th of correct object
The physical location of body, wi、hiThe size of i-th of target object is detected for neutral net,For i-th of correct object
The actual size of body, Ci、For the classification and concrete class of i-th of target object, α, β, λ are respectively empirical parameter.
In described step 13), the image of mark and training is had neither part nor lot in as test set by 20%.
Described step 1) is further comprising the steps of：
14) when neutral net target detection model produces a desired effect accuracy rate, by neutral net target detection model
Mobile GPU platform is arranged on, real-time detection and statistics for regional aim, while assess whether real-time target detection performance reaches
To expection.
Compared with prior art, the present invention has advantages below：
First, statistics is accurate：The present invention aids in contactless RFID technique in self-service knot using neural network image identification
The number of target object in label detection region is checked in calculation, meets the requirement of statistics zero defect.Solve RFID application limitations,
Than the self-help settlement mode of prevalence, accuracy and confidence level increase substantially.
2nd, self-studying mode：Using sample cumulative and iterative manner, algorithm is set to have from ripe ability, during with application
Between elapse, the degree of accuracy improves constantly, and in finite time, reaches 98% correct recognition rata, and as empirical data adds up,
Recognition accuracy tends to 100%, and the present invention is different from the Car license recognition of fixed meta schema, is more closely similar to face recognition technology, has
There is more general promotional value.
3rd, embedded system, plug and play：Cost is low, ageing height, is the optimal auxiliary of self-service checkout system.Make base
In RFID Digital Logistics information, observable object image information is obtained.Possibility is started for more applications.
Brief description of the drawings
Fig. 1 is flow chart of the method for the present invention.
Embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.
Embodiment
As shown in figure 1, " based on RFID technique, improved the purpose of the present invention is to propose to a kind of with reference to image recognition algorithm
The method of statistics accuracy rate ".The information of detection target is obtained using RFID, using image recognition to specific object physical examination
Survey, there is high accuracy, high robust, high real-time, so as to be accurate detection of the realization for target object, it is ensured that Bu Duojian
Survey also not missing inspection.
Compared to traditional neural network, deep neural network has more multi-neuron, can be to extensive diverse location, form
Region subject image sample learnt；, can be to same image by strengthening transform method simultaneously in data preparation stage
The operation such as turning-over changed, tone reversal, change of scale, noise disturbance, colour switching is carried out, it is special so as to generate a variety of varying environments
The training image of property so that Neural Network Detector is for different scenes, and varying environment, region object of different shapes can
The detection of stable performance is carried out, ensures the sample data volume of abundance, avoids over-fitting from occurring.In addition, the Dropout of neutral net
Mechanism can be rejected at random in the feed forward operation of training process partial nerve member, so as to reduce between neutral net it is mutual according to
The relation of relying, allows neutral net to possess stronger generalization ability, avoids over-fitting.
Idea of the invention is that：RFID system is established in self-help settlement equipment first, setting identification detects in the plane
Region, this setting regions realtime graphic is obtained.RFID detects object, while triggers image capture.Image detection algorithm
Visual identity module is imported based on the advance training for being sent into sample, and training result.When capture images input algoritic module, soon
Fast identification and statistics detectable substance.In view of self-help settlement is ageing, setting RFID information obtains the time and the image recognition time is same
Step, less than 2 seconds, that is, system obtained the result of statistical magnitude from two separate channels in 2 seconds.Using can be according to result
Make follow-up clearing action, it is ensured that settlement process is perfectly safe.
According to above-mentioned inventive concept, inventive algorithm module uses following technical proposals：
Target detection means based on neutral net, its concrete operation step are as follows：
(1), data prepare and demarcated：The image for including detection target is obtained from the multiple means such as network and reality shooting.
Sufficient training data is ensured by the mode such as data prediction and data enhancing conversion (data augmentation).Pass through mark
Note instrument is labeled to the target object of training sample data, obtains position and dimensional information.
(2), data are trained：Correct object detection information is obtained by annotation tool, the object that neutral net is detected
Classification, position, the error between size information and training data target object classification, actual size position is as target letter
Number.Training terminates when make it that object function reaches minimum value by tens thousand of training iterative process.Objectives function is as follows：
Wherein x, y, w, h are position and the size information of detection block, and C is the classification information of detection.Will by object function
Size, position and the classification information of detection combine, α, β, and λ is empirical parameter.
(3) data model test and evaluation：The data set for having neither part nor lot in mark and experiment by 20% evaluates mould as test set
Type performance, determine the accuracy rate of neutral net target detection model, recall rate, accurate rate and F1-Score.
