CN109635690A - The commodity recognition detection method and device of view-based access control model - Google Patents
The commodity recognition detection method and device of view-based access control model Download PDFInfo
- Publication number
- CN109635690A CN109635690A CN201811457416.0A CN201811457416A CN109635690A CN 109635690 A CN109635690 A CN 109635690A CN 201811457416 A CN201811457416 A CN 201811457416A CN 109635690 A CN109635690 A CN 109635690A
- Authority
- CN
- China
- Prior art keywords
- commodity
- product features
- image
- features model
- sample
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- 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
Abstract
This application discloses a kind of commodity recognition detection method and devices of view-based access control model, are related to article identification field, this method includes the image of collecting sample commodity；According to the merchandise news of default sample commodity, the sample commodity are labeled in described image；Described image after mark is subjected to deep learning by CNN neural network, obtains the product features model of the sample commodity；The product features model is sent to default commodity identification terminal；The image of the commodity identification terminal acquisition end article, and image recognition is carried out to the end article according to the product features model, obtain the corresponding merchandise news of the end article.Present application addresses the training process of commodity recognition detection method in the related technology and identification process all on a terminal device, leads to the problem that recognition efficiency is low.
Description
Technical field
Identify field this application involves article, a kind of commodity recognition detection method in particular to view-based access control model and
Device.
Background technique
The training process of the commodity recognition detection method based on machine learning and identification process are all at one in the related technology
On terminal device, do not have a real-time, general training one secondary at least one day or more, the second talent can be carried out knowledge after more new commodity
Not, multiple devices repetition training and then cause at high cost, and if when identifying that more than one piece commodity or commodity block simultaneously, know
Other poor effect.
Therefore, it is badly in need of a kind of commodity recognition detection method and device of view-based access control model, to solve commodity in the related technology
The training process and identification process of recognition detection method all on a terminal device, lead to the problem that recognition efficiency is low.
Summary of the invention
The main purpose of the application is to provide a kind of commodity recognition detection method and device of view-based access control model, to solve phase
The training process and identification process of commodity recognition detection method in the technology of pass lead to recognition effect all on a terminal device
Very poor problem.
To achieve the goals above, according to a first aspect of the present application, the embodiment of the present application provides a kind of view-based access control model
Commodity recognition detection method, which comprises the image of collecting sample commodity；Believed according to the commodity of default sample commodity
Breath, is labeled the sample commodity in described image；Described image after mark is carried out deeply by CNN neural network
Degree study, obtains the product features model of the sample commodity；The product features model is sent to default commodity identification eventually
End；The image of commodity identification terminal acquisition end article, and according to the product features model to the end article into
Row image recognition obtains the corresponding merchandise news of the end article.
With reference to first aspect, the embodiment of the present application provides the first possible embodiment of first aspect, wherein institute
The image for stating collecting sample commodity includes: to carry out image information collecting to the sample commodity by image collecting device, wherein
The placement state of the sample commodity is practical placement state when selling.
With reference to first aspect, the embodiment of the present application provides second of possible embodiment of first aspect, wherein institute
Stating merchandise news includes: commodity material, commodity color, commodity shape, commercial size, Brand, the type of merchandise, commodity class
Not, at least one of product name and illuminance.
With reference to first aspect, the embodiment of the present application provides the third possible embodiment of first aspect, wherein institute
It states after being labeled in the picture to the sample commodity, the described image after mark is carried out deeply by CNN neural network
It include: to be able to carry out the Data Format Transform of the merchandise news of the sample commodity after mark to be corresponding before degree study
The data format of CNN neural metwork training.
With reference to first aspect, the embodiment of the present application provides the 4th kind of possible embodiment of first aspect, wherein institute
It states and the described image after mark is subjected to deep learning by CNN neural network, obtain the product features mould of the sample commodity
Type includes: to carry out RCNN neural metwork training to described image, obtains the error amount of training data and truthful data；To the mistake
Difference carries out correction operation；Judge whether the presently described error amount after correcting reaches predetermined target value；If it is determined that after correcting
Presently described error amount reach predetermined target value, then terminate to train, obtain the product features model of the sample commodity.
With reference to first aspect, the embodiment of the present application provides the 5th kind of possible embodiment of first aspect, wherein institute
Stating product features model being sent to default commodity identification terminal includes: to judge the product features model and the default commodity
Whether current product features model is identical on identification terminal；If the product features model and the default commodity identification are eventually
Current product features model is different on end, then the product features model is sent to the default commodity identification terminal.
