CN112862762A - Deep learning-based food material feature extraction and compression method - Google Patents
Deep learning-based food material feature extraction and compression method Download PDFInfo
- Publication number
- CN112862762A CN112862762A CN202110078834.4A CN202110078834A CN112862762A CN 112862762 A CN112862762 A CN 112862762A CN 202110078834 A CN202110078834 A CN 202110078834A CN 112862762 A CN112862762 A CN 112862762A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- food material
- feature extraction
- image
- compression method
- 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
- 239000000463 material Substances 0.000 title claims abstract description 31
- 238000013135 deep learning Methods 0.000 title claims abstract description 30
- 238000000605 extraction Methods 0.000 title claims abstract description 28
- 238000000034 method Methods 0.000 title claims abstract description 20
- 230000006835 compression Effects 0.000 title claims abstract description 13
- 238000007906 compression Methods 0.000 title claims abstract description 13
- 238000005516 engineering process Methods 0.000 abstract description 3
- 239000011159 matrix material Substances 0.000 description 6
- 238000001514 detection method Methods 0.000 description 5
- 230000009466 transformation Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 3
- 230000007423 decrease Effects 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 229920006395 saturated elastomer Polymers 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- 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
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20164—Salient point detection; Corner detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to the field of image search and the field of intelligent home furnishing, and particularly relates to a food material feature extraction and compression method based on deep learning, which comprises the steps of configuring a refrigerator camera, and enabling the camera to clearly shoot a process of a user for storing and taking food materials; step two, training a deep learning algorithm model; step three, feature extraction; step four, compressing the characteristic values; and step five, warehousing the characteristic values. The method solves the problem that the features of all dimensions of the image cannot be well extracted by the traditional feature extraction algorithm, the features extracted by the deep learning algorithm can well reflect the features of all dimensions of the image, the feature value extraction of the intelligent refrigerator food material is completed by utilizing the deep learning technology, the problem that the feature extraction of the traditional feature value extraction method is insufficient is solved, the features of all dimensions of the image can be extracted by the deep learning algorithm, and the accuracy of image search feature value comparison and refrigerator food material identification is improved.
Description
Technical Field
The invention belongs to the field of image search and the field of smart home, and particularly relates to a deep learning-based food material feature extraction and compression method.
Background
With the development of intelligent refrigerators, image search technologies are increasingly applied to refrigerator food identification, image search comparison is performed on refrigerator food, features of an image are extracted firstly, and a traditional image feature extraction method such as a Harris (detection corner) SIFT (detection blob) SURF (detection blob) FAST (detection corner) BRIEF (detection blob) ORB (FAST algorithm with direction and BRIEF algorithm with rotation invariance) mainly extracts features of the image according to gradients of the image, and features of all dimensions of the image cannot be extracted sufficiently, so that image search accuracy is low, and a better feature extraction method is found.
Disclosure of Invention
To address the deficiencies noted in the background, the present invention is presented.
In order to achieve the above purpose, the following technical solutions are provided:
a deep learning-based food material feature extraction and compression method comprises the steps of
The method comprises the following steps that firstly, a refrigerator camera is configured, so that the process that a user accesses food materials can be clearly shot by the camera;
step two, training a deep learning algorithm model;
step three, feature extraction;
step four, compressing the characteristic values;
and step five, warehousing the characteristic values.
Further, the deep learning model is Resnet50, and the deep network is difficult to train because the gradient disappears. Because the gradient propagates back to the previous layer, repeating the multiplication may make the gradient infinitesimal small. The result is that as the layer number of the network is deeper, the performance of the network tends to be saturated and even rapidly decreases, after the residual error is added, the network can be optimized more easily, and a Resnet50 deep learning algorithm model is obtained through training by using a camera to capture food material images as training data.
Further, after the Resnet50 algorithm model is obtained through training, the output of the full connection layer of the Resnet50 deep learning algorithm model is taken as the feature extracted from the food material picture by the whole network, and the feature can contain feature information of each dimension of the original image.
Further, since the image captured by the refrigerator camera contains not only food materials but also a lot of background information, in order to improve the accuracy of food material graph search, PCA (principal components analysis)/principle of principal component analysis is adopted to compress the characteristic values, and the PCA compresses a matrix formed by characteristic vectors corresponding to the characteristic values of the covariance matrix as a coordinate transformation matrix, and then performs coordinate transformation on the characteristic vectors to remove common parts in the image characteristics, so that the background characteristic information in the image characteristics can be well taken out, and the difference of the characteristics between different images can be better reflected.
