CN112115902A - Dish identification method based on single-stage target detection algorithm - Google Patents
Dish identification method based on single-stage target detection algorithm Download PDFInfo
- Publication number
- CN112115902A CN112115902A CN202011023187.9A CN202011023187A CN112115902A CN 112115902 A CN112115902 A CN 112115902A CN 202011023187 A CN202011023187 A CN 202011023187A CN 112115902 A CN112115902 A CN 112115902A
- Authority
- CN
- China
- Prior art keywords
- dish
- identification
- feature extraction
- target detection
- stage target
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 33
- 238000000034 method Methods 0.000 title claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 16
- 238000007689 inspection Methods 0.000 claims abstract description 4
- 238000012986 modification Methods 0.000 claims abstract description 4
- 230000004048 modification Effects 0.000 claims abstract description 4
- 238000012706 support-vector machine Methods 0.000 claims description 16
- 238000012937 correction Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 description 4
- 235000003166 Opuntia robusta Nutrition 0.000 description 3
- 244000218514 Opuntia robusta Species 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000010411 cooking Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000003205 fragrance Substances 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 210000001525 retina Anatomy 0.000 description 1
- 235000015067 sauces Nutrition 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- 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
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/68—Food, e.g. fruit or vegetables
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a dish identification method based on a single-stage target detection algorithm, which comprises the steps of collecting dish image data, carrying out label inspection and label modification on the collected dish characteristic data, and generating a training data label file; transmitting the training data label file into a single-stage target detection algorithm for training, and reserving best weight to be stored in a background to obtain a convolutional neural network model for dish feature extraction; and calling a convolutional neural network model for dish feature extraction to perform feature extraction, and sending the extracted features to a classifier model of the SVM to perform training classification to obtain the classifier model. The convolutional neural network model is used for extracting the characteristics of dishes, training and classifying are carried out in the classifier model of the SVM, dish identification results are obtained, the process of extracting candidate areas is not explicitly given due to single-stage target detection, the characteristic extraction and detection are combined into a whole, the dish detection results can be directly obtained, and the advantages of high identification speed and high identification precision are considered.
Description
Technical Field
The invention relates to the technical field of image recognition, in particular to a dish recognition method based on a single-stage target detection algorithm.
Background
With the rapid development of society, in order to embody convenience and rapidness, most restaurants adopt the method of independently selecting dishes and queuing for settlement at the present stage. As the checkout is divided into manual pricing and automatic pricing, the defects that the manual pricing efficiency is low, the accuracy cannot be guaranteed and the like are revealed along with the increase of the number of eaters. With the rapid increase of the scale of artificial intelligence and the mobile-end internet, people have more and more demands on "intellectualization", so that some automatic pricing methods appear in recent years. In the aspect of catering industry, intelligent dish identification and settlement equipment is introduced into dining halls of more and more colleges and large-scale enterprises at present.
The existing intelligent settlement system for the RFID tableware is characterized in that an RFID radio frequency chip is embedded in the bottom of a dinner plate, the intelligent settlement table identifies the serial number of the RFID chip in the dinner plate and calculates the amount of dishes, the RFID chip is embedded in each dinner plate in the mode, the chip is easy to damage at high temperature, the manufacture is troublesome, hot dishes are not contained conveniently, the dishes and the tableware are required to be in one-to-one correspondence, and the operation is complex.
The dish identification system comprises an acquisition module, a picture processing module, a picture identification module, a control module and a code generation module, wherein the acquisition module comprises a photographing module and a receiving module, the photographing module is used for photographing a picture for identification and directly photographing information by using a photographing function, the receiving module is used for receiving the transmitted picture through a network and then identifying the picture, the picture processing module comprises a filtering module, a binarization and denoising processing module and a segmentation module, the filtering module is used for filtering illegal pictures with illegal characteristics, and the binarization and denoising processing module is used for removing useless frames and boundaries of the pictures. The method uses the traditional image contrast matching technology, and as the dishes made in restaurants and dining halls are various and the color, the fragrance, the taste and the shape after cooking are different each time, the method has larger errors and uncontrollable properties in identifying the outline, the characteristics, the color and the like of the dishes, and has low identification accuracy, low speed and low efficiency.
Disclosure of Invention
The dish identification method based on the single-stage target detection algorithm is high in identification precision and identification speed.