With reference to first aspect, the embodiment of the present application provides the 6th kind of possible embodiment of first aspect, wherein institute
The image of commodity identification terminal acquisition end article is stated, and image is carried out to the end article according to the product features model
Identification, obtaining the corresponding merchandise news of the end article includes: before by third party's product features model according to the commodity
The data format of characteristic model carries out format conversion；Third party's product features model after conversion is sent to default commodity
Identification terminal.
With reference to first aspect, the embodiment of the present application provides the 7th kind of possible embodiment of first aspect, wherein institute
The image of commodity identification terminal acquisition end article is stated, and image is carried out to the end article according to the product features model
Identification, obtaining the corresponding merchandise news of the end article includes: the image information for judging the end article and the commodity
Whether the likelihood probability numerical value of characteristic model reaches default recognizable probability numbers；If it is determined that the image of the end article is believed
The likelihood probability numerical value of breath and the product features model reaches default recognizable probability numbers, then obtains the end article
Merchandise news.
To achieve the goals above, according to a second aspect of the present application, the embodiment of the present application provides a kind of view-based access control model
Commodity identification and detection device, comprising: sample commodity image acquisition unit, the image for collecting sample commodity；Sample commodity
Image labeling unit, for being marked to the sample commodity in described image according to the merchandise news for presetting sample commodity
Note；Product features model acquiring unit carries out deep learning by CNN neural network for the described image after marking, obtains
To the product features model of the sample commodity；Product features model transmission unit, for sending the product features model
To default commodity identification terminal；End article recognition unit, for the image of commodity identification terminal acquisition end article, and
Image recognition is carried out to the end article according to the product features model, obtains the corresponding commodity letter of the end article
Breath.
In the embodiment of the present application, by the way of being labeled to the sample commodity in image, after it will mark
Image carries out deep learning by CNN neural network, obtains the product features model of sample commodity, and product features model is sent out
It send to default commodity identification terminal, has reached commodity identification terminal according to product features model and image recognition is carried out to end article
Purpose, thus realize provide commodity recognition efficiency technical effect, and then solve in the related technology commodity identification inspection
The training process and identification process of survey method all on a terminal device, lead to the problem that recognition efficiency is low.
Detailed description of the invention
The attached drawing constituted part of this application is used to provide further understanding of the present application, so that the application's is other
Feature, objects and advantages become more apparent upon.The illustrative examples attached drawing and its explanation of the application is for explaining the application, not
Constitute the improper restriction to the application.In the accompanying drawings:
Fig. 1 is the flow chart of the commodity recognition detection method provided according to the embodiment of the present application one；
Fig. 2 is the flow chart of the commodity recognition detection method provided according to the embodiment of the present application two；
Fig. 3 is the flow chart of the commodity recognition detection method provided according to the embodiment of the present application three；
Fig. 4 is the flow chart of the commodity recognition detection method provided according to the embodiment of the present application four；
Fig. 5 is the flow chart of the commodity recognition detection method provided according to the embodiment of the present application five；
Fig. 6 is the schematic diagram of the commodity identification and detection device provided according to the embodiment of the present application six；And
Fig. 7 is the detailed maps according to herein described product features model acquiring unit 30.
Specific embodiment
In order to make those skilled in the art more fully understand application scheme, below in conjunction in the embodiment of the present application
Attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is only
The embodiment of the application a part, instead of all the embodiments.Based on the embodiment in the application, ordinary skill people
Member's every other embodiment obtained without making creative work, all should belong to the model of the application protection
It encloses.
It should be noted that the description and claims of this application and term " first " in above-mentioned attached drawing, "
Two " etc. be to be used to distinguish similar objects, without being used to describe a particular order or precedence order.It should be understood that using in this way
Data be interchangeable under appropriate circumstances, so as to embodiments herein described herein.In addition, term " includes " and " tool
Have " and their any deformation, it is intended that cover it is non-exclusive include, for example, containing a series of steps or units
Process, method, system, product or equipment those of are not necessarily limited to be clearly listed step or unit, but may include without clear
Other step or units listing to Chu or intrinsic for these process, methods, product or equipment.
In this application, term " on ", "lower", "left", "right", "front", "rear", "top", "bottom", "inner", "outside",
" in ", "vertical", "horizontal", " transverse direction ", the orientation or positional relationship of the instructions such as " longitudinal direction " be orientation based on the figure or
Positional relationship.These terms are not intended to limit indicated dress primarily to better describe the application and embodiment
Set, element or component must have particular orientation, or constructed and operated with particular orientation.
