CN109948736A - Commodity identification model active training method, system, equipment and storage medium - Google Patents
Commodity identification model active training method, system, equipment and storage medium Download PDFInfo
- Publication number
- CN109948736A CN109948736A CN201910269443.3A CN201910269443A CN109948736A CN 109948736 A CN109948736 A CN 109948736A CN 201910269443 A CN201910269443 A CN 201910269443A CN 109948736 A CN109948736 A CN 109948736A
- Authority
- CN
- China
- Prior art keywords
- commodity
- image
- identification model
- magnitude
- recognition accuracy
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The present invention provides a kind of commodity identification model active training method, system, equipment and storage mediums, include the following steps: the commodity image for obtaining first order of magnitude that an at least camera acquires a target area, it is identified by commodity image of the preset commodity identification model to first order of magnitude, generates commodity and identify data;The commodity image of first order of magnitude is sent to backstage labeling system, and receives the mark commodity image that the backstage labeling system returns;Commodity identification data are compared with the commodity image that marked, determine the recognition accuracy of each commodity in the commodity image, when the recognition accuracy is greater than the recognition accuracy threshold value pre-seted, then end step, when the recognition accuracy is less than the recognition accuracy threshold value pre-seted, then the commodity identification model is trained again.The present invention can be avoided the lower problem of commodity identification model recognition accuracy in being actually used in environment.
Description
Technical field
The present invention relates to new retail domains, and in particular, to a kind of commodity identification model active training method, is set system
Standby and storage medium.
Background technique
There is the refrigerator of many commercial articles vendings in market, carry out the automatic vending of commodity, such as carries out selling for beverage.In order to
The sales volume of commodity is improved, many suppliers place refrigerator in many markets and carry out selling for commodity, such as in many markets all
It can be seen that laughable refrigerator is sold by Coca-Cola.
When commodity need to track the commodity display case and merchandise sales situation of each refrigerator, need to pay great labor
Power cost needs to carry out the acquisition of data in face of the person of sending someone to each refrigerator.But in refrigerator the condition of sales of commodity be with
Shi Bianhua's, without the supplement of progress commodity after selling preferable commodity and selling off, it will strong influence to the commodity
Total sales volume.
In addition, each refrigerator probably can all have five, six layers or so of laying rack, puts the commodity in middle position and be easier quilt
Consumer sees, so as to cause the buying behavior of consumer.Therefore, supplier can carry out commodity in refrigerator according to sales target
Put, but refrigerator is generally all placed in market and is managed by market, and market is possible to can not be by the arrangement of supplier
Putting for commodity is carried out, the marketing plan of supplier is influenced.
It will can be most so if the display case of commodity and merchandise sales situation can be understood in each refrigerator in real time
The maximum sales potential of the big each refrigerator of performance, realizes profit bigizationner of single refrigerator.Some Enterprises pass through in refrigerator
Camera is installed to carry out in refrigerator after the acquisition of commodity image, by commodity identification model to the commodity area in commodity image on door
Domain is identified, realizes the effective monitoring to commodity in refrigerator.But the position range that refrigerator is placed is wider, in refrigerator doors
Camera acquisition photo there are the difference in brightness, visual angle and product category so that the commodity identification model of pre-training
Apply the problem that recognition accuracy is lower on some refrigerators.
Summary of the invention
For the defects in the prior art, the object of the present invention is to provide a kind of commodity identification model active training method,
System, equipment and storage medium.
The commodity identification model active training method provided according to the present invention, includes the following steps:
Step S1: the commodity image for first order of magnitude that an at least camera acquires a target area is obtained, by pre-
If commodity identification model the commodity image of first order of magnitude is identified, generate commodity identify data;
Step S2: being sent to backstage labeling system for the commodity image of first order of magnitude, and receives the rear logo
The mark commodity image that injection system returns;
Step S3: commodity identification data are compared with the commodity image that marked, determine the commodity figure
The recognition accuracy of each commodity as in then terminates when the recognition accuracy is greater than the recognition accuracy threshold value pre-seted
Step then carries out again the commodity identification model when the recognition accuracy is less than the recognition accuracy threshold value pre-seted
Secondary training.
