CN109657681A - Mask method, device, electronic equipment and the computer readable storage medium of picture - Google Patents
Mask method, device, electronic equipment and the computer readable storage medium of picture Download PDFInfo
- Publication number
- CN109657681A CN109657681A CN201811623619.2A CN201811623619A CN109657681A CN 109657681 A CN109657681 A CN 109657681A CN 201811623619 A CN201811623619 A CN 201811623619A CN 109657681 A CN109657681 A CN 109657681A
- Authority
- CN
- China
- Prior art keywords
- picture
- item
- detection block
- processed
- information
- 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
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
Abstract
This application provides a kind of mask method of picture, device, electronic equipment and computer readable storage mediums.This method comprises: identifying to picture to be processed, the corresponding detection block information of each single-item in picture to be processed is determined；Each corresponding detection block of detection block information is ranked up, and arrangement information is identified based on preset single-item, the corresponding single-item mark of each detection block after determining sequence；According to each detection block information, each single-item picture for carrying single-item mark is cut out from picture to be processed.The application by each detection block information for identifying by each single-item in picture to be processed outline come, picture to be processed is cut based on each detection block, a large amount of single-item picture can quickly be obtained, greatly improve collecting efficiency, and based on the corresponding single-item mark of each detection block, the single-item mark of single-item picture can be directly obtained, a large amount of mark task is quickly completed, human cost is saved, annotating efficiency is promoted.
Description
Technical field
This application involves machine vision and field of artificial intelligence, specifically, this application involves a kind of marks of picture
Injecting method, device, electronic equipment and computer readable storage medium.
Background technique
Artificial intelligence technology gradually affects people's lives, and many enterprises change by the strength of artificial intelligence silently
The consumption pattern for becoming us brings subversive change to traditional retail structure.By taking unmanned supermarket as an example, as retail trade
An important application scene under Internet of Things and the Internet converged, in particular with the unmanned clearing under new public safety and fastly
Prompt cash register demand it is growing, unmanned supermarket's scene based on machine vision need to train deep neural network (such as CNN,
Convolutional Neural Network, convolutional neural networks) it goes to detect cargo and classified, this is just needed largely
Cargo picture and corresponding markup information just can preferably complete to train.
Through investigating, the existing excessively artificial Direct Mark of mark task multi-pass artificially acquires picture by camera, and
Position mark and attributive classification are carried out to the cargo in picture.For every kind of cargo in the scenes such as unmanned supermarket, or even also need
The photo of different angle is acquired, and general scheme is cargo to be placed on rotatable pallet to be taken pictures or recorded a video at present.
However, not only cost of labor is higher with notation methods for artificial acquisition, it is low with annotating efficiency to fully rely on artificial acquisition
Under, and inevitably there is fault in manual operation, and markup information quality caused by casual mistake reduces, and influences whether depth
The precision of neural network model is finally made troubles to the use of user.Particularly with the retail trade of enormous amount, work as cargo
When excessive, how more rapid and better to complete mark task becomes the technical problem of urgent need to resolve.
Summary of the invention
To overcome above-mentioned technical problem or at least being partially solved above-mentioned technical problem, spy proposes following technical scheme:
In a first aspect, this application provides a kind of mask methods of picture, this method comprises:
Picture to be processed is identified, determines the corresponding detection block letter of each single-item in the picture to be processed
Breath；
Each corresponding detection block of detection block information is ranked up, and based on preset single-item mark arrangement letter
Breath, the corresponding single-item mark of each detection block after determining sequence；
According to each detection block information, each single-item figure for carrying single-item mark is cut out from the picture to be processed
Piece.
In an optional implementation manner, described that picture to be processed is identified, it determines in the picture to be processed
The corresponding detection block information of each single-item, comprising:
Picture to be processed is identified by the detector of pre-training, determines each single-item in the picture to be processed
Corresponding detection block information.