Also, above-mentioned part term is other than it can be used to indicate that orientation or positional relationship, it is also possible to for indicating it
His meaning, such as term " on " also are likely used for indicating certain relations of dependence or connection relationship in some cases.For ability
For the those of ordinary skill of domain, the concrete meaning of these terms in this application can be understood as the case may be.
In addition, term " installation ", " setting ", " being equipped with ", " connection ", " connected ", " socket " shall be understood in a broad sense.For example,
It may be a fixed connection, be detachably connected or monolithic construction；It can be mechanical connection, or electrical connection；It can be direct phase
It even, or indirectly connected through an intermediary, or is two connections internal between device, element or component.
For those of ordinary skills, the concrete meaning of above-mentioned term in this application can be understood as the case may be.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Consider: the training process of the commodity recognition detection method based on machine learning and identification process be all in the related technology
It is not only at high cost on a terminal device, and if recognition effect is very when identifying that more than one piece commodity or commodity block simultaneously
Difference, therefore, this application provides a kind of commodity recognition detection method and devices of view-based access control model, to solve quotient in the related technology
The training process and identification process of product recognition detection method lead to the problem that recognition efficiency is low all on a terminal device.
As shown in Figure 1, this method includes the following steps, namely S101 to step S105:
Step S101, the image of collecting sample commodity；
Preferably, by image collecting device to the sample commodity carry out Image Acquisition, the sample commodity be for
The commodity to restocking sold, described image acquisition device can be high-definition camera, or other can acquire image
Device.
Step S102 marks the sample commodity in described image according to the merchandise news of default sample commodity
Note；
Preferably, customized in advance to the merchandise news progress of every sample commodity in system database, specifically, from
Definition the merchandise news include: commodity material, commodity color, commodity shape, commercial size, Brand, the type of merchandise,
At least one of merchandise classification, product name and illuminance, in the other embodiments of the application, the merchandise news can also
Think other information related with commodity；By artificial mode by the position of sample commodity in the image of the sample commodity
It marks off and, and mark the merchandise news of the upper sample commodity.
Described image after mark is carried out deep learning by CNN neural network, obtains the sample quotient by step S103
The product features model of product；
Preferably, will divide has the image of the sample product locations, the merchandise news for being labeled with the sample commodity to make
For training data, RCNN neural metwork training is carried out, the product features model of the sample commodity is obtained.
The product features model is sent to default commodity identification terminal by step S104；
Preferably, judge product features model current on the product features model and the default commodity identification terminal
It is whether identical, if the product features model is different from product features model current on the default commodity identification terminal,
The product features model is then sent to the default commodity identification terminal, specifically, judge product features model whether phase
It is same can also to use other manner of comparison by the way of comparing two product features model data sizes, it is described default
Commodity identification terminal can take stock device for commodity cashier and hand-held commodity, or other are for carrying out commodity identification
Device.
Step S105, the image of the commodity identification terminal acquisition end article, and according to the product features model pair
The end article carries out image recognition, obtains the corresponding merchandise news of the end article.
Preferably, when the commodity identification terminal identifies end article, pass through image acquisition device first
The image of the end article, the product features model then contained according to itself carry out image to the image of the end article
Identification, obtains the corresponding merchandise news of the end article, and the end article is carried out position division in described image
It is shown with information.
Embodiment one:
By taking laughable commodity identification as an example, firstly, the image for all colas sold is wanted in acquisition, for example pass through camera
Image Acquisition is carried out to " Coca-Cola ", and pre-defines the merchandise news of Coca-Cola in systems, such as " palatable can
Happy, 330ml, world cup commemorates 01 section of version, and red outer packing, cylindrical bottle body, high 8cm, radius 3cm, illumination are sufficient ", and lead to
It crosses artificial mode to select Coca-Cola frame in Coca-Cola image, there is Coca-Cola position, frame to select division
Training data as RCNN neural network of top left co-ordinate and the picture of bottom right angular coordinate and the merchandise news of Coca-Cola
Deep learning training is carried out, the product features model of Coca-Cola is obtained, then, by the product features model of described Coca-Cola
It is sent to commodity cashier, finally, the commodity cashier acquires end article when carrying out commodity clearing, through camera
Image, the product features model of the Coca-Cola then contained according to itself judge whether the end article is described palatable
Cola, if it is, the merchandise news for extracting described Coca-Cola is settled accounts for commodity.
It can be seen from the above description that the present invention realizes following technical effect:
In the embodiment of the present application, by the way of being labeled to the sample commodity in image, after it will mark
Image carries out deep learning by CNN neural network, obtains the product features model of sample commodity, and product features model is sent out
It send to default commodity identification terminal, has reached commodity identification terminal according to product features model and image recognition is carried out to end article
Purpose, thus realize provide commodity recognition efficiency technical effect, and then solve in the related technology commodity identification inspection
The training process and identification process of survey method all on a terminal device, lead to the problem that recognition efficiency is low.