Preferably, when train again to the commodity identification model, include the following steps:
Step M1: the commodity image of second order of magnitude of the camera acquisition is obtained, by the quotient of second order of magnitude
Product image is sent to backstage labeling system;
Step M2: the training returned after the backstage labeling system marks the commodity image of second order of magnitude is received
Image set;
Step M3: the commodity identification model is trained again by the training image collection.
Preferably, first order of magnitude is less than second order of magnitude.
Preferably, include the following steps: when the commodity identification model identifies commodity image
Step N1: by the angle end identification model that pre-sets at least two stratoses of intelligent refrigerator in the commodity image
Stratose frame angle end identified, at least identify four be located at the commodity image both side ends stratose frame angle petiolarea
Domain;
Step N2: according to the reference layer at least four stratose frame angle end regions and the reference commodity image pre-seted
Positional relationship between the end regions of column frame angle generates homography matrix;
Step N3: commodity image after correction generation is corrected is carried out to the commodity image according to the homography matrix;
Step N4: identifying commodity image after the correction by the commodity identification model pre-seted, to know
It Chu not the corresponding goods number pre-seted in each commodity region in the commodity image.
Preferably, the step S1 includes the following steps:
Step N101: obtaining multiple training images for the identification of angle end, to each training for the identification of angle end
The mark of image progress angle end regions;
Step N102: angle end identification model is established using the training image of mark angle of departure end regions;
Step N103: commodity image input angle end identification model is subjected to the identification of angle end regions.
Preferably, the step S2 includes the following steps:
Step N201: establishing two-dimensional coordinate system, determines that at least four with reference in commodity image refer to stratose frame angle
The coordinate of end regions;
Step N202: in the two-dimensional coordinate system, the coordinate of four stratose framves angle end regions is determined;
Step N203: according between four stratose frame angle end regions and corresponding reference stratose frame angle end regions
Positional relationship generate the homography matrix.
Preferably, the step S4 includes the following steps:
Step N401: obtaining multiple training images for commodity region recognition, to each described for the knowledge of commodity region
Other training image carries out the mark in commodity region;
Step N402: the commodity region recognition model is established using the training image for marking out commodity region;
Step N403: the commodity image is inputted into the commodity region recognition model and carries out commodity region recognition.
Commodity identification model active training system provided by the invention is actively instructed for realizing the commodity identification model
Practice method, comprising:
Commodity identify data generation module, first order of magnitude acquired for obtaining an at least camera to a target area
Commodity image, identified by commodity image of the preset commodity identification model to first order of magnitude, generate commodity
Identify data;
Image generation module is marked, for the commodity image of first order of magnitude to be sent to backstage labeling system, and
Receive the mark commodity image that the backstage labeling system returns;
Recognition accuracy judgment module, for comparing commodity identification data with the commodity image that marked
It is right, the recognition accuracy of each commodity in the commodity image is determined, when the recognition accuracy is greater than the identification standard pre-seted
When true rate threshold value, then end step, when the recognition accuracy is less than the recognition accuracy threshold value pre-seted, then to the quotient
Product identification model is trained again.
Commodity identification model active training equipment provided by the invention, comprising:
Processor;
Memory, wherein being stored with the executable instruction of the processor;
Wherein, the processor is configured to execute the commodity identification model via the executable instruction is executed actively
The step of training method.
Computer readable storage medium provided by the invention, for storing program, described program is performed described in realization
The step of commodity identification model active training method.
Compared with prior art, the present invention have it is following the utility model has the advantages that
The present invention can be applied in the terminal data acquisition device being monitored to the commodity in refrigerator, when by terminal data
Acquisition device is installed in refrigerator doors, can will be on terminal data acquisition device when being monitored to the commodity in refrigerator
The commodity image of first order of magnitude of camera acquisition is sent to backstage labeling system, and is returned according to backstage labeling system
Mark commodity image the recognition accuracy of commodity identification model is judged in time, in order to the commodity identification model into
Row training, avoids the lower problem of commodity identification model recognition accuracy in being actually used in environment.
Detailed description of the invention
Upon reading the detailed description of non-limiting embodiments with reference to the following drawings, other feature of the invention,
Objects and advantages will become more apparent upon:
Fig. 1 is the step flow chart of commodity identification model active training method in the present invention;
Fig. 2 is to carry out step flow chart trained again in the present invention to commodity identification model;
Fig. 3 is the step flow chart that commodity identification model identifies commodity image in the present invention;
Fig. 4 is the step flow chart of training angle end identification model in the present invention;
Fig. 5 is the step flow chart of homography matrix in the present invention;
Fig. 6 is the step flow chart of the training commodity region recognition model in the present invention;
Fig. 7 is the module diagram of commodity identification model active training system in the present invention;
Fig. 8 is the structural schematic diagram of commodity identification model active training equipment in the present invention;And
Fig. 9 is the structural schematic diagram of computer readable storage medium in the present invention.