In an optional implementation manner, the detector by pre-training carries out picture to be processed to identify it
Before, further includes:
The inspection of training sample picture based on acquisition and each single-item in the correspondence training sample picture marked
Survey frame information, the training detector.
In an optional implementation manner, the detection block information, comprising:
The dimension information of location information of the detection block in picture to be processed, detection block.
It is in an optional implementation manner, described that each corresponding detection block of detection block information is ranked up, comprising:
According to location information of each detection block in picture to be processed, each detection block is ranked up.
In an optional implementation manner, it is described picture to be processed is identified before, further includes:
The picture to be processed including multiple single-items is acquired by picture collection equipment, the multiple single-item is according to the single-item
Mark arrangement information is arranged.
In an optional implementation manner, described that the figure to be processed including multiple single-items is acquired by picture collection equipment
Piece, comprising:
Pass through picture collection equipment, to be processed picture of the acquisition for the different angle of the multiple single-item；
It is described that each single-item picture for carrying single-item mark is cut out from the picture to be processed, comprising:
Each difference for carrying single-item mark is cut out from the picture to be processed of the different angle of the multiple single-item
The single-item picture of angle.
Second aspect, this application provides a kind of annotation equipment of picture, which includes:
Identification module determines each single-item difference in the picture to be processed for identifying to picture to be processed
Corresponding detection block information；
Sorting module for being ranked up to each corresponding detection block of detection block information, and is based on preset list
Product identify arrangement information, the corresponding single-item mark of each detection block after determining sequence；
Module is cut, each carries list for being cut out from the picture to be processed according to each detection block information
The single-item picture of product mark.
In an optional implementation manner, the identification module is specifically used for the detector by pre-training to be processed
Picture is identified, determines the corresponding detection block information of each single-item in the picture to be processed.
In an optional implementation manner, the device further include:
Training module, in the training sample picture based on acquisition and the correspondence marked the training sample picture
The detection block information of each single-item, the training detector.
In an optional implementation manner, the detection block information, comprising:
The dimension information of location information of the detection block in picture to be processed, detection block.
In an optional implementation manner, the sorting module is specifically used for according to each detection block in picture to be processed
In location information, each detection block is ranked up.
In an optional implementation manner, the device further include:
Acquisition module, for including the picture to be processed of multiple single-items, the multiple list by the acquisition of picture collection equipment
Product are arranged according to single-item mark arrangement information.
In an optional implementation manner, the acquisition module is specifically used for through picture collection equipment, and acquisition is directed to
The picture to be processed of the different angle of the multiple single-item；
The cutting module is specifically used for cutting out from the picture to be processed of the different angle of the multiple single-item each
Carry the single-item picture of the different angle of single-item mark.
The third aspect, this application provides a kind of electronic equipment, which includes:
Processor and memory, memory are stored at least one instruction, at least a Duan Chengxu, code set or instruction set,
At least one instruction, an at least Duan Chengxu, code set or instruction set loads by processor and are executed the to realize such as disclosure
Method shown in one side.
Fourth aspect, present disclose provides a kind of computer readable storage mediums, and computer storage medium is based on storing
The instruction of calculation machine, program, code set or instruction set, when run on a computer, so that computer executes the such as the disclosure
Method shown in one side.
Technical solution provided by the present application have the benefit that the mask method of picture provided by the present application, device,
Electronic equipment and computer readable storage medium, by each detection block information for identifying by each list in picture to be processed
Product, which outline, to be come, and picture to be processed is cut based on each detection block, can quickly obtain a large amount of single-item picture, without to every kind
Single-item does individual Image Acquisition, greatly improves collecting efficiency, and based on the corresponding single-item mark of each detection block, can be straight
Obtain single-item picture single-item mark, quickly complete a large amount of mark task, without cumbersome artificial mark, save manpower
Cost improves annotating efficiency while reducing mark error probability.