The dish identification method based on the single-stage target detection algorithm comprises the following steps:
s1, collecting dish image data;
s2, performing label inspection and label modification on the collected dish image data to generate a training data label file;
s3, transmitting the training data label file into a single-stage target detection algorithm for training, reserving best weight and storing the best weight in a background to obtain a convolutional neural network model for dish feature extraction capable of identifying dishes;
s4, calling a convolutional neural network model for dish feature extraction to perform feature extraction, and sending the extracted features to a classifier model of an SVM (support vector machine) to perform training classification to obtain a classifier model;
and S5, identifying the dish to be identified, transmitting the dish picture to a convolutional neural network model for dish feature extraction after the dish picture is acquired by the system, and transmitting the dish picture to a trained SVM (support vector machine) classifier for classification after the feature extraction to obtain a dish identification result.
According to the dish identification method based on the single-stage target detection algorithm, the collected dish image data is trained to form a convolutional neural network model for dish feature extraction by the single-stage target detection algorithm, the convolutional neural network model is used for dish feature extraction, training and classification are further carried out in a classifier model of an SVM (support vector machine) to obtain a dish identification result, the process of extracting a candidate area is not explicitly given due to single-stage target detection, the feature extraction and detection are combined into one, the dish detection result can be directly obtained, dish features are accurately extracted and intelligently detected under the condition that the detection speed is guaranteed, and the advantages of high identification speed and high identification precision are considered.
Drawings
Fig. 1 is a block diagram of the steps of a dish identification method based on a single-stage target detection algorithm.
Detailed Description
As shown in fig. 1, a dish identification method based on a single-stage target detection algorithm includes the following steps:
s1, collecting dish image data;
s2, performing label inspection and label modification on the collected dish image data to generate a training data label file;
s3, transmitting the training data label file into a single-stage target detection algorithm for training, reserving best weight and storing the best weight in a background to obtain a convolutional neural network model for dish feature extraction capable of identifying dishes;
s4, calling a convolutional neural network model for dish feature extraction to perform feature extraction, and sending the extracted features to a classifier model of an SVM (support vector machine) to perform training classification to obtain a classifier model;
and S5, identifying the dish to be identified, transmitting the dish picture to a convolutional neural network model for dish feature extraction after the dish picture is acquired by the system, and transmitting the dish picture to a trained SVM (support vector machine) classifier for classification after the feature extraction to obtain a dish identification result.
According to the dish identification method based on the single-stage target detection algorithm, the collected dish image data is trained to form a convolutional neural network model for dish feature extraction by the single-stage target detection algorithm, the convolutional neural network model is used for dish feature extraction, training and classification are further carried out in a classifier model of an SVM (support vector machine) to obtain the classifier model, the process of extracting a candidate region is not explicitly given due to single-stage target detection, the feature extraction and detection are combined into one, the dish detection result can be directly obtained, dish features are accurately extracted and intelligently detected under the condition that the detection speed is guaranteed, and the advantages of high identification speed and high identification precision are considered.
The dish identification method based on the single-stage target detection algorithm comprises YOO series, SSD series and RetinaNet (retina network). And (4) transmitting the training data label file into a single-stage target detection algorithm for training, and reserving best weight to be stored in a background to obtain a dish identification model based on the single-stage target detection algorithm.
The dish identification method based on the single-stage target detection algorithm further comprises an alarm mechanism for identification settlement, the dishes are selected through the display frame, an identification threshold value of confidence coefficient identification scores is set, when the dishes are identified, when new dishes are increased and the confidence coefficient identification scores are close, an alarm is given to prompt on a dish identification interface, and manual intervention correction identification is carried out. If the dish categories are too similar and the dish categories are identified incorrectly, the confidence coefficient identification score is too high, dish identification category correction can be performed, the dish with the identification error is selected, a plurality of identified approximate dishes and corresponding confidence coefficient identification score values are given, and correct dish categories are manually selected from the plurality of identified approximate dishes. When dishes are made, the dishes are subject to various objective factors, such as: the influence of the surface light reflection of the dishes, the fresh-keeping film wrapped by the dishes, sauce, ingredients and the like can often cause the appearance of similar dishes which are very similar in surface characteristics and easy to identify and confuse. The method can be used for reminding workers that the recognition result under the condition is not necessarily credible and needs to be manually checked, so that the false recognition risk brought by similar dishes is reduced. Dishes are subject to external factors in the identification process, such as: the hand/mobile phone can reduce the dish identification effect due to the influences of shielding, shadow walking, overlapping or external strong light interference and the like. Therefore, for the recognition result with the high confidence score, the recognition frames with different colors are different from the normal recognition frame to remind that the recognition result of the current dish needs to be manually checked, so that the risk of false recognition is reduced.