Specific embodiment
The present invention is described in detail combined with specific embodiments below.Following embodiment will be helpful to the technology of this field
Personnel further understand the present invention, but the invention is not limited in any way.It should be pointed out that the ordinary skill of this field
For personnel, without departing from the inventive concept of the premise, various modifications and improvements can be made.These belong to the present invention
Protection scope.
In the present embodiment, Fig. 1 is the step flow chart of commodity identification model active training method in the present invention, such as Fig. 1
Shown, commodity identification model active training method provided by the invention, the commodity identification model is applied to in intelligent refrigerator
The collected commodity image of commodity identified that the intelligent refrigerator is provided with the refrigerator doors that can be opened and closed, the refrigerator doors
On be at least arranged one for in refrigerator commodity carry out commodity image acquisition camera, include the following steps:
Step S1: the commodity image for first order of magnitude that an at least camera acquires a target area is obtained, by pre-
If commodity identification model the commodity image of first order of magnitude is identified, generate commodity identify data;
Step S2: being sent to backstage labeling system for the commodity image of first order of magnitude, and receives the rear logo
The mark commodity image that injection system returns;
Step S3: commodity identification data are compared with the commodity image that marked, determine the commodity figure
The recognition accuracy of each commodity as in then terminates when the recognition accuracy is greater than the recognition accuracy threshold value pre-seted
Step then carries out again the commodity identification model when the recognition accuracy is less than the recognition accuracy threshold value pre-seted
Secondary training.
In the present embodiment, the target area can be the shelf of intelligent refrigerator, or common shelf.When described
When refrigerator doors are opened to a preset angle, such as 45 °, by camera the commodity in refrigerator are carried out with the acquisition of commodity image.
In the present embodiment, the present invention can be applied in the terminal data acquisition dress being monitored to the commodity in refrigerator
It sets, is installed in refrigerator doors when by terminal data acquisition device, when being monitored to the commodity in refrigerator, can be taken the photograph described
As the commodity image of first order of magnitude of head acquisition is sent to backstage labeling system, and the mark returned according to backstage labeling system
Note commodity image in time judges the recognition accuracy of commodity identification model, in order to carry out to the commodity identification model
Training, avoids the lower problem of commodity identification model recognition accuracy in being actually used in environment.
Fig. 2 is to carry out step flow chart trained again in the present invention to commodity identification model, as shown in Fig. 2, when to institute
State commodity identification model carry out again train when, include the following steps:
Step M1: the commodity image of second order of magnitude of the camera acquisition is obtained, by the quotient of second order of magnitude
Product image is sent to backstage labeling system;
Step M2: the training returned after the backstage labeling system marks the commodity image of second order of magnitude is received
Image set;
Step M3: the commodity identification model is trained again by the training image collection.
In the present embodiment, the backstage labeling system can be backstage marking software, then be grasped by staff
Make the mark that the backstage marking software carries out commodity image.Can certainly automatic marking software carry out commodity image it is automatic
Mark.
In the present embodiment, it is possible to the commodity image of second order of magnitude directly acquired by camera, and pass through backstage
Labeling system is labeled generation training image collection to the commodity image, realize the commodity identification model line interation more
Newly, so that commodity identification model can adapt to different installation environments, commodity identification model is avoided because of the actual installation of refrigerator
Environment there are illumination, visual angle or increase new product in the case where, commodity identification model recognition accuracy reduce the problem of.
In the present embodiment, first order of magnitude is less than second order of magnitude;First order of magnitude is 10 left sides
The right side, such as 5 to 15;Second order of magnitude is 50 to 100.