Detailed description of the invention
In order to more clearly explain the technical solutions in the embodiments of the present application, institute in being described below to the embodiment of the present application
Attached drawing to be used is needed to do simple introduction.
Fig. 1 is a kind of exemplary diagram of mark task provided by the embodiments of the present application；
Fig. 2 is a kind of flow diagram of the mask method of picture provided by the embodiments of the present application；
Fig. 3 is a kind of exemplary diagram of the picture to be processed with detection block provided by the embodiments of the present application；
Fig. 4 is the flow diagram of a kind of acquisition and mask method provided by the embodiments of the present application；
Fig. 5 is a kind of structural schematic diagram of the annotation equipment of picture provided by the embodiments of the present application；
Fig. 6 is the structural schematic diagram of a kind of electronic equipment provided by the embodiments of the present application.
Specific embodiment
Embodiments herein is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end
Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached
The embodiment of figure description is exemplary, and is only used for explaining the application, and cannot be construed to the limitation to the application.
Those skilled in the art of the present technique are appreciated that unless expressly stated, singular " one " used herein, " one
It is a ", " described " and "the" may also comprise plural form.It is to be further understood that being arranged used in the description of the present application
Diction " comprising " refer to that there are the feature, integer, step, operation, element and/or component, but it is not excluded that in the presence of or addition
Other one or more features, integer, step, operation, element, component and/or their group.It should be understood that when we claim member
Part is " connected " or when " coupled " to another element, it can be directly connected or coupled to other elements, or there may also be
Intermediary element.In addition, " connection " used herein or " coupling " may include being wirelessly connected or wirelessly coupling.It is used herein to arrange
Diction "and/or" includes one or more associated wholes for listing item or any cell and all combinations.
To keep the purposes, technical schemes and advantages of the application clearer, below in conjunction with attached drawing to the application embodiment party
Formula is described in further detail.
As shown in Figure 1, being placed with 20 commodity on the shelf of Fig. 1 by taking the shelf of unmanned supermarket as an example.Mark task is pair
Each commodity shooting includes the picture of complete commodity, and selects corresponding product name from known 20 kinds of product names
It is assigned to the picture of shooting.
According to existing artificial notation methods, it can artificially shoot the picture of a commodity and choose a kind of correct
Product name can also artificially shoot the picture comprising a variety of commodity, and artificially in picture to each commodity mark
The external frame that can cover its complete commodity is outpoured, and corresponding quotient artificially is selected from 20 kinds of product names to each external frame
The name of an article claims to be assigned to external frame.Even if marking an external frame and choosing correct product name also needs using latter approach
Want the time of 3-5s.So, mark task is executed to the entire service in Fig. 1 and needs 1-2 minutes.In addition, being directed to every kind
Commodity also need the photo of acquisition different angle, and general scheme is that cargo is placed in on rotatable pallet different angles of taking pictures at present
The plurality of pictures or video recording of degree.When commodity amount is more, more times will be expended.
As it can be seen that not only cost of labor is higher with notation methods for artificial acquisition, it is low with annotating efficiency to fully rely on artificial acquisition
Under, and inevitably there is fault in manual operation, and markup information quality caused by casual mistake reduces, and influences whether depth
The precision of neural network model is finally made troubles to the use of user.
Mask method, device, electronic equipment and the computer readable storage medium of picture provided by the present application, core are
The picture of high quality and the picture markup information of high precision rate are obtained, by way of short period and lower cost to solve
The technical problem as above of the prior art.
How the technical solution of the application and the technical solution of the application are solved with specific embodiment below above-mentioned
Technical problem is described in detail.
The embodiment of the present application provides a kind of mask method of picture, as shown in Fig. 2, this method comprises:
Step S201: identifying picture to be processed, determines the corresponding inspection of each single-item in picture to be processed
Survey frame information；
Wherein, picture to be processed is to carry out Image Acquisition for the multiple single-items for needing to be implemented mark task in advance to obtain
, i.e., it include the multiple single-items for needing to be implemented mark task in picture to be processed.