Claims (2)
1. A dish identification method based on a single-stage target detection algorithm is characterized by comprising the following steps:
s1, collecting dish image data;
s2, performing label inspection and label modification on the collected dish image data to generate a training data label file;
s3, transmitting the training data label file into a single-stage target detection algorithm for training, reserving best weight and storing the best weight in a background to obtain a convolutional neural network model for dish feature extraction capable of identifying dishes;
s4, calling a convolutional neural network model for dish feature extraction to perform feature extraction, and sending the extracted features to a classifier model of an SVM (support vector machine) to perform training classification to obtain a classifier model;
and S5, identifying the dish to be identified, transmitting the dish picture to a convolutional neural network model for dish feature extraction after the dish picture is acquired by the system, and transmitting the dish picture to a trained SVM (support vector machine) classifier for classification after the feature extraction to obtain a dish identification result.
2. A dish identification method based on a single-stage target detection algorithm is characterized in that an alarm mechanism is further included for identification settlement, dishes are selected through a display frame, an identification threshold value of confidence identification scores is set, when the dishes are identified, when new dishes are increased and the confidence identification scores are close, an alarm is timely given out on a dish identification interface for reminding, and manual intervention correction identification is carried out.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011023187.9A CN112115902A (en) | 2020-09-25 | 2020-09-25 | Dish identification method based on single-stage target detection algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011023187.9A CN112115902A (en) | 2020-09-25 | 2020-09-25 | Dish identification method based on single-stage target detection algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112115902A true CN112115902A (en) | 2020-12-22 |
Family
ID=73798116
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011023187.9A Pending CN112115902A (en) | 2020-09-25 | 2020-09-25 | Dish identification method based on single-stage target detection algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112115902A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113205133A (en) * | 2021-04-30 | 2021-08-03 | 成都国铁电气设备有限公司 | Tunnel water stain intelligent identification method based on multitask learning |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203330A (en) * | 2016-07-08 | 2016-12-07 | 西安理工大学 | A kind of vehicle classification method based on convolutional neural networks |
| CN106845527A (en) * | 2016-12-29 | 2017-06-13 | 南京江南博睿高新技术研究院有限公司 | A kind of vegetable recognition methods |
| CN106874913A (en) * | 2016-12-29 | 2017-06-20 | 南京江南博睿高新技术研究院有限公司 | A kind of vegetable detection method |
| CN109214409A (en) * | 2018-07-10 | 2019-01-15 | 上海斐讯数据通信技术有限公司 | A kind of vegetable recognition methods and system |
| CN109508664A (en) * | 2018-10-26 | 2019-03-22 | 浙江师范大学 | A kind of vegetable identification pricing method based on deep learning |
| CN109684922A (en) * | 2018-11-20 | 2019-04-26 | 浙江大学山东工业技术研究院 | A kind of recognition methods based on the multi-model of convolutional neural networks to finished product dish |
| CN109830072A (en) * | 2019-02-26 | 2019-05-31 | 魔珐(上海)信息科技有限公司 | The valuation of view-based access control model identification and cash device, control system and method |
| CN111160428A (en) * | 2019-12-17 | 2020-05-15 | 金陵科技学院 | Automatic vegetable identification method based on CNN-SVM algorithm |
| CN111325084A (en) * | 2019-08-29 | 2020-06-23 | 西安铱食云餐饮管理有限公司 | Dish information identification method and terminal based on YOLO neural network |
| CN111680603A (en) * | 2020-05-28 | 2020-09-18 | 浙江师范大学 | Dish detection and identification method |
-
2020
- 2020-09-25 CN CN202011023187.9A patent/CN112115902A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203330A (en) * | 2016-07-08 | 2016-12-07 | 西安理工大学 | A kind of vehicle classification method based on convolutional neural networks |