Fig. 3 is the step flow chart that commodity identification model identifies commodity image in the present invention, as shown in figure 3, working as
The commodity identification model includes the following steps: when identifying to commodity image
Step N1: by the angle end identification model that pre-sets at least two stratoses of intelligent refrigerator in the commodity image
Stratose frame angle end identified, at least identify four be located at the commodity image both side ends stratose frame angle petiolarea
Domain;
Step N2: according to the reference layer at least four stratose frame angle end regions and the reference commodity image pre-seted
Positional relationship between the end regions of column frame angle generates homography matrix;
Step N3: commodity image after correction generation is corrected is carried out to the commodity image according to the homography matrix;
Step N4: identifying commodity image after the correction by the commodity identification model pre-seted, to know
It Chu not the corresponding goods number pre-seted in each commodity region in the commodity image.
In the present embodiment, the present invention will be carried out by the commodity identification model again after collecting the correction of commodity image part
Identification, when avoiding refrigerator doors open angle difference, there are commodity identification models caused by the difference of visual angle to know for acquired image
The lower problem of other accuracy rate.
Fig. 4 is the step flow chart of training angle end identification model in the present invention, as shown in figure 4, the step S1 includes such as
Lower step:
Step N101: obtaining multiple training images for the identification of angle end, to each training for the identification of angle end
The mark of image progress angle end regions;
Step N102: angle end identification model is established using the training image of mark angle of departure end regions;
Step N103: commodity image input angle end identification model is subjected to the identification of angle end regions.
Fig. 5 is the step flow chart of homography matrix in the present invention, as shown in figure 5, the step S2 includes the following steps:
Step N201: establishing two-dimensional coordinate system, determines that at least four with reference in commodity image refer to stratose frame angle
The coordinate of end regions;
Step N202: in the two-dimensional coordinate system, the coordinate of four stratose framves angle end regions is determined;
Step N203: according between four stratose frame angle end regions and corresponding reference stratose frame angle end regions
Positional relationship generate the homography matrix.
In the present embodiment, angle end identification model and the commodity identification model use convolutional neural networks structure,
It is trained to obtain under deep learning frame.
According to general knowledge known in this field, the stratose frame angle end regions of same stratose are located on a horizontal linear, are located at same
The stratose frame angle end regions of side end are located in a vertical straight line.It is successively used as known conditions, by four in the commodity image
The stratose frame angle end regions of a both side ends being located at are moved, and four are located at stratose frame angle end regions in rectangle as target
As a result, the homography matrix can be generated, and then realize the correction to the commodity image.
Fig. 6 is the step flow chart of the training commodity region recognition model in the present invention, as shown in fig. 6, the step
S4 includes the following steps:
Step N401: obtaining multiple training images for commodity region recognition, to each described for the knowledge of commodity region
Other training image carries out the mark in commodity region;
Step N402: the commodity region recognition model is established using the training image for marking out commodity region;
Step N403: the commodity image is inputted into the commodity region recognition model and carries out commodity region recognition.
Fig. 7 is the module diagram of commodity identification model active training system in the present invention, as shown in fig. 7, the present invention mentions
The commodity identification model active training system of confession, for realizing the commodity identification model active training method, comprising:
Commodity identify data generation module, first order of magnitude acquired for obtaining an at least camera to a target area
Commodity image, identified by commodity image of the preset commodity identification model to first order of magnitude, generate commodity
Identify data;
Image generation module is marked, for the commodity image of first order of magnitude to be sent to backstage labeling system, and
Receive the mark commodity image that the backstage labeling system returns;
Recognition accuracy judgment module, for comparing commodity identification data with the commodity image that marked
It is right, the recognition accuracy of each commodity in the commodity image is determined, when the recognition accuracy is greater than the identification standard pre-seted
When true rate threshold value, then end step, when the recognition accuracy is less than the recognition accuracy threshold value pre-seted, then to the quotient
Product identification model is trained again.
The commodity identification model active training equipment also provided in the embodiment of the present invention, including processor.Memory, wherein
It is stored with the executable instruction of processor.Wherein, processor is configured to execute commodity identification mould via executable instruction is executed
The step of type active training method.
As above, the commodity image for first order of magnitude that the camera acquires can be sent to rear logo in the embodiment
Injection system, and according to backstage labeling system return marked commodity image in time to the recognition accuracy of commodity identification model into
Row judgement, in order to be trained to the commodity identification model, avoids commodity identification model and knows in being actually used in environment
The lower problem of other accuracy rate.
Person of ordinary skill in the field it is understood that various aspects of the invention can be implemented as system, method or
Program product.Therefore, various aspects of the invention can be embodied in the following forms, it may be assumed that complete hardware embodiment, complete
The embodiment combined in terms of full Software Implementation (including firmware, microcode etc.) or hardware and software, can unite here
Referred to as " circuit ", " module " or " platform ".