In practical application, single-item is different the single item for needing to complete mark task under scene.For example, super at nobody
In city's scene, single-item can be any commodity, such as one bottle of beverage, one bag of instant food etc.；In another example in automatic library
In scene, single-item can be books, etc.；The embodiment of the present application to the concrete type of single-item it is not limited here.
In the embodiment of the present application, picture to be processed is identified, is determined for each single-item in picture to be processed
Corresponding detection block information includes how many single-item in picture that is, to be processed, will determine how many detection block information.Wherein,
The corresponding detection block of each detection block information can cover its corresponding complete single-item.
In practical application, detection block can be rectangle, circular, oval, diamond shape etc., but not limited to this,
As long as detection block can be can be used as convenient for the shape of marker location and size, in practical application, it is contemplated that covering it is reliable
Property, detection block is usually the boundary rectangle frame of single-item.
Step S202: being ranked up each corresponding detection block of detection block information, and is based on preset single-item mark
Know arrangement information, the corresponding single-item mark of each detection block after determining sequence；
Wherein, the true row of each single-item when preset single-item mark arrangement information is according to acquisition picture to be processed
Column sequence is preset.
In the embodiment of the present application, each detection block is ranked up, since the arrangement of the corresponding single-item of each detection block is suitable
It is identical, that is, after can determine that sequence each detection block difference that sequence, which identifies putting in order for corresponding single-item with each single-item,
Corresponding single-item mark.
Wherein, single-item mark can include but is not limited to single-item title, single-item ID (identification, identity
Number), single-item type, single-item attribute etc..
Step S203: according to each detection block information, each single-item mark of carrying is cut out from picture to be processed
Single-item picture.
It, can be according to detection block since the corresponding detection block of each detection block information can cover its corresponding complete commodity
Corresponding single-item picture is cut (crop) from picture to be processed and come out by information.
And due to defining the corresponding single-item mark of detection block in step S202, that is, cutting out single-item picture can carry
Corresponding single-item mark.
In this way, for each single-item, so that it may obtain corresponding single-item picture and corresponding single-item mark, that is, be rapidly completed
The mark task of each single-item.
As it can be seen that without artificially going to select from numerous alternative single-item marks, therefore being avoided by the embodiment of the present application
The generation of human error.
The mask method of picture provided by the embodiments of the present application, by each detection block information for identifying by figure to be processed
Each single-item in piece, which outlines, to be come, and picture to be processed is cut based on each detection block, can quickly obtain a large amount of single-item figure
Piece greatly improves collecting efficiency, and corresponding based on each detection block without doing individual Image Acquisition to every kind of single-item
Single-item mark can directly obtain the single-item mark of single-item picture, a large amount of mark task be quickly completed, without cumbersome artificial
Mark saves human cost, while reducing mark error probability, improves annotating efficiency.
Wherein, arrangement information is identified by preset single-item to determine that each detection block after sequence is corresponding
Single-item mark, i.e., by the mask method of fixed relative position, especially suitable for the scene of single-item disposing way rule relatively, such as
Unmanned supermarket etc..
In a kind of feasible implementation, in step s 201, the embodiment of the present application can pass through the detector of pre-training
Picture to be processed is identified, determines the corresponding detection block information of each single-item in picture to be processed.
Specifically, detector can relatively accurately return to be processed by the detector of picture to be processed input pre-training
The detection block information of each single-item in picture, as shown in figure 3, the corresponding detection block of detection block information can be by picture to be processed
In each single-item outline come (dotted line frame in such as Fig. 3, the dotted line frame can by detector export detection block information obtain
To), accurately each single-item can be separated from whole picture to be processed.
As shown in figure 4, being needed before the embodiment of the present application identifies picture to be processed by the detector of pre-training
Detector is trained.
Specifically, each single-item in training sample picture based on acquisition and the correspondence training sample picture marked
Detection block information, training detector.