Model is disposed：When neutral net object module produces a desired effect accuracy rate, model is deployed in mobile GPU and put down
Platform, real-time detection and statistics for regional aim, while assess whether real-time target detection performance reaches expected.
The present invention can be used in the self-service checkout system based on RFID technique.Common applicable cases are non-using RFID
Reading manner is contacted, obtains the object information by settling accounts area.Multiple objects simultaneously by when, once some object information read
Failure is taken, performance is exactly that statistics is omitted.In actual applications, it has to arrange human attendance, it is found that when discrepancy of quantity closes, by people
Work is corrected.Although now the workload of people is little, the unmanned clearing original intention apart from the Automation Design is still quite defective.
The present invention is used as supplementary means, and purpose is exactly the drawbacks of overcoming " automatic equipment needs eye-observation ", to accomplish machine
It is automatic to find mistake of statistics.Notice operation " this time clearing needs re-start ", or " need artificial nucleus to ".Substantially nothing is realized
People's self-help settlement on duty, homodyne are wrong.
Concrete practice case is as illustrated：
When RFID clearing area's activation (measured object occur), notice vision module starts；
Vision module captures the image of effective coverage, starts quick analysis；
Vision module exports：The figure of capture, and the physical quantities counted by algorithm.
Self-help settlement main frame compares the data that two kinds of technologies obtain, and determines to operate in next step, and preserve the figure.
Claims (6)
1. a kind of statistical magnitude method of view-based access control model, to check target object in label detection region in self-help settlement
Number, it is characterised in that comprise the following steps：
1) depth neuroid training is carried out on target object, builds neutral net target detection model；
2) target object in label detection region, number of lines of going forward side by side statistics are detected by RFID system；
3) triggered after RFID system detects the RFID information of target object and obtain the target object in label detection region
Image, and will be identified in the neutral net target detection model that trains of image input and statistics numbers；
4) statistics numbers of RFID system and neutral net target detection model are combined, it is final to obtain target in label detection region
The number of object：
If the counting of RFID system is identical with the counting of neutral net target detection model, using the counting as final meter
Number；
If the counting of RFID system is different from the counting of neutral net target detection model, alarmed, request carries out artificial
Count.
A kind of 2. statistical magnitude method of view-based access control model according to claim 1, it is characterised in that described step 1) tool
Body comprises the following steps：
11) data prepare and marked：Obtain several include target object images be used as training sample, by data prediction with
Data enhancing conversion ensures sufficient training data, and the target object in training sample is labeled by annotation tool；
12) data are trained：The detection information of correct target object is obtained by annotation tool, neutral net is detected into target
Error between classification, position and the size data of object and the classification of correct target object, physical location and actual size
As object function, by iteration so that object function completes training when reaching minimum value；
13) data model test and evaluation：It is iterated, accumulated and self study using test set, it is ensured that neutral net target is examined
Survey the accuracy rate of model.
A kind of 3. statistical magnitude method of view-based access control model according to claim 1, it is characterised in that described step 11)
In, several images comprising target object are obtained as training sample from by way of network crawl and reality shooting.
A kind of 4. statistical magnitude method of view-based access control model according to claim 1, it is characterised in that described step 12)
In, described object function is：
<mrow>
<mi>min</mi>
<mi> </mi>
<mi>J</mi>
<mo>=</mo>
<mi>&alpha;</mi>
<mo>*</mo>
<mi>&Sigma;</mi>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mover>
<mi>x</mi>
<mo>~</mo>
</mover>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mover>
<mi>y</mi>
<mo>~</mo>
</mover>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<mi>&beta;</mi>
<mo>*</mo>
<mi>&Sigma;</mi>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>w</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mover>
<mi>w</mi>
<mo>~</mo>
</mover>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>h</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mover>
<mi>h</mi>
<mo>~</mo>
</mover>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<mi>&lambda;</mi>
<mo>*</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>C</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mover>
<mi>C</mi>
<mo>~</mo>
</mover>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
Wherein, xi、yiThe position of i-th of target object is detected for neutral net,For the reality of i-th of correct target object
Border position, wi、hiThe size of i-th of target object is detected for neutral net,For the reality of i-th of correct target object
Border size, Ci、For the classification and concrete class of i-th of target object, α, β, λ are respectively empirical parameter.