Fig. 8 is the structural schematic diagram of the online shopping ancillary equipment in the present invention based on augmented reality.It is described referring to Fig. 8
The electronic equipment 600 of this embodiment according to the present invention.The electronic equipment 600 that Fig. 8 is shown is only an example, is not answered
Any restrictions are brought to the function and use scope of the embodiment of the present invention.
As shown in figure 8, electronic equipment 600 is showed in the form of universal computing device.The component of electronic equipment 600 can wrap
Include but be not limited to: at least one processing unit 610, at least one storage unit 620, connection different platform component (including storage
Unit 620 and processing unit 610) bus 630, display unit 640 etc..
Wherein, storage unit is stored with program code, and program code can be executed with unit 610 processed, so that processing is single
Member 610 executes various exemplary implementations according to the present invention described in this specification above-mentioned electronic prescription circulation processing method part
The step of mode.For example, processing unit 610 can execute step as shown in fig. 1.
Storage unit 620 may include the readable medium of volatile memory cell form, such as Random Access Storage Unit
(RAM) 6201 and/or cache memory unit 6202, it can further include read-only memory unit (ROM) 6203.
Storage unit 620 can also include program/utility with one group of (at least one) program module 6205
6204, such program module 6205 includes but is not limited to: operating system, one or more application program, other program moulds
It may include the realization of network environment in block and program data, each of these examples or certain combination.
Bus 630 can be to indicate one of a few class bus structures or a variety of, including storage unit bus or storage
Cell controller, peripheral bus, graphics acceleration port, processing unit use any bus structures in a variety of bus structures
Local bus.
Electronic equipment 600 can also be with one or more external equipments 700 (such as keyboard, sensing equipment, bluetooth equipment
Deng) communication, can also be enabled a user to one or more equipment interact with the electronic equipment 600 communicate, and/or with make
Any equipment (such as the router, modulation /demodulation that the electronic equipment 600 can be communicated with one or more of the other calculating equipment
Device etc.) communication.This communication can be carried out by input/output (I/O) interface 650.Also, electronic equipment 600 can be with
By network adapter 660 and one or more network (such as local area network (LAN), wide area network (WAN) and/or public network,
Such as internet) communication.Network adapter 660 can be communicated by bus 630 with other modules of electronic equipment 600.It should
Understand, although being not shown in Fig. 8, other hardware and/or software module can be used in conjunction with electronic equipment 600, including unlimited
In: microcode, device driver, redundant processing unit, external disk drive array, RAID system, tape drive and number
According to backup storage platform etc..
A kind of computer readable storage medium is also provided in the embodiment of the present invention, for storing program, program is performed
The step of image split-joint method of realization.In some possible embodiments, various aspects of the invention are also implemented as
A kind of form of program product comprising program code, when program product is run on the terminal device, program code is for making
Terminal device executes various exemplary according to the present invention described in this specification above-mentioned electronic prescription circulation processing method part
The step of embodiment.
As it appears from the above, the program of the computer readable storage medium of the embodiment is when being executed, the present invention can will be described
The commodity image of first order of magnitude of camera acquisition is sent to backstage labeling system, and is returned according to backstage labeling system
Mark commodity image the recognition accuracy of commodity identification model is judged in time, in order to the commodity identification model into
Row training, avoids the lower problem of commodity identification model recognition accuracy in being actually used in environment.
Fig. 9 is the structural schematic diagram of computer readable storage medium of the invention.Refering to what is shown in Fig. 9, describing according to this
The program product 800 for realizing the above method of the embodiment of invention can use the read-only storage of portable compact disc
Device (CD-ROM) and including program code, and can be run on terminal device, such as PC.However, journey of the invention
Sequence product is without being limited thereto, and in this document, readable storage medium storing program for executing can be any tangible medium for including or store program, the journey
Sequence can be commanded execution system, device or device use or in connection.
Program product can be using any combination of one or more readable mediums.Readable medium can be readable signal Jie
Matter or readable storage medium storing program for executing.Readable storage medium storing program for executing for example can be but be not limited to electricity, magnetic, optical, electromagnetic, infrared ray or partly lead
System, device or the device of body, or any above combination.More specific example (the non exhaustive column of readable storage medium storing program for executing
Table) it include: the electrical connection with one or more conducting wires, portable disc, hard disk, random access memory (RAM), read-only storage
Device (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read only memory (CD-
ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.