Wherein, training sample picture can be gathered in advance, is also possible to obtain from local or cloud, training sample figure
It is needed in piece comprising multiple single-items, therefore for the embodiment of the present application, targetedly can acquire or obtain and is such
Training sample picture.
In practical application, the quantity of the training sample picture of acquisition can be multiple, such as 20 or so.It is appreciated that
Excessive amount of training sample picture will increase trained cost, and very few training sample picture can be such that training effect not enough manages
Think.Those skilled in the art can the quantity according to the actual situation to training sample picture be configured, the embodiment of the present application exists
This is without limitation.
In addition, the single-item quantity and single-item type in different training sample pictures can be the same or different, ability
Field technique personnel can also select suitable training sample picture for training according to the actual situation.
Then, these are obtained by original artificial notation methods or other notation methods to the training sample picture of acquisition
The detection block information of every kind of single-item on training sample picture.Based on these training sample pictures and the detection block information marked,
Network is detected using existing basis, so that it may detector needed for training the embodiment of the present application.
In the embodiment of the present application, detection block information be can include but is not limited to: position of the detection block in picture to be processed
Information (bounding box location), dimension information (bounding box size) of detection block etc..
In practical application, by detection block be rectangle frame for, detection block information can refer to detection block four-dimensional information x,
Y, w, h }, wherein coordinate system can be established in picture to be processed, { x, y } indicates that the left upper apex of rectangle frame is (or right
Upper vertex, bottom left vertex, bottom right vertex, central point of rectangle etc.) coordinate (i.e. location information) in a coordinate system；{ w, h } table
Show the length and width (i.e. dimension information) of rectangle frame.
Alternatively, detector output can be the three-dimensional information { x, y, r } of detection block so that detection block is circular frame as an example,
Similarly, { x, y } indicates that the coordinate (i.e. location information) of the centre point of circular frame in a coordinate system, { r } indicate the half of rectangle frame
Diameter (i.e. dimension information).
Alternatively, detection block can refer to position of each single-item in picture to be processed in the location information in picture to be processed
Relationship, such as the positional relationship of some single-item and its upper and lower, left and right side single-item are set, in another example as shown in Figure 1, being located at xth
The position of y-th of single-item of row is array [x, y].When being applied to the scenes such as unmanned supermarket, positional relationship is also denoted as
The position of y-th of single-item of xth row on n-th of shelf is array [n, x, y].
The dimension information of detection block may also mean that the information such as pixel, area.
Those skilled in the art will be understood that above-mentioned several detection block information are only for example, and be carried out based on these examples
Appropriate variation is equally applicable to the application, therefore should also be included within the protection scope of the application.
Based on this, the embodiment of the present application provides a kind of feasible implementation to each detection block to be ranked up,
Specifically, the location information according to each detection block in picture to be processed is ranked up each detection block.
For example, according to each detection block coordinate in a coordinate system or the positional relationship of each single-item, it can be very simple
Be mapped as the true position of each single-item, and be ranked up according to this.
For above, as shown in figure 4, picture to be processed to be sent into the detector of the embodiment of the present application training, by examining
Survey the detection block information that device exports each single-item in picture to be processed, the location information including detection block in picture to be processed
With the dimension information of detection block.The picture to be processed with detection block can be obtained at this time.First according to the location information of detection block
Detection block is ranked up, then the corresponding single-item of each detection block is determined based on preset single-item mark arrangement information
Mark.Then, according to the dimension information of the detection block location information in picture to be processed and detection block, by each single-item picture
It cuts out and from picture to be processed, due to being believed according to the size of location information of the detection block in picture to be processed and detection block
Breath, can determine detection block part occupied in picture to be processed, and since to cover its corresponding complete for detection block
The part is extracted or is copied by single-item, corresponding single-item picture can be cut out from picture to be processed and come.This
When, each single-item picture being cut out has been set to correspond to respective single-item mark.