| CN106845527A (en) * | 2016-12-29 | 2017-06-13 | 南京江南博睿高新技术研究院有限公司 | A kind of vegetable recognition methods |
| CN106874913A (en) * | 2016-12-29 | 2017-06-20 | 南京江南博睿高新技术研究院有限公司 | A kind of vegetable detection method |
| CN109214409A (en) * | 2018-07-10 | 2019-01-15 | 上海斐讯数据通信技术有限公司 | A kind of vegetable recognition methods and system |
| CN109508664A (en) * | 2018-10-26 | 2019-03-22 | 浙江师范大学 | A kind of vegetable identification pricing method based on deep learning |
| CN109684922A (en) * | 2018-11-20 | 2019-04-26 | 浙江大学山东工业技术研究院 | A kind of recognition methods based on the multi-model of convolutional neural networks to finished product dish |
| CN109830072A (en) * | 2019-02-26 | 2019-05-31 | 魔珐(上海)信息科技有限公司 | The valuation of view-based access control model identification and cash device, control system and method |
| CN111325084A (en) * | 2019-08-29 | 2020-06-23 | 西安铱食云餐饮管理有限公司 | Dish information identification method and terminal based on YOLO neural network |
| CN111160428A (en) * | 2019-12-17 | 2020-05-15 | 金陵科技学院 | Automatic vegetable identification method based on CNN-SVM algorithm |
| CN111680603A (en) * | 2020-05-28 | 2020-09-18 | 浙江师范大学 | Dish detection and identification method |
Non-Patent Citations (1)
| Title |
|---|
| 朱瑶 等: "基于深度卷积神经网络的菜品识别", 常州信息职业技术学院学报, vol. 19, no. 04, pages 35 - 40 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113205133A (en) * | 2021-04-30 | 2021-08-03 | 成都国铁电气设备有限公司 | Tunnel water stain intelligent identification method based on multitask learning |
| CN113205133B (en) * | 2021-04-30 | 2024-01-26 | 成都国铁电气设备有限公司 | Tunnel water stain intelligent identification method based on multitask learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Aguilar et al. | Grab, pay, and eat: Semantic food detection for smart restaurants | |
| CN109508664B (en) | Dish identification pricing method based on deep learning | |
| US10853702B2 (en) | Method and apparatus for checkout based on image identification technique of convolutional neural network | |
| CN111178197B (en) | Mass R-CNN and Soft-NMS fusion based group-fed adherent pig example segmentation method | |
| CN106980852B (en) | Based on Corner Detection and the medicine identifying system matched and its recognition methods | |
| Huang et al. | Real-time classification of green coffee beans by using a convolutional neural network | |
| CN110059654A (en) | A kind of vegetable Automatic-settlement and healthy diet management method based on fine granularity identification | |
| CN110705424B (en) | Method and device for positioning commodity display position and storage medium | |
| CN112115906A (en) | Open dish identification method based on deep learning target detection and metric learning | |
| CN111680603A (en) | Dish detection and identification method | |
| CN111161295B (en) | Dish image background stripping method | |
| CN106056487A (en) | Tableware-pattern-based pricing method of dish automatic identification system | |
| CN105389581A (en) | Germinated rice germ integrity intelligent identification system and identification method thereof | |
| CN111832590A (en) | Article identification method and system | |
| CN104200249A (en) | Automatic clothes matching method, device and system | |
| CN111080493A (en) | Dish information identification method and device and dish self-service settlement system | |
| CN108647696B (en) | Picture color value determining method and device, electronic equipment and storage medium | |
| CN109685030A (en) | A kind of mug rim of a cup defects detection classification method based on convolutional neural networks | |
| CN111640268A (en) | Intelligent settlement method and system based on dinner plate shape and color | |
| CN108090517A (en) | A kind of cereal recognition methods, device and computer storage media | |
| CN112037440A (en) | Intelligent settlement system and settlement method based on machine learning | |
| CN112508132A (en) | Training method and device for identifying SKU | |
| CN112784640A (en) | Menu making method and device and cooking machine | |
| CN116385430A (en) | Machine vision flaw detection method, device, medium and equipment | |
| CN112115902A (en) | Dish identification method based on single-stage target detection algorithm |
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 |