According to embodiments of the present invention, as preferred in the embodiment of the present application, the image of the collecting sample commodity includes:
Image information collecting is carried out to the sample commodity by image collecting device, wherein the placement state of the sample commodity is
Practical placement state when selling.
Preferably, when carrying out Image Acquisition to the sample commodity, the placement state of the sample commodity is when selling
Practical placement state, comprising being collected by practical placement state by different circumstance of occlusion, different shooting angle
Real image compactness of the image and end article of sample commodity when identifying clearing is higher, using by practical placement state
Training data of the image of the sample commodity collected as RCNN neural network, the resolution of obtained product features model
It is higher.
According to embodiments of the present invention, as preferred in the embodiment of the present application, the merchandise news include: commodity material,
In commodity color, commodity shape, commercial size, Brand, the type of merchandise, merchandise classification, product name and illuminance extremely
Few one kind.
Preferably, in the other embodiments of the application, the merchandise news may be other letters related with commodity
Breath.
According to embodiments of the present invention, described in the picture to the sample commodity as preferred in the embodiment of the present application
After being labeled, including: before by CNN neural network progress deep learning by the described image after mark will be after mark
The Data Format Transform of the merchandise news of the sample commodity is the corresponding data lattice for being able to carry out CNN neural metwork training
Formula.
Preferably, when carrying out RCNN neural metwork training, there are strict requirements to the data format of training data, therefore
It will be the data format of RCNN neural metwork training requirement by the Data Format Transform of the customized merchandise news of system.
According to embodiments of the present invention, as preferred in the embodiment of the present application, as shown in Fig. 2, the institute by after mark
It states image and deep learning is carried out by CNN neural network, the product features model for obtaining the sample commodity includes following step
Rapid S201 to step S204:
Step S201 carries out RCNN neural metwork training to described image, obtains the error of training data and truthful data
Value；
Preferably, RCNN neural metwork training is carried out to described image, obtains predicted value, and calculate predicted value and mark
Infuse the error amount of the true value of data.
Step S202 carries out correction operation to the error amount；
Preferably, the error amount is the calculated result of loss function, the model calculated by loss function it is pre-
The inconsistent degree of measured value and true value optimizes RCNN according to the parameter of the loss function after the error amount, and optimization
Neural network parameter, and being trained again, the error between the continuous data corrected and marked obtain new closer
It in the error amount of true value, is corrected again, 30,000 error correction behaviour is probably carried out in the CNN neural network of the application
Make.
Step S203, judges whether the presently described error amount after correcting reaches predetermined target value；
Preferably, predetermined target value is equivalent to the value i.e. true value of markup information, is equal to or converges to true value, in advance
Time value is the equal of the degree of closeness of true value, that is, the size of error amount.
Step S204, if it is decided that the presently described error amount after correction reaches predetermined target value, then terminates to train, obtain
The product features model of the sample commodity.
Preferably, if it is decided that the current error value after error correction operates within a preset range, then terminates
Neural metwork training, and the product features model of the sample commodity is obtained according to training result.
According to embodiments of the present invention, as preferred in the embodiment of the present application, as shown in figure 3, described by product features mould
Type is sent to default commodity identification terminal and includes the following steps, namely S301 to step S302:
Step S301 judges product features mould current on the product features model and the default commodity identification terminal
Whether type is identical；
Preferably, judge in the product features model obtained by RCNN neural metwork training and default commodity identification terminal
Current commodity characteristic model it is whether identical, the judgment mode can be using comparing two product features model data sizes
Mode can also use other manner of comparison, and the default commodity identification terminal can be commodity cashier and hand-held commodity disk
Goods device, or other are used to carry out the device of commodity identification.
Step S302, if product features mould current on the product features model and the default commodity identification terminal
Type is different, then the product features model is sent to the default commodity identification terminal.
Preferably, if product features model current on the product features model and the default commodity identification terminal
Difference then determines that the product features model in the default commodity identification terminal is not latest edition, will pass through RCNN nerve net
The product features model that network training obtains is sent to the commodity identification terminal, and replaces wherein already contg product features mould
Type.