In the embodiment of the present application, time that whole single-items in picture to be processed for one are labeled can be met in 1s
It is interior, and at the same time the single-item picture of every kind of single-item can be exported.
As can be seen that the embodiment of the present application relies on detector without a large amount of mark personnel compared to traditional labelling schemes
Notation methods, the process manually marked is replaced with into machine mark, can be accomplished without cumbersome artificial mark
The markup information of every kind of single-item is obtained, to save human cost, promotes annotating efficiency.And without being done individually to every kind of single-item
Image Acquisition, so that it may quickly obtain a large amount of single-item picture, greatly improve the collecting efficiency of single-item picture.In addition, this Shen
Please embodiment be not necessarily to mark personnel and go to select markup information from numerous alternative possibility, therefore avoid the hair of human error
It is raw, reduce the probability of markup information error.
In the embodiment of the present application, continue as shown in figure 4, before being identified to picture to be processed, it is also necessary to which acquisition should
Picture to be processed, specifically, the picture to be processed including multiple single-items, multiple single-items can be acquired by picture collection equipment
It is arranged according to single-item mark arrangement information.
Wherein, picture collection equipment may include preconfigured camera, robot, unmanned plane etc..
In practical application, single-item can be placed on pre-set placement platform, such as in unmanned supermarket's scene,
Placement platform can be shelf.In other scenes, placement platform can be bookshelf, shelf, Vending Machine etc..Wherein, scheme
The visual angle of piece acquisition equipment is set as that complete placement platform can be covered.
Since the embodiment of the present application has preset single-item mark arrangement information, picture collection equipment pair can be configured
The multiple single-items arranged according to single-item mark arrangement information are shot.For example, can be according to preset single-item mark
Know arrangement information, the single-item of identical several amount and type is put according to the identical position of single-item mark arrangement information, and
It is shot after being well placed.Alternatively, single-item mark arrangement information can also be set according to true display case, and according to list
The single-item number amount and type that product mark arrangement information includes are shot, i.e., need to only shoot the list that single-item mark arrangement information includes
The picture to be processed of product.
Furthermore, it is contemplated that needing the picture of each single-item different angle under the application scenarios such as unmanned supermarket, the application is implemented
In example, the placement angle (rotating horizontally, spin upside down) of adjustable single-item acquires different pictures to be processed afterwards, or adjusts
The shooting angle of whole picture collection equipment acquires different pictures to be processed, i.e., by picture collection equipment, acquisition is for single
Product identify the picture to be processed of the different angle of the corresponding multiple single-items of arrangement information.
Wherein, since there is no variations for the position of each single-item, that is, relative position is still fixed, then adopting
The whole of collection picture to be processed inputs trained detector, and it is different can to complete each single-item according to processing mode above
The mark task of angle.That is it in step S203, can be cut out from the picture to be processed of the different angle of multiple single-item each
The single-item picture of a different angle for carrying single-item mark.
It can be seen that the embodiment of the present application is only needed without being rotated and being shot after every time individually selecting some single-item
Detector is integrally shot and be sent into after whole single-items to be done to the adjustment of angle, i.e., can disposably be obtained with markup information
A large amount of single-item photos, so that the acquisition of single-item and annotating efficiency have obtained great promotion.And due to the opposite position of each single-item
It sets and does not change, different angle will not be adjusted to because of single-item by mistake mark at other commodity signs, further reduced mark
Infuse the probability of information error.
In the following, by taking unmanned supermarket's scene as an example, the application of the technical solution of the application under concrete scene is introduced.
Wherein, which can configure a camera in deployment, and one or more shelf and one have GPU
The calculating equipment of operation, the visual angle of camera are configured to that all shelf can be covered, and method below is executed by calculating equipment.
Firstly, it is necessary to one detector of training, can choose and directly use after being trained under line, it can also be to be instructed on line
Carried out after white silk using.