Further, the extracted image characteristic value is stored in a characteristic value database to be used for image search.
The invention has the beneficial effects that:
the invention provides an image feature extraction method based on deep learning, which solves the problem that the traditional feature extraction algorithm cannot well extract the features of all dimensions of an image. By utilizing the deep learning technology, the characteristic value extraction of the intelligent refrigerator food materials is completed, the problem of insufficient characteristic extraction of the traditional characteristic value extraction method is solved, the characteristics of all dimensions of the image can be extracted by a deep learning algorithm, the accuracy of image searching characteristic value comparison is improved, and the accuracy of refrigerator food material identification is greatly improved.
Drawings
Fig. 1 is a schematic flow chart of a deep learning-based food material feature extraction and compression method;
FIG. 2 is a schematic diagram of a residual error unit in step two;
fig. 3 is a diagram of an image feature extraction and storage structure.
Detailed Description
In order to make the technical solution of the present invention more clear and definite for those skilled in the art, the technical solution of the present invention is described in detail below with reference to the accompanying drawings, but the embodiments of the present invention are not limited thereto.
A deep learning-based food material feature extraction and compression method comprises the steps of
The method comprises the following steps that firstly, a refrigerator camera is configured, so that the process that a user accesses food materials can be clearly shot by the camera;
step two, training a deep learning algorithm model; the deep learning model is Resnet50, and the deep network is difficult to train because the gradient disappears. Because the gradient propagates back to the previous layer, repeating the multiplication may make the gradient infinitesimal small. The result is that as the layer number of the network is deeper, the performance of the network tends to be saturated and even rapidly decreases, after the residual error is added, the network can be optimized more easily, and a Resnet50 deep learning algorithm model is obtained through training by using a camera to capture food material images as training data.
Step three, feature extraction; after the Resnet50 algorithm model is obtained through training, the output of the full connection layer of the Resnet50 deep learning algorithm model is taken as the characteristics extracted from the food material picture by the whole network, and the characteristics can contain characteristic information of each dimension of the original image.
Step four, compressing the characteristic values; the image captured by the refrigerator camera contains not only food materials but also a lot of background information, so in order to improve the accuracy of food material graph search, the feature values are compressed by adopting PCA (principal component analysis)/principal component analysis principle, the PCA compression is carried out by solving a matrix formed by feature vectors corresponding to the feature values of the covariance matrix as a coordinate transformation matrix, then coordinate transformation is carried out on the feature vectors, and the common part in the image features is removed, so that the background feature information in the image features can be well taken out, and the difference of the features among different pictures can be better embodied.
And step five, storing the characteristic values into a database, and storing the extracted image characteristic values into a characteristic value database for image search.
While the invention has been described with respect to the preferred embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (5)
1. A deep learning-based food material feature extraction and compression method is characterized by comprising the following steps: comprises that
The method comprises the following steps that firstly, a refrigerator camera is configured, so that the process that a user accesses food materials can be clearly shot by the camera;
step two, training a deep learning algorithm model;
step three, feature extraction;
step four, compressing the characteristic values;
and step five, warehousing the characteristic values.
2. The food material feature extraction and compression method based on deep learning of claim 1, wherein: the deep learning algorithm model is Resnet50, and is trained by using a camera to capture food material images as training data to obtain a Resnet50 deep learning algorithm model.
3. The food material feature extraction and compression method based on deep learning of claim 2, wherein: after the Resnet50 deep learning algorithm model is obtained through training, the output of the full connection layer of the Resnet50 deep learning algorithm model is taken as the feature extracted from the food material picture by the whole network, and the feature can contain feature information of each dimension of the original image.
4. The food material feature extraction and compression method based on deep learning of claim 3, wherein: and the PCA is adopted to compress the characteristic values, so that the accuracy of searching the food material graph is improved.