According to embodiments of the present invention, as preferred in the embodiment of the present application, as shown in figure 4, the commodity identification terminal
The image of end article is acquired, and image recognition is carried out to the end article according to the product features model, is obtained described
Include the following steps, namely S401 to step S402 before the corresponding merchandise news of end article:
Third party's product features model is carried out format according to the data format of the product features model and turned by step S401
Change；
Preferably, the product features model obtained after RCNN neural metwork training has fixed data format, can
Commodity identification, but third party's product features are carried out so that third party's product features model is sent to default commodity identification terminal
Model has to comply with defined data format, therefore by third party's product features model according to the data of the product features model
Format carries out format conversion.
Third party's product features model after conversion is sent to default commodity identification terminal by step S402.
Preferably, third party's product features model after Data Format Transform default commodity are sent to identify
Terminal is for carrying out commodity identification.
According to embodiments of the present invention, as preferred in the embodiment of the present application, as shown in figure 5, the commodity identification terminal
The image of end article is acquired, and image recognition is carried out to the end article according to the product features model, is obtained described
The corresponding merchandise news of end article includes the following steps, namely S501 to step S502:
The likelihood probability numerical value of step S501, the image information and the product features model that judge the end article is
It is no to reach default recognizable probability numbers；
Preferably, after the default commodity identification terminal collects the image information of the end article, contain with itself
The product features model carry out similarity comparison, obtain likelihood probability numerical value, judge the likelihood probability numerical value and default
It can recognize the size relation of probability numbers.
Step S502, if it is decided that the likelihood probability number of the image information of the end article and the product features model
Value reaches default recognizable probability numbers, then obtains the merchandise news of the end article.
Preferably, if being greater than default recognizable probability numbers by the likelihood probability numerical value that comparison obtains, judgement can
To carry out commodity identification, and obtain the merchandise news of the end article.
It should be noted that step shown in the flowchart of the accompanying drawings can be in such as a group of computer-executable instructions
It is executed in computer system, although also, logical order is shown in flow charts, and it in some cases, can be with not
The sequence being same as herein executes shown or described step.
According to embodiments of the present invention, it additionally provides a kind of for implementing the commodity recognition detection method of above-mentioned view-based access control model
Device, as shown in fig. 6, the device includes: sample commodity image acquisition unit 10, the image for collecting sample commodity；Sample
Commodity image marks unit 20, for the merchandise news according to default sample commodity, to the sample commodity in described image
It is labeled；Product features model acquiring unit 30 carries out depth by CNN neural network for the described image after marking
Study, obtains the product features model of the sample commodity；Product features model transmission unit 40 is used for the product features
Model is sent to default commodity identification terminal；End article recognition unit 50 acquires target quotient for the commodity identification terminal
The image of product, and image recognition is carried out to the end article according to the product features model, obtain the end article pair
The merchandise news answered.
The image of collecting sample commodity is used for according to sample commodity image acquisition unit 10 described herein, it is preferred that
Image Acquisition is carried out to the sample commodity by image collecting device, the sample commodity are the quotient to restocking for selling
Product, described image acquisition device can be high-definition camera, or other can acquire the device of image.
Unit 20 is marked according to sample commodity image described herein to be used for according to the merchandise news for presetting sample commodity,
The sample commodity are labeled in described image, it is preferred that the commodity of every sample commodity in system database
Information progress is customized in advance, specifically, the customized merchandise news includes: commodity material, commodity color, commodity shape
At least one of shape, commercial size, Brand, the type of merchandise, merchandise classification, product name and illuminance, in the application
Other embodiments in, the merchandise news may be other information related with commodity；In the image of the sample commodity
In the position of sample commodity is marked off and to mark the merchandise news of the upper sample commodity by artificial mode.
Described image after being used to mark according to product features model acquiring unit 30 described herein passes through CNN mind
Deep learning is carried out through network, obtains the product features model of the sample commodity, it is preferred that division there are into the sample commodity
Position, be labeled with the sample commodity merchandise news image as training data, carry out RCNN neural metwork training, obtain
The product features model of the sample commodity.
It is used to for the product features model being sent to according to product features model transmission unit 40 described herein pre-
If commodity identification terminal, it is preferred that judge commodity current on the product features model and the default commodity identification terminal
Whether characteristic model is identical, if product features mould current on the product features model and the default commodity identification terminal
Type is different, then the product features model is sent to the default commodity identification terminal, specifically, judging product features model
Whether it is identical can by the way of comparing two product features model data sizes, can also use other manner of comparison, institute
Stating default commodity identification terminal can take stock device for commodity cashier and hand-held commodity, or other are for carrying out commodity
The device of identification.
