CN110378311A - Violation judgment method in kitchen after food and drink based on Encoder-Decoder model and mixed Gauss model - Google Patents
Violation judgment method in kitchen after food and drink based on Encoder-Decoder model and mixed Gauss model Download PDFInfo
- Publication number
- CN110378311A CN110378311A CN201910679127.3A CN201910679127A CN110378311A CN 110378311 A CN110378311 A CN 110378311A CN 201910679127 A CN201910679127 A CN 201910679127A CN 110378311 A CN110378311 A CN 110378311A
- Authority
- CN
- China
- Prior art keywords
- violation
- encoder
- decoder
- model
- picture
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses violation judgment methods in kitchen after a kind of food and drink based on Encoder-Decoder model and mixed Gauss model, are related to technical field of image processing.Including the modelling phase: preparing violation picture, construct violation atlas；The violation region in violation picture is marked, violation region collection is constructed；Construct Encoder-Decoder model, including encoder Encoder and decoder Decoder；It is restrained using the violation atlas as sample training Encoder-Decoder model to cost function up to standard；It is input building mixed Gauss model with the coding result of trained encoder Encoder, output result is violation probability.Implementation phase: the coding result of picture to be detected is extracted using trained encoder；Coding result is inputted into mixed Gauss model, calculates violation probability；Judged in the picture to be detected according to violation probability with the presence or absence of unlawful practice.The present invention can precisely extract the visual signature of violation, and not by rear kitchen scene restriction, preferable for the violation identification Generalization Capability of different food and drink shops.
Description
Technical field
The present invention relates to technical field of image processing, one kind being based on Encoder-Decoder model and mixed Gauss model
Food and drink after kitchen violation judgment method.
Background technique
Computer vision is a technology ripe day by day in recent years, the intelligent video prison based on computer vision technique
Control technology is widely used currently with people's life water in the scenes such as food and drink, company, gymnasium, building site and railway station
Flat raising, the requirement to catering industry also increase accordingly, and in order to standardize catering industry increasingly to the management in rear kitchen, are based on
The food and drink intelligent Video Surveillance Technology of computer vision technique also comes into being
Food and drink intelligent video monitoring common practices based on computer vision technique is the method for using target detection first
The target for detecting the refinement in current video frame is then based on the semantic information in some rear kitchens to the operator in present frame
Movement and dressing make a decision, so that whether the operator judged in present frame that makes a policy all meets rear kitchen standard, look for
There are the pictures of operator's violation to be pushed to Catering Management person in video frame out, saves catering industry and looks into rear kitchen violation operation
The human capital looked for effectively manages and supervises each shops for catering industry and provides reliable solution
But current algorithm has some limitations, and is in the rear kitchen scene of catering industry first, visual background is miscellaneous
Unrest and there are a large amount of redundancy interference informations, certain interference can be caused to algorithm of target detection, leads to algorithm of target detection
Performance boost is limited and followed by detects that target combines some semantic informations in rear kitchen artificially to formulate phase based on algorithm of target detection
The rule answered finds out the violation item in rear kitchen tour standard, and this method is simple in rule in processing, when visual signature is single clear, intelligence
Energy algorithm can accurately find the problem, and video tour amount be reduced, to improve efficiency.But it is practical in the tour item in rear kitchen
Clear Rulemaking can not be much used, causes intelligent algorithm that can only find out the doubtful item of some violations, rate of false alarm can be relatively high, to mentioning
It is limited to rise working efficiency help.
Summary of the invention
The purpose of the present invention is to provide after a kind of food and drink based on Encoder-Decoder model and mixed Gauss model
Violation judgment method in kitchen can precisely extract the visual signature of violation, and not by rear kitchen scene restriction, for different food and drink doors
The violation identification Generalization Capability in shop is preferable.
To achieve the above object, the invention provides the following technical scheme:
Violation judgment method in kitchen after a kind of food and drink based on Encoder-Decoder model and mixed Gauss model, it is special
Sign is, including modelling phase and implementation phase；
The modelling phase includes:
S1 prepares violation picture, constructs violation atlas；
S2 marks the violation region in violation picture, constructs violation region collection；
S3 constructs Encoder-Decoder model, including encoder Encoder and decoder Decoder；
S4, using the violation atlas as sample training Encoder-Decoder model, until Encoder-Decoder mould
Cost function between the input and output of type converges to first threshold；
S5, is input building mixed Gauss model with the coding result of trained encoder Encoder, and output result is
Violation probability；
The implementation phase includes:
S6 extracts the coding result of picture to be detected using trained encoder；
S7, the coding result that S6 is acquired input mixed Gauss model, calculate violation probability；
S8 judges in the picture to be detected according to violation probability with the presence or absence of unlawful practice.