Computer readable storage medium may include in a base band or as carrier wave a part propagate data-signal,
In carry readable program code.The data-signal of this propagation can take various forms, including but not limited to electromagnetic signal,
Optical signal or above-mentioned any appropriate combination.Readable storage medium storing program for executing can also be any readable Jie other than readable storage medium storing program for executing
Matter, the readable medium can send, propagate or transmit for by instruction execution system, device or device use or and its
The program of combined use.The program code for including on readable storage medium storing program for executing can transmit with any suitable medium, including but not
It is limited to wireless, wired, optical cable, RF etc. or above-mentioned any appropriate combination.
The program for executing operation of the present invention can be write with any combination of one or more programming languages
Code, programming language include object oriented program language-Java, C++ etc., further include conventional process
Formula programming language-such as " C " language or similar programming language.Program code can be calculated fully in user
It executes in equipment, partly execute on a user device, executing, as an independent software package partially in user calculating equipment
Upper part executes on a remote computing or executes in remote computing device or server completely.It is being related to remotely counting
In the situation for calculating equipment, remote computing device can pass through the network of any kind, including local area network (LAN) or wide area network
(WAN), it is connected to user calculating equipment, or, it may be connected to external computing device (such as utilize ISP
To be connected by internet).
In the present embodiment, the present invention can be applied in the terminal data acquisition dress being monitored to the commodity in refrigerator
It sets, is installed in refrigerator doors when by terminal data acquisition device, it, can be by the end when being monitored to the commodity in refrigerator
The commodity image of first order of magnitude of the camera acquisition on end data acquisition device is sent to backstage labeling system, and according to rear
The commodity image that marked that logo injection system returns in time judges the recognition accuracy of commodity identification model, in order to right
The commodity identification model is trained, and avoids that commodity identification model recognition accuracy in being actually used in environment is lower to ask
Topic.
Specific embodiments of the present invention are described above.It is to be appreciated that the invention is not limited to above-mentioned
Particular implementation, those skilled in the art can make various deformations or amendments within the scope of the claims, this not shadow
Ring substantive content of the invention.
Claims (10)
1. a kind of commodity identification model active training method, which comprises the steps of:
Step S1: the commodity image for first order of magnitude that an at least camera acquires a target area is obtained, by preset
Commodity identification model identifies the commodity image of first order of magnitude, generates commodity and identifies data;
Step S2: being sent to backstage labeling system for the commodity image of first order of magnitude, and receives backstage mark system
The mark commodity image that system returns;
Step S3: commodity identification data are compared with the commodity image that marked, are determined in the commodity image
The recognition accuracy of each commodity, when the recognition accuracy is greater than the recognition accuracy threshold value pre-seted, then end step,
When the recognition accuracy is less than the recognition accuracy threshold value pre-seted, then the commodity identification model is instructed again
Practice.
2. commodity identification model active training method according to claim 1, which is characterized in that identified when to the commodity
When model train again, include the following steps:
Step M1: the commodity image of second order of magnitude of the camera acquisition is obtained, by the commodity figure of second order of magnitude
As being sent to backstage labeling system;
Step M2: the training image returned after the backstage labeling system marks the commodity image of second order of magnitude is received
Collection;
Step M3: the commodity identification model is trained again by the training image collection.
3. commodity identification model active training method according to claim 2, which is characterized in that first order of magnitude is small
In second order of magnitude.
4. commodity identification model active training method according to claim 1, which is characterized in that when the commodity identify mould
Include the following steps: when type identifies commodity image
Step N1: by the angle end identification model that pre-sets to the layer of at least two stratoses of intelligent refrigerator in the commodity image
Column frame angle end is identified, at least identifies the stratose frame angle end regions for the both side ends that four are located at the commodity image;
Step N2: according to the reference stratose frame at least four stratose frame angle end regions and the reference commodity image pre-seted
Positional relationship between the end regions of angle generates homography matrix;
Step N3: commodity image after correction generation is corrected is carried out to the commodity image according to the homography matrix;
Step N4: commodity image after the correction is identified by the commodity identification model pre-seted, to identify
The corresponding goods number pre-seted in each commodity region in the commodity image.