According to end article recognition unit 50 described herein for commodity identification terminal acquisition end article
Image, and image recognition is carried out to the end article according to the product features model, it is corresponding to obtain the end article
Merchandise news, it is preferred that when the commodity identification terminal identifies end article, pass through image acquisition device first
The image of the end article, the product features model then contained according to itself carry out image to the image of the end article
Identification, obtains the corresponding merchandise news of the end article, and the end article is carried out position division in described image
It is shown with information.
According to embodiments of the present invention, as preferred in the embodiment of the present application, as shown in fig. 7, the product features model
Acquiring unit 30 includes: that training error value obtains module 31, for carrying out RCNN neural metwork training to described image, is instructed
Practice the error amount of data and truthful data；Training error value corrects module 32, for carrying out correction operation to the error amount；Instruction
Practice error amount judgment module 33, for judging whether the presently described error amount after correcting reaches predetermined target value；Characteristic model
Obtain module 34 then terminates to train if it is determined that the presently described error amount for after correcting reaches predetermined target value, obtains institute
State the product features model of sample commodity.
Module 31 is obtained according to training error value described herein to be used to carry out RCNN neural network instruction to described image
Practice, obtain the error amount of training data and truthful data, it is preferred that RCNN neural metwork training is carried out to described image, is obtained
Predicted value, and calculate the error amount of the true value of predicted value and labeled data.
Module 32 is corrected for carrying out correction operation to the error amount, preferably according to training error value described herein
, the error amount is loss function, the predicted value of the model calculated by loss function and the inconsistent journey of true value
Degree is trained again according to the parameter of the loss function after the error amount, and optimization, continuous to correct and mark
Data between error, obtain the new error amount closer to true value, corrected again, the application CNN mind
30,000 error correction operations are probably carried out in network.
It is used to judge that the presently described error amount after correction to be according to training error value judgment module 33 described herein
It is no to reach predetermined target value, it is preferred that predetermined target value is equivalent to the value i.e. true value of markup information, is equal to or infinitely connects
It is bordering on true value, predetermined target value is the equal of the degree of closeness of true value, that is, the size of error amount.
If it is determined that obtaining the presently described error amount after module 34 is used to correct according to characteristic model described herein
Reach predetermined target value, then terminate to train, obtains the product features model of the sample commodity, it is preferred that if it is determined that by
Current error value after error correction operation within a preset range, then terminates neural metwork training, and according to training result
Obtain the product features model of the sample commodity.
Obviously, those skilled in the art should be understood that each module of the above invention or each step can be with general
Computing device realize that they can be concentrated on a single computing device, or be distributed in multiple computing devices and formed
Network on, optionally, they can be realized with the program code that computing device can perform, it is thus possible to which they are stored
Be performed by computing device in the storage device, perhaps they are fabricated to each integrated circuit modules or by they
In multiple modules or step be fabricated to single integrated circuit module to realize.In this way, the present invention is not limited to any specific
Hardware and software combines.
The foregoing is merely preferred embodiment of the present application, are not intended to limit this application, for the skill of this field
For art personnel, various changes and changes are possible in this application.Within the spirit and principles of this application, made any to repair
Change, equivalent replacement, improvement etc., should be included within the scope of protection of this application.
Claims (10)
1. a kind of commodity recognition detection method of view-based access control model, which is characterized in that the described method includes:
The image of collecting sample commodity；
According to the merchandise news of default sample commodity, the sample commodity are labeled in described image；
Described image after mark is subjected to deep learning by CNN neural network, obtains the product features of the sample commodity
Model；
The product features model is sent to default commodity identification terminal；And
The image of commodity identification terminal acquisition end article, and according to the product features model to the end article into
Row image recognition obtains the corresponding merchandise news of the end article.
2. commodity recognition detection method according to claim 1, which is characterized in that the image packet of the collecting sample commodity
It includes:
Image information collecting is carried out to the sample commodity by image collecting device, wherein the sample commodity put shape
State is practical placement state when selling.
3. commodity recognition detection method according to claim 1, which is characterized in that the merchandise news includes: merchantable timber
In matter, commodity color, commodity shape, commercial size, Brand, the type of merchandise, merchandise classification, product name and illuminance
It is at least one.
4. commodity recognition detection method according to claim 1, which is characterized in that described in the picture to the sample quotient
After product are labeled, the described image after mark, which is carried out deep learning before by CNN neural network, includes:
The Data Format Transform of the merchandise news of the sample commodity after mark is able to carry out CNN neural network to be corresponding
Trained data format.