5. The food material feature extraction and compression method based on deep learning of claim 4, wherein: and storing the extracted image characteristic value into a characteristic value database for image search.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110078834.4A CN112862762A (en) | 2021-01-21 | 2021-01-21 | Deep learning-based food material feature extraction and compression method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110078834.4A CN112862762A (en) | 2021-01-21 | 2021-01-21 | Deep learning-based food material feature extraction and compression method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112862762A true CN112862762A (en) | 2021-05-28 |
Family
ID=76008428
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110078834.4A Pending CN112862762A (en) | 2021-01-21 | 2021-01-21 | Deep learning-based food material feature extraction and compression method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112862762A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113761942A (en) * | 2021-09-14 | 2021-12-07 | 合众新能源汽车有限公司 | Semantic analysis method and device based on deep learning model and storage medium |
CN114998891A (en) * | 2022-05-17 | 2022-09-02 | 电子科技大学 | Intelligent refrigerator food material accurate detection method based on transfer learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110069643A (en) * | 2019-03-08 | 2019-07-30 | 杭州晶彩数字科技有限公司 | The method and system of similar building picture searching |
CN110647649A (en) * | 2019-09-29 | 2020-01-03 | 腾讯云计算(北京)有限责任公司 | Feature retrieval method, device and storage medium |
CN111503990A (en) * | 2020-04-10 | 2020-08-07 | 海信集团有限公司 | Refrigerator and food material identification method |
CN111797267A (en) * | 2020-07-14 | 2020-10-20 | 西安邮电大学 | Medical image retrieval method and system, electronic device and storage medium |
-
2021
- 2021-01-21 CN CN202110078834.4A patent/CN112862762A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110069643A (en) * | 2019-03-08 | 2019-07-30 | 杭州晶彩数字科技有限公司 | The method and system of similar building picture searching |
CN110647649A (en) * | 2019-09-29 | 2020-01-03 | 腾讯云计算(北京)有限责任公司 | Feature retrieval method, device and storage medium |
CN111503990A (en) * | 2020-04-10 | 2020-08-07 | 海信集团有限公司 | Refrigerator and food material identification method |
CN111797267A (en) * | 2020-07-14 | 2020-10-20 | 西安邮电大学 | Medical image retrieval method and system, electronic device and storage medium |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113761942A (en) * | 2021-09-14 | 2021-12-07 | 合众新能源汽车有限公司 | Semantic analysis method and device based on deep learning model and storage medium |
CN113761942B (en) * | 2021-09-14 | 2023-12-05 | 合众新能源汽车股份有限公司 | Semantic analysis method, device and storage medium based on deep learning model |
CN114998891A (en) * | 2022-05-17 | 2022-09-02 | 电子科技大学 | Intelligent refrigerator food material accurate detection method based on transfer learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Chen et al. | Person re-identification by camera correlation aware feature augmentation | |
Du et al. | Correlation-guided attention for corner detection based visual tracking | |
Zheng et al. | SIFT meets CNN: A decade survey of instance retrieval | |
Huang et al. | A benchmark and comparative study of video-based face recognition on cox face database | |
Nagesh et al. | A compressive sensing approach for expression-invariant face recognition | |
Yu et al. | Stratified pooling based deep convolutional neural networks for human action recognition | |
WO2010144259A1 (en) | Ultra-low dimensional representation for face recognition under varying expressions | |
CN112862762A (en) | Deep learning-based food material feature extraction and compression method | |
US9600738B2 (en) | Discriminative embedding of local color names for object retrieval and classification | |
Matas et al. | Object recognition methods based on transformation covariant features | |
CN111723692B (en) | Near-repetitive video detection method based on label features of convolutional neural network semantic classification | |
CN112308128B (en) | Image matching method based on attention mechanism neural network | |
CN101515286B (en) | Image matching method based on image feature multi-level filtration | |
Liu et al. | Overview of image inpainting and forensic technology | |
CN112131431B (en) | Data processing method, device and computer readable storage medium | |
Huang et al. | Estimating human pose from occluded images | |
Liu et al. | Loop closure detection based on improved hybrid deep learning architecture | |
Tian et al. | Improved bag-of-words model for person re-identification | |
CN111931850A (en) | Cross-modal fusion significance detection method based on feature refinement | |
CN112487994A (en) | Smoke and fire detection method and system, storage medium and terminal | |
Novozámský et al. | Extended IMD2020: a large‐scale annotated dataset tailored for detecting manipulated images | |
Zhu et al. | Statistical image modeling for semantic segmentation | |
Zhao et al. | Robust object tracking via sparse representation based on compressive collaborative Haar-like feature space | |
CN114549938B (en) | Model training method, image information management method, image recognition method and device | |
Xiang et al. | Spatial pyramid pooling based convolutional autoencoder network for loop closure detection |
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: 20210528 |