5. commodity identification model active training method according to claim 4, which is characterized in that the step S1 includes such as
Lower step:
Step N101: obtaining multiple training images for the identification of angle end, to each training image for the identification of angle end
Carry out the mark of angle end regions;
Step N102: angle end identification model is established using the training image of mark angle of departure end regions;
Step N103: commodity image input angle end identification model is subjected to the identification of angle end regions.
6. commodity identification model active training method according to claim 4, which is characterized in that the step S2 includes such as
Lower step:
Step N201: establishing two-dimensional coordinate system, determines that at least four with reference in commodity image refer to stratose frame angle petiolarea
The coordinate in domain;
Step N202: in the two-dimensional coordinate system, the coordinate of four stratose framves angle end regions is determined;
Step N203: according to four stratose frame angle end regions and the corresponding position with reference between the end regions of stratose frame angle
It sets relationship and generates the homography matrix.
7. commodity identification model active training method according to claim 4, which is characterized in that the step S4 includes such as
Lower step:
Step N401: obtaining multiple training images for commodity region recognition, to each described for commodity region recognition
The mark in training image progress commodity region;
Step N402: the commodity region recognition model is established using the training image for marking out commodity region;
Step N403: the commodity image is inputted into the commodity region recognition model and carries out commodity region recognition.
8. a kind of commodity identification model active training system, for realizing the identification of commodity described in any one of claims 1 to 7
Model active training method characterized by comprising
Commodity identify data generation module, for obtaining the quotient for first order of magnitude that an at least camera acquires a target area
Product image is identified by commodity image of the preset commodity identification model to first order of magnitude, generates commodity identification
Data;
Image generation module is marked, for the commodity image of first order of magnitude to be sent to backstage labeling system, and is received
The mark commodity image that the backstage labeling system returns;
Recognition accuracy judgment module, for commodity identification data to be compared with the commodity image that marked, really
The recognition accuracy of each commodity in the fixed commodity image, when the recognition accuracy is greater than the recognition accuracy threshold pre-seted
When value, then end step then identifies the commodity when the recognition accuracy is less than the recognition accuracy threshold value pre-seted
Model is trained again.
9. a kind of commodity identification model active training equipment characterized by comprising
Processor;
Memory, wherein being stored with the executable instruction of the processor;
Wherein, the processor is configured to come any one of perform claim requirement 1 to 7 institute via the execution executable instruction
The step of stating commodity identification model active training method.
10. a kind of computer readable storage medium, for storing program, which is characterized in that described program is performed realization power
Benefit require any one of 1 to 7 described in commodity identification model active training method the step of.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910269443.3A CN109948736A (en) | 2019-04-04 | 2019-04-04 | Commodity identification model active training method, system, equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910269443.3A CN109948736A (en) | 2019-04-04 | 2019-04-04 | Commodity identification model active training method, system, equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109948736A true CN109948736A (en) | 2019-06-28 |
Family
ID=67013913
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910269443.3A Pending CN109948736A (en) | 2019-04-04 | 2019-04-04 | Commodity identification model active training method, system, equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109948736A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110781813A (en) * | 2019-10-24 | 2020-02-11 | 北京市商汤科技开发有限公司 | Image recognition method and device, electronic equipment and storage medium |
CN112270664A (en) * | 2020-04-16 | 2021-01-26 | 上海扩博智能技术有限公司 | Transparent packaging product detection method, system, equipment and storage medium |
WO2023130429A1 (en) * | 2022-01-10 | 2023-07-13 | 烟台创迹软件有限公司 | Object identification method, object identification apparatus and model learning method |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102982332A (en) * | 2012-09-29 | 2013-03-20 | 顾坚敏 | Retail terminal goods shelf image intelligent analyzing system based on cloud processing method |