5. commodity recognition detection method according to claim 1, which is characterized in that the described image by after mark is led to
It crosses CNN neural network and carries out deep learning, the product features model for obtaining the sample commodity includes:
RCNN neural metwork training is carried out to described image, obtains the error of the error amount of training data and truthful data data
Value；
Correction operation is carried out to the error amount；
Judge whether the presently described error amount after correcting reaches predetermined target value；And
If it is determined that the presently described error amount after correcting reaches predetermined target value, then terminates to train, obtain the sample commodity
Product features model.
6. commodity recognition detection method according to claim 1, which is characterized in that described to be sent to product features model
Presetting commodity identification terminal includes:
Judge whether the product features model and product features model current on the default commodity identification terminal are identical；
If the product features model is different from product features model current on the default commodity identification terminal, by institute
It states product features model and is sent to the default commodity identification terminal.
7. commodity recognition detection method according to claim 1, which is characterized in that the commodity identification terminal acquires target
The image of commodity, and image recognition is carried out to the end article according to the product features model, obtain the end article
Include: before corresponding merchandise news
Third party's product features model is subjected to format conversion according to the data format of the product features model；
Third party's product features model after conversion is sent to default commodity identification terminal.
8. commodity recognition detection method according to claim 1, which is characterized in that the commodity identification terminal acquires target
The image of commodity, and image recognition is carried out to the end article according to the product features model, obtain the end article
Corresponding merchandise news includes:
The likelihood probability numerical value of the image information that judges the end article and the product features model whether reach it is default can
Identification probability numerical value；
If it is determined that the likelihood probability numerical value of the image information of the end article and the product features model reach it is default can
Identification probability numerical value then obtains the merchandise news of the end article.
9. a kind of commodity identification and detection device of view-based access control model characterized by comprising
Sample commodity image acquisition unit, the image for collecting sample commodity；
Sample commodity image marks unit, for the merchandise news according to default sample commodity, to the sample in described image
This commodity is labeled；
Product features model acquiring unit carries out deep learning by CNN neural network for the described image after marking, obtains
To the product features model of the sample commodity；
Product features model transmission unit, for the product features model to be sent to default commodity identification terminal；And
End article recognition unit, for the image of commodity identification terminal acquisition end article, and according to the commodity spy
It levies model and image recognition is carried out to the end article, obtain the corresponding merchandise news of the end article.
10. commodity identification and detection device according to claim 9, which is characterized in that the product features model obtains single
Member includes:
Training error value obtains module, for carrying out RCNN neural metwork training to described image, obtains training data and true
The error amount of data；
Training error value corrects module, for carrying out correction operation to the error amount；
Training error value judgment module, for judging whether the presently described error amount after correcting reaches predetermined target value；And
Characteristic model obtains module then to be terminated if it is determined that the presently described error amount for after correcting reaches predetermined target value
Training, obtains the product features model of the sample commodity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811457416.0A CN109635690A (en) | 2018-11-30 | 2018-11-30 | The commodity recognition detection method and device of view-based access control model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811457416.0A CN109635690A (en) | 2018-11-30 | 2018-11-30 | The commodity recognition detection method and device of view-based access control model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109635690A true CN109635690A (en) | 2019-04-16 |
Family
ID=66070659
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811457416.0A Pending CN109635690A (en) | 2018-11-30 | 2018-11-30 | The commodity recognition detection method and device of view-based access control model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109635690A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110321797A (en) * | 2019-05-31 | 2019-10-11 | 苏宁云计算有限公司 | Commodity recognition method and device |
| CN110942035A (en) * | 2019-11-28 | 2020-03-31 | 浙江由由科技有限公司 | Method, system, device and storage medium for acquiring commodity information |