Further, the decoder Decoder is the decoupling-structure of encoder Encoder.
Further, the cost function is specific as follows:
Wherein, WijFor weight, the position of pixel (i, j) is depended on, if pixel (i, j) is located at the violation of corresponding picture
In region, then Wij=W, 0.6≤W≤1, otherwise Wij=1-W；I (i, j) is the pixel value of pixel (i, j) in violation picture；
Id (i, j) is with the pixel of pixel (i, j) in the output result for the Encoder-Decoder model that the violation picture is input
Value.
Further, the W takes 0.7.
Further, the first threshold is 0.001.
Further, the mixed Gauss model are as follows:
Wherein, K is the number of the Gaussian component in mixed Gauss model；wk、μk、ΣkRespectively k-th Gaussian component
Weight, mean value, covariance matrix are determined by EM algorithm；0≤wk≤ 1 and
Further, the particular content judged in the S8 is as follows: second threshold is preset, if violation probability is greater than institute
Second threshold is stated, then there are unlawful practices in the picture to be detected, and unlawful practice is otherwise not present in the picture to be detected.
Further, the violation picture and picture to be detected are the video frame that video monitoring obtains.
Compared with prior art, the beneficial effects of the present invention are: present invention combination violation picture and corresponding violation region
Training Encoder-Decoder model, lays particular emphasis on the image within violation region, and it is dry to eliminate a large amount of redundancy in rear kitchen background
Disturb influence of the information to model accuracy.Finally, the present invention is using the output of mixed Gauss model as violation probability, final judgement
It whether there is unlawful practice in picture to be detected.The present invention can precisely extract the visual signature of violation, and not by rear kitchen
Scape limitation, it is preferable for the violation identification Generalization Capability of different food and drink shops.
Detailed description of the invention
Fig. 1 is overall flow figure of the invention.
Fig. 2 is Encoder-Decoder model schematic of the invention.
Specific embodiment
The following is a clear and complete description of the technical scheme in the embodiments of the invention, it is clear that described embodiment
Only a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, the common skill in this field
Art personnel every other embodiment obtained without making creative work belongs to the model that the present invention protects
It encloses.
Referring to Fig. 1, after the present invention provides a kind of food and drink based on Encoder-Decoder model and mixed Gauss model
Kitchen violation judgment method, including modelling phase and implementation phase；
The modelling phase includes:
S1, the video monitoring in kitchen after acquisition take 30000 video frames for including unlawful practice as violation picture, building
Violation atlas I=I1, I2 ..., IN }；Wherein, N is the total quantity of violation picture, i.e. 30000 in the present embodiment.
S2 marks the violation region in each violation picture, constructs violation region collection R={ R1, R2 ..., RN }；In violation of rules and regulations
Region is specially the set of pixel position of the unlawful practice in corresponding violation picture.Such as: for there are unlawful practice " kitchens
Shi Wei is with cook's cap " violation picture for, violation region is the position of the pixel in the violation picture on cook head.
S3 constructs Encoder-Decoder model, including encoder Encoder and decoder Decoder.Please refer to figure
The input of 2, encoder Encoder are picture I, and output is coding result X, which is the vector of m dimension；Decoding
The input of device Decoder is the coding result X of encoder Encoder, and output is picture Id, it may be assumed that
X=Encoder (I)
Id=Decoder (X)
It is noted that the encoder Encoder is the mainstream deep learning using computer vision field
Network is realized, such as Resnet50, VGG16 etc.；The decoder Decoder is the decoupling-structure of encoder Encoder.
S4, using the violation atlas as sample, using stochastic gradient descent method training Encoder-Decoder model, directly
First threshold is converged to the cost function between the input and output of Encoder-Decoder model, first threshold is preferably
0.001.Wherein, the cost function is specific as follows:
Wherein, WijFor weight, the position of pixel (i, j) is depended on, if pixel (i, j) is located at the violation of corresponding picture
In region, then Wij=W, 0.6≤W≤1, otherwise Wij=1-W, W preferably take 0.7.I (i, j) be violation picture in pixel (i,
J) pixel value；It is pixel in the output result of the Encoder-Decoder model inputted that Id (i, j), which is with the violation picture,
The pixel value of (i, j).