CN106951925A (en) * | 2017-03-27 | 2017-07-14 | 成都小多科技有限公司 | Data processing method, device, server and system |
CN108647671A (en) * | 2018-06-28 | 2018-10-12 | 武汉市哈哈便利科技有限公司 | A kind of optical indicia visual identity method and the self-service cabinet based on this method |
CN108647720A (en) * | 2018-05-10 | 2018-10-12 | 上海扩博智能技术有限公司 | Iterative cycles recognition methods, system, equipment and the storage medium of commodity image |
CN109190110A (en) * | 2018-08-02 | 2019-01-11 | 厦门快商通信息技术有限公司 | A kind of training method of Named Entity Extraction Model, system and electronic equipment |
CN109190919A (en) * | 2018-08-10 | 2019-01-11 | 上海扩博智能技术有限公司 | It is sold critical sales index generation method, system, equipment and storage medium |
-
2019
- 2019-04-04 CN CN201910269443.3A patent/CN109948736A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102982332A (en) * | 2012-09-29 | 2013-03-20 | 顾坚敏 | Retail terminal goods shelf image intelligent analyzing system based on cloud processing method |
CN106951925A (en) * | 2017-03-27 | 2017-07-14 | 成都小多科技有限公司 | Data processing method, device, server and system |
CN108647720A (en) * | 2018-05-10 | 2018-10-12 | 上海扩博智能技术有限公司 | Iterative cycles recognition methods, system, equipment and the storage medium of commodity image |
CN108647671A (en) * | 2018-06-28 | 2018-10-12 | 武汉市哈哈便利科技有限公司 | A kind of optical indicia visual identity method and the self-service cabinet based on this method |
CN109190110A (en) * | 2018-08-02 | 2019-01-11 | 厦门快商通信息技术有限公司 | A kind of training method of Named Entity Extraction Model, system and electronic equipment |
CN109190919A (en) * | 2018-08-10 | 2019-01-11 | 上海扩博智能技术有限公司 | It is sold critical sales index generation method, system, equipment and storage medium |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110781813A (en) * | 2019-10-24 | 2020-02-11 | 北京市商汤科技开发有限公司 | Image recognition method and device, electronic equipment and storage medium |
CN112270664A (en) * | 2020-04-16 | 2021-01-26 | 上海扩博智能技术有限公司 | Transparent packaging product detection method, system, equipment and storage medium |
WO2023130429A1 (en) * | 2022-01-10 | 2023-07-13 | 烟台创迹软件有限公司 | Object identification method, object identification apparatus and model learning method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108647553B (en) | Method, system, device and storage medium for rapidly expanding images for model training | |
JP7000555B2 (en) | Augmented reality display method, attitude information determination method and equipment | |
CA3111279A1 (en) | Deep learning-based actionable digital receipts for cashier-less checkout | |
Lambrecht et al. | Spatial programming for industrial robots through task demonstration | |
CN109556596A (en) | Air navigation aid, device, equipment and storage medium based on ground texture image | |
CN108416403A (en) | The automatic correlation method of commodity and label, system, equipment and storage medium | |
CN109948736A (en) | Commodity identification model active training method, system, equipment and storage medium | |
US20210174299A1 (en) | Method, system and device for collecting sales information of commodities in cabinet, and storage medium | |
CN109886169B (en) | Article identification method, device, equipment and storage medium applied to unmanned container | |
CN110322300B (en) | Data processing method and device, electronic equipment and storage medium | |
CN109190919A (en) | It is sold critical sales index generation method, system, equipment and storage medium | |
CN109902636A (en) | Commodity identification model training method, system, equipment and storage medium | |
CN110516628A (en) | Shelf vacant locations merchandise news acquisition methods, system, equipment and storage medium | |
CN108389230A (en) | Refrigerator capacity automatic testing method, system, equipment and storage medium | |
CN109918513A (en) | Image processing method, device, server and storage medium | |
CN109635198A (en) | The method, apparatus of presentation user's search result, medium and electronic equipment on merchandise display platform | |
CN109814545A (en) | Replenishing method, device and the storage medium of the unmanned vending machine of automatic Pilot | |
US11508105B2 (en) | Using augmented reality for accessing legacy transaction terminals | |
CN109242060A (en) | New restocking product fast searching method, system, equipment and storage medium | |
CN111553889A (en) | Method, system, equipment and storage medium for comparing commodity placement positions on goods shelf | |
CN109784172A (en) | Ground heap commodity occupied area evaluation method, system, equipment and medium in market | |
CN109978668A (en) | Based on the matched transaction customer recognition methods of timestamp, system, equipment and medium | |
CN107862557A (en) | Customer's dynamic tracing system and method | |
CN109733783A (en) | A kind of cargo restocking method, apparatus, electronic equipment and storage medium | |
US11636663B2 (en) | Localizing relevant objects in multi-object images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190628 |