| CN111079699A (en) * | 2019-12-30 | 2020-04-28 | 北京每日优鲜电子商务有限公司 | Commodity identification method and device |
| CN111209911A (en) * | 2020-01-07 | 2020-05-29 | 创新奇智(合肥)科技有限公司 | Custom tag identification system and identification method based on semantic segmentation network |
| CN111242167A (en) * | 2019-12-30 | 2020-06-05 | 深圳奇迹智慧网络有限公司 | Distributed image annotation method and device, computer equipment and storage medium |
| CN111460888A (en) * | 2020-02-27 | 2020-07-28 | 平安科技（深圳）有限公司 | Article identification method and device based on machine learning |
| CN111626150A (en) * | 2020-05-11 | 2020-09-04 | 广东顺德工业设计研究院(广东顺德创新设计研究院) | Commodity identification method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108491799A (en) * | 2018-03-23 | 2018-09-04 | 海深科技（宁波）有限公司 | A kind of intelligent sales counter merchandise control method and system based on image recognition |
| CN108596137A (en) * | 2018-05-02 | 2018-09-28 | 济南浪潮高新科技投资发展有限公司 | A kind of commodity scanning record method based on image recognition algorithm |
| CN108875602A (en) * | 2018-05-31 | 2018-11-23 | 珠海亿智电子科技有限公司 | Monitor the face identification method based on deep learning under environment |
-
2018
- 2018-11-30 CN CN201811457416.0A patent/CN109635690A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108491799A (en) * | 2018-03-23 | 2018-09-04 | 海深科技（宁波）有限公司 | A kind of intelligent sales counter merchandise control method and system based on image recognition |
| CN108596137A (en) * | 2018-05-02 | 2018-09-28 | 济南浪潮高新科技投资发展有限公司 | A kind of commodity scanning record method based on image recognition algorithm |
| CN108875602A (en) * | 2018-05-31 | 2018-11-23 | 珠海亿智电子科技有限公司 | Monitor the face identification method based on deep learning under environment |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110321797A (en) * | 2019-05-31 | 2019-10-11 | 苏宁云计算有限公司 | Commodity recognition method and device |
| CN110942035A (en) * | 2019-11-28 | 2020-03-31 | 浙江由由科技有限公司 | Method, system, device and storage medium for acquiring commodity information |
| CN111079699A (en) * | 2019-12-30 | 2020-04-28 | 北京每日优鲜电子商务有限公司 | Commodity identification method and device |
| CN111242167A (en) * | 2019-12-30 | 2020-06-05 | 深圳奇迹智慧网络有限公司 | Distributed image annotation method and device, computer equipment and storage medium |
| CN111242167B (en) * | 2019-12-30 | 2023-11-14 | 深圳奇迹智慧网络有限公司 | Distributed image labeling method, device, computer equipment and storage medium |
| CN111209911A (en) * | 2020-01-07 | 2020-05-29 | 创新奇智(合肥)科技有限公司 | Custom tag identification system and identification method based on semantic segmentation network |
| CN111460888A (en) * | 2020-02-27 | 2020-07-28 | 平安科技（深圳）有限公司 | Article identification method and device based on machine learning |
| CN111626150A (en) * | 2020-05-11 | 2020-09-04 | 广东顺德工业设计研究院(广东顺德创新设计研究院) | Commodity identification method |
| CN111626150B (en) * | 2020-05-11 | 2023-08-18 | 广东顺德工业设计研究院(广东顺德创新设计研究院) | Commodity identification method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109635690A (en) | The commodity recognition detection method and device of view-based access control model | |
| CN104506857B (en) | A kind of camera position deviation detection method and apparatus | |
| US6845296B2 (en) | Object taking out apparatus | |
| CN108388641B (en) | Traffic facility map generation method and system based on deep learning | |
| CN108647553A (en) | Rapid expansion method, system, equipment and the storage medium of model training image | |
| CN109815932A (en) | A kind of paper corrects method, apparatus, electronic equipment and storage medium | |
| CN105678322A (en) | Sample labeling method and apparatus | |
| US9839941B2 (en) | System and method for facilitating manual sorting of objects | |
| CN111738036B (en) | Image processing method, device, equipment and storage medium | |
| CN104392232A (en) | Method for correcting video metadata in traffic scenes | |
| CN109741551A (en) | A kind of commodity identification settlement method, apparatus and system | |
| CN106372613A (en) | Test paper statistical method and apparatus | |
| CN110569856A (en) | sample labeling method and device, and damage category identification method and device | |
| CN110321837A (en) | A kind of recognition methods, device, terminal and the storage medium of examination question score | |
| CN109215075B (en) | Positioning and identifying system and method for workpiece in material grabbing of industrial robot | |
| CN108596115A (en) | A kind of vehicle checking method, apparatus and system based on convolutional neural networks | |
| CN113095338B (en) | Automatic labeling method and device for industrial product image, electronic equipment and storage medium | |
| JP2007003636A (en) | Teaching material processing apparatus, teaching material processing method, and teaching material processing program | |
| CN109063855A (en) | Aid in treatment method and device for the identification of vehicle vehicle condition | |
| KR102259451B1 (en) | Device and Method For Providing Map Information | |
| CN114723977A (en) | Stable feature point identification method for visual SLAM system | |
| CN113516705A (en) | Method and device for calibrating key points of hands | |
| US11010903B1 (en) | Computer vision and machine learning techniques for item tracking | |
| JP5975484B2 (en) | Image processing device | |
| Hile et al. | Microbiology tray and pipette tracking as a proactive tangible user interface |
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: 20190416 |
|
| RJ01 | Rejection of invention patent application after publication |