S5, is input building mixed Gauss model with the coding result of trained encoder Encoder, and output result is
Violation probability；The mixed Gauss model is established using EM algorithm, specific as follows:
Wherein, K is the number of the Gaussian component in mixed Gauss model；wk、μk、ΣkRespectively k-th Gaussian component
Weight, mean value, covariance matrix are determined by EM algorithm；0≤wk≤ 1 and
The implementation phase includes:
S6 extracts the coding result of picture I to be detected using trained encoder；
X=Encoder (I)；
S7, the coding result that S6 is acquired input mixed Gauss model, calculate violation probability；
P=GMM (X)；
S8, judged in the picture to be detected according to violation probability P with the presence or absence of unlawful practice: particular content is as follows: in advance
It is 0.8 that second threshold, which is arranged, if violation probability is greater than the second threshold, there are unlawful practices in the picture to be detected, no
Unlawful practice is then not present in the picture to be detected.
Further, the violation picture and picture to be detected are the video frame that video monitoring obtains.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included within the present invention.
Claims (8)
1. kitchen violation judgment method, feature after a kind of food and drink based on Encoder-Decoder model and mixed Gauss model
It is, including modelling phase and implementation phase；
The modelling phase includes:
S1 prepares violation picture, constructs violation atlas；
S2 marks the violation region in violation picture, constructs violation region collection；
S3 constructs Encoder-Decoder model, including encoder Encoder and decoder Decoder；
S4, using the violation atlas as sample training Encoder-Decoder model, until Encoder-Decoder model
Cost function between input and output converges to first threshold；
S5, is input building mixed Gauss model with the coding result of trained encoder Encoder, and output result is in violation of rules and regulations
Probability；
The implementation phase includes:
S6 extracts the coding result of picture to be detected using trained encoder；
S7, the coding result that S6 is acquired input mixed Gauss model, calculate violation probability；
S8 judges in the picture to be detected according to violation probability with the presence or absence of unlawful practice.
2. kitchen violation after the food and drink according to claim 1 based on Encoder-Decoder model and mixed Gauss model
Judgment method, which is characterized in that the decoder Decoder is the decoupling-structure of encoder Encoder.
3. kitchen violation after the food and drink according to claim 1 based on Encoder-Decoder model and mixed Gauss model
Judgment method, which is characterized in that the cost function is specific as follows:
Wherein, WijFor weight, the position of pixel (i, j) is depended on, if pixel (i, j) is located at the violation region of corresponding picture
It is interior, then Wij=W, 0.6≤W≤1, otherwise Wij=1-W；I (i, j) is the pixel value of pixel (i, j) in violation picture；Id(i,
J) for the pixel value of pixel (i, j) in the output result for the Encoder-Decoder model that the violation picture is input.
4. kitchen violation after the food and drink according to claim 3 based on Encoder-Decoder model and mixed Gauss model
Judgment method, which is characterized in that the W takes 0.7.
5. kitchen violation after the food and drink according to claim 1 based on Encoder-Decoder model and mixed Gauss model
Judgment method, which is characterized in that the first threshold is 0.001.
6. kitchen violation after the food and drink according to claim 1 based on Encoder-Decoder model and mixed Gauss model
Judgment method, which is characterized in that the mixed Gauss model are as follows:
Wherein, K is the number of the Gaussian component in mixed Gauss model；wk、μk、ΣkThe weight of respectively k-th Gaussian component,
Mean value, covariance matrix are determined by EM algorithm；0≤wk≤ 1 and
7. kitchen violation after the food and drink according to claim 1 based on Encoder-Decoder model and mixed Gauss model
Judgment method, which is characterized in that the particular content judged in the S8 is as follows: presetting second threshold, if violation probability is big
In the second threshold, then there are unlawful practices in the picture to be detected, and unlawful practice is otherwise not present in the picture to be detected.