A kind of 5. statistical magnitude method of view-based access control model according to claim 1, it is characterised in that described step 13)
In, the image of mark and training is had neither part nor lot in as test set by 20%.
6. the statistical magnitude method of a kind of view-based access control model according to claim 1, it is characterised in that described step 1) is also
Comprise the following steps：
14) when neutral net target detection model produces a desired effect accuracy rate, neutral net target detection model is set
In mobile GPU platform, real-time detection and statistics for regional aim, while assess whether real-time target detection performance reaches pre-
Phase.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710525089.7A CN107341538A (en) | 2017-06-30 | 2017-06-30 | A kind of statistical magnitude method of view-based access control model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710525089.7A CN107341538A (en) | 2017-06-30 | 2017-06-30 | A kind of statistical magnitude method of view-based access control model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107341538A true CN107341538A (en) | 2017-11-10 |
Family
ID=60219407
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710525089.7A Pending CN107341538A (en) | 2017-06-30 | 2017-06-30 | A kind of statistical magnitude method of view-based access control model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107341538A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107862775A (en) * | 2017-11-29 | 2018-03-30 | 深圳易伙科技有限责任公司 | A kind of supermarket's commodity anti-theft early warning system and method based on artificial intelligence |
| CN110245596A (en) * | 2019-06-05 | 2019-09-17 | 浙江大华技术股份有限公司 | A kind of monitoring method, monitor terminal and the monitoring system of special animal |
| CN110415240A (en) * | 2019-08-01 | 2019-11-05 | 国信优易数据有限公司 | Sample image generation method and device, circuit board defect detection method and device |
| CN111428684A (en) * | 2020-04-13 | 2020-07-17 | 捻果科技(深圳)有限公司 | Automatic identification method for operation specifications and number of airport apron operators |
| CN112037206A (en) * | 2020-09-01 | 2020-12-04 | 成都睿畜电子科技有限公司 | Target characteristic data calculation method and device and computer readable storage medium |
| CN112149768A (en) * | 2020-09-17 | 2020-12-29 | 北京计算机技术及应用研究所 | Method for counting number of cinema audiences by combining video monitoring and radio frequency identification |
| CN112215311A (en) * | 2020-09-07 | 2021-01-12 | 上海原能细胞生物低温设备有限公司 | Sample warehousing method and device, computer equipment and storage medium |
| CN112435479A (en) * | 2020-11-09 | 2021-03-02 | 浙江大华技术股份有限公司 | Target object violation detection method and device, computer equipment and system |
| CN112581016A (en) * | 2020-12-28 | 2021-03-30 | 深圳硅纳智慧科技有限公司 | Material management system and material management method adopting same |
| CN112991256A (en) * | 2020-12-11 | 2021-06-18 | 中国石油天然气股份有限公司 | Oil pipe material statistical method based on machine vision and deep learning |
| WO2021226893A1 (en) * | 2020-05-13 | 2021-11-18 | 鸿富锦精密工业(武汉)有限公司 | Object identification system and related device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1881271A (en) * | 2005-06-17 | 2006-12-20 | 东芝泰格有限公司 | Rfid tag recognition apparatus and method and article sales data processing apparatus |
| CN201853244U (en) * | 2010-08-25 | 2011-06-01 | 东莞市凌康信息技术有限公司 | Automatic error correction RFID (radio frequency identification) reading-writing device based on image processing |
| CN106845621A (en) * | 2017-01-18 | 2017-06-13 | 山东大学 | Dense population number method of counting and system based on depth convolutional neural networks |
| CN106845374A (en) * | 2017-01-06 | 2017-06-13 | 清华大学 | Pedestrian detection method and detection means based on deep learning |
-
2017
- 2017-06-30 CN CN201710525089.7A patent/CN107341538A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1881271A (en) * | 2005-06-17 | 2006-12-20 | 东芝泰格有限公司 | Rfid tag recognition apparatus and method and article sales data processing apparatus |
| CN201853244U (en) * | 2010-08-25 | 2011-06-01 | 东莞市凌康信息技术有限公司 | Automatic error correction RFID (radio frequency identification) reading-writing device based on image processing |
| CN106845374A (en) * | 2017-01-06 | 2017-06-13 | 清华大学 | Pedestrian detection method and detection means based on deep learning |
| CN106845621A (en) * | 2017-01-18 | 2017-06-13 | 山东大学 | Dense population number method of counting and system based on depth convolutional neural networks |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107862775A (en) * | 2017-11-29 | 2018-03-30 | 深圳易伙科技有限责任公司 | A kind of supermarket's commodity anti-theft early warning system and method based on artificial intelligence |