8. kitchen violation after the food and drink according to claim 1 based on Encoder-Decoder model and mixed Gauss model
Judgment method, the violation picture and picture to be detected are the video frame that video monitoring obtains.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910679127.3A CN110378311A (en) | 2019-07-25 | 2019-07-25 | Violation judgment method in kitchen after food and drink based on Encoder-Decoder model and mixed Gauss model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910679127.3A CN110378311A (en) | 2019-07-25 | 2019-07-25 | Violation judgment method in kitchen after food and drink based on Encoder-Decoder model and mixed Gauss model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110378311A true CN110378311A (en) | 2019-10-25 |
Family
ID=68256203
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910679127.3A Pending CN110378311A (en) | 2019-07-25 | 2019-07-25 | Violation judgment method in kitchen after food and drink based on Encoder-Decoder model and mixed Gauss model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110378311A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507320A (en) * | 2020-07-01 | 2020-08-07 | 平安国际智慧城市科技股份有限公司 | Detection method, device, equipment and storage medium for kitchen violation behaviors |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101729911A (en) * | 2009-12-23 | 2010-06-09 | 宁波大学 | Multi-view image color correction method based on visual perception |
| CN105578183A (en) * | 2015-12-16 | 2016-05-11 | 西安交通大学 | Compression sensing video encoding and decoding method based on Gaussian mixture model (GMM) |
| EP3223253A1 (en) * | 2016-03-23 | 2017-09-27 | Thomson Licensing | Multi-stage audio activity tracker based on acoustic scene recognition |
| CN107341790A (en) * | 2017-06-12 | 2017-11-10 | 广州大学 | A kind of image processing method of environment cleanliness detection |
| CN108052900A (en) * | 2017-12-12 | 2018-05-18 | 成都睿码科技有限责任公司 | A kind of method by monitor video automatic decision dressing specification |
| CN109815851A (en) * | 2019-01-03 | 2019-05-28 | 深圳壹账通智能科技有限公司 | Kitchen hygiene detection method, device, computer equipment and storage medium |
-
2019
- 2019-07-25 CN CN201910679127.3A patent/CN110378311A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101729911A (en) * | 2009-12-23 | 2010-06-09 | 宁波大学 | Multi-view image color correction method based on visual perception |
| CN105578183A (en) * | 2015-12-16 | 2016-05-11 | 西安交通大学 | Compression sensing video encoding and decoding method based on Gaussian mixture model (GMM) |
| EP3223253A1 (en) * | 2016-03-23 | 2017-09-27 | Thomson Licensing | Multi-stage audio activity tracker based on acoustic scene recognition |
| CN107341790A (en) * | 2017-06-12 | 2017-11-10 | 广州大学 | A kind of image processing method of environment cleanliness detection |
| CN108052900A (en) * | 2017-12-12 | 2018-05-18 | 成都睿码科技有限责任公司 | A kind of method by monitor video automatic decision dressing specification |
| CN109815851A (en) * | 2019-01-03 | 2019-05-28 | 深圳壹账通智能科技有限公司 | Kitchen hygiene detection method, device, computer equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 王硕，等: ""基于一维卷积自编码器—高斯混合模型的间歇过程故障检测"", 《信息与控制》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507320A (en) * | 2020-07-01 | 2020-08-07 | 平安国际智慧城市科技股份有限公司 | Detection method, device, equipment and storage medium for kitchen violation behaviors |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105868689B (en) | A kind of face occlusion detection method based on concatenated convolutional neural network | |
| Huang et al. | Detection algorithm of safety helmet wearing based on deep learning | |
| CN104992167B (en) | A kind of method for detecting human face and device based on convolutional neural networks | |
| CN110569772B (en) | Method for detecting state of personnel in swimming pool | |
| CN103984915B (en) | Pedestrian's recognition methods again in a kind of monitor video | |
| EP3869459A1 (en) | Target object identification method and apparatus, storage medium and electronic apparatus | |
| CN105095856B (en) | Face identification method is blocked based on mask | |
| CN109101865A (en) | A kind of recognition methods again of the pedestrian based on deep learning | |
| CN108052900A (en) | A kind of method by monitor video automatic decision dressing specification | |
| CN106815566A (en) | A kind of face retrieval method based on multitask convolutional neural networks | |
| CN109190475A (en) | A kind of recognition of face network and pedestrian identify network cooperating training method again | |
| CN113903081A (en) | Visual identification artificial intelligence alarm method and device for images of hydraulic power plant | |
| CN206224639U (en) | A kind of face recognition door control system with occlusion detection function | |
| CN205427946U (en) | Hotel moves in management system | |
| CN111611874A (en) | Face mask wearing detection method based on ResNet and Canny | |
| CN109492575A (en) | A kind of staircase safety monitoring method based on YOLOv3 | |
| CN109902613A (en) | A kind of human body feature extraction method based on transfer learning and image enhancement | |
| CN103886305A (en) | Specific face searching method for grassroots policing, safeguard stability and counter-terrorism | |
| WO2022121498A1 (en) | Identity recognition method, model training method, apparatuses, and device and storage medium | |
| CN109117774A (en) | A kind of multi-angle video method for detecting abnormality based on sparse coding | |
| CN116563797B (en) | Monitoring management system for intelligent campus | |
| CN109344720A (en) | A kind of affective state detection method based on adaptive features select | |
| CN110378311A (en) | Violation judgment method in kitchen after food and drink based on Encoder-Decoder model and mixed Gauss model | |
| CN114821647A (en) | Sleeping post identification method, device, equipment and medium | |
| CN113111733A (en) | Posture flow-based fighting behavior recognition method |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191025 |