| CN110245596A (en) * | 2019-06-05 | 2019-09-17 | 浙江大华技术股份有限公司 | A kind of monitoring method, monitor terminal and the monitoring system of special animal |
| CN110415240A (en) * | 2019-08-01 | 2019-11-05 | 国信优易数据有限公司 | Sample image generation method and device, circuit board defect detection method and device |
| CN111428684A (en) * | 2020-04-13 | 2020-07-17 | 捻果科技(深圳)有限公司 | Automatic identification method for operation specifications and number of airport apron operators |
| WO2021226893A1 (en) * | 2020-05-13 | 2021-11-18 | 鸿富锦精密工业(武汉)有限公司 | Object identification system and related device |
| US11429800B2 (en) | 2020-05-13 | 2022-08-30 | Hong Fu Jin Precision Industry (Wuhan) Co., Ltd. | Object recognition system and related device |
| CN112037206A (en) * | 2020-09-01 | 2020-12-04 | 成都睿畜电子科技有限公司 | Target characteristic data calculation method and device and computer readable storage medium |
| CN112215311A (en) * | 2020-09-07 | 2021-01-12 | 上海原能细胞生物低温设备有限公司 | Sample warehousing method and device, computer equipment and storage medium |
| CN112149768A (en) * | 2020-09-17 | 2020-12-29 | 北京计算机技术及应用研究所 | Method for counting number of cinema audiences by combining video monitoring and radio frequency identification |
| CN112149768B (en) * | 2020-09-17 | 2024-05-14 | 北京计算机技术及应用研究所 | Method for counting cinema audience number by integrating video monitoring and radio frequency identification |
| CN112435479A (en) * | 2020-11-09 | 2021-03-02 | 浙江大华技术股份有限公司 | Target object violation detection method and device, computer equipment and system |
| CN112991256A (en) * | 2020-12-11 | 2021-06-18 | 中国石油天然气股份有限公司 | Oil pipe material statistical method based on machine vision and deep learning |
| CN112581016A (en) * | 2020-12-28 | 2021-03-30 | 深圳硅纳智慧科技有限公司 | Material management system and material management method adopting same |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107341538A (en) | A kind of statistical magnitude method of view-based access control model | |
| Li et al. | Automatic pavement crack detection by multi-scale image fusion | |
| CN108596277B (en) | Vehicle identity recognition method and device and storage medium | |
| KR101977174B1 (en) | Apparatus, method and computer program for analyzing image | |
| Tran et al. | A two-step sequential automated crack detection and severity classification process for asphalt pavements | |
| Krotosky et al. | Mutual information based registration of multimodal stereo videos for person tracking | |
| CN110110646B (en) | Gesture image key frame extraction method based on deep learning | |
| CN108875600A (en) | A kind of information of vehicles detection and tracking method, apparatus and computer storage medium based on YOLO | |
| CN101443817B (en) | Method and device for determining correspondence, preferably for the three-dimensional reconstruction of a scene | |
| CN107392885A (en) | A kind of method for detecting infrared puniness target of view-based access control model contrast mechanism | |
| CN103591940B (en) | Method of evaluating confidence of matching signature of hyperspectral image | |
| CN104166841A (en) | Rapid detection identification method for specified pedestrian or vehicle in video monitoring network | |
| CN109948684A (en) | Quality detecting method, device and its relevant device of point cloud data mark quality | |
| Liu et al. | A night pavement crack detection method based on image‐to‐image translation | |
| JPH11175730A (en) | Human body detection and trace system | |
| CN104715238A (en) | Pedestrian detection method based on multi-feature fusion | |
| Yuan et al. | Learning to count buildings in diverse aerial scenes | |
| Dib et al. | A review on negative road anomaly detection methods | |
| CN107452015A (en) | A kind of Target Tracking System with re-detection mechanism | |
| CN106991419A (en) | Method for anti-counterfeit based on tire inner wall random grain | |
| Chatterjee et al. | Intelligent Road Maintenance: a Machine Learning Approach for surface Defect Detection. | |
| CN104517095A (en) | Head division method based on depth image | |
| CN108875754A (en) | A kind of vehicle recognition methods again based on more depth characteristic converged network | |
| JP2008033462A (en) | Vehicle retrieval system | |
| CN109740609A (en) | A kind of gauge detection method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171110 |
|
| RJ01 | Rejection of invention patent application after publication |