Specifically, training process are as follows: acquire some shelf pictures full of commodity, the commodity amount in every shelf picture
It can be different with type.It can refer to Fig. 1, at this point, the different shape in Fig. 1 represents different commodity.Or under actual scene, no
Similar shape is also possible to the similar commodity with Different Package appearance.Different shape herein is only for example, and should not be understood as pair
The restriction of commodity.Then by these shelf pictures it is original manually mark by way of, obtain every kind of quotient of these planogram on pieces
The detection block information of product.By these shelf pictures and detection block information, the detector of unmanned supermarket's scene is trained.
Then, which in application process, can relatively accurately return to the quotient in input picture (picture to be processed)
The detection block information of product.The corresponding detection block of each detection block information can be accurately by commodity from input pictures point
Out.Wherein, input picture is acquired after the commodity that will need to mark are placed on shelf.During being somebody's turn to do, it is also necessary to determine quotient
Product are placed in the arrangement information on shelf, and using these detection block information and arrangement information, we can pass through picture cutting
Mode obtain each commodity single-item photo and corresponding product name, i.e. markup information.
In addition, the arrangement information due to commodity is fixed and invariable, can only rotate each commodity do not change commodity it
Preceding relative positional relationship and be acquired and mark.Such as with reference to Fig. 3, it is assumed that one layer of the top of shelf respectively food A,
Food B, food C and food D, the second layer are food E, food F and food G.Every time in order to collect different food products different angle
Single-item photo, only need to do Simple rotary to every kind of food and keep its relative position not change simultaneously (such as food A is being eaten
On the left of the same layer of product B, food G is on the right side of the same layer of food F, upper one layer etc. in food E of food A), it then claps and is rounded goods
The picture of frame is simultaneously sent into detector.Detector can export the detection block information of this 7 kinds of food, only need to be by these detection block information
It is ranked up according to the exact arrangement mode of food A~G, then input picture is cut out to get corresponding to every kind of food
Markup information.
Those skilled in the art will be understood that above-mentioned application scenarios are only for example, which has very strong transfer ability,
As long as passing through the detector of the embodiment of the present application pre-training, then can be easily by this by the mask method of fixed relative position
The mask method for the picture that application embodiment provides moves to similar scene, such as counter, automated beverage cabinet, gift apparatus, automatic
The scenes such as cabinet adapted for storing books, that is to say, that when based on the appropriate variation of example progress to be used for other scenes, also may belong to this Shen
Spirit or scope please.
The embodiment of the present application also provides a kind of annotation equipments of picture, as shown in figure 5, the annotation equipment 50 can wrap
It includes: identification module 501, sorting module 502 and cutting module 503, wherein
Identification module 501 determines that each single-item in picture to be processed is right respectively for identifying to picture to be processed
The detection block information answered；
Sorting module 502 is based on preset for being ranked up to each corresponding detection block of detection block information
Single-item identifies arrangement information, the corresponding single-item mark of each detection block after determining sequence；
It cuts module 503 to be used for according to each detection block information, is cut out from picture to be processed and each carry single-item
The single-item picture of mark.
In a kind of feasible implementation, identification module 501 is specifically used for the detector by pre-training to figure to be processed
Piece is identified, determines the corresponding detection block information of each single-item in picture to be processed.
In a kind of feasible implementation, which can also include: training module, wherein
Training module is for each in the training sample picture based on acquisition and the correspondence training sample picture marked
The detection block information of single-item, training detector.
In a kind of feasible implementation, detection block information, comprising:
The dimension information of location information of the detection block in picture to be processed, detection block.
In a kind of feasible implementation, sorting module 502 is specifically used for according to each detection block in picture to be processed
Location information, each detection block is ranked up.
In a kind of feasible implementation, which can also include: acquisition module, wherein
Acquisition module is used for through to be processed picture of the picture collection equipment acquisition including multiple single-items, above-mentioned multiple single-items
It is arranged according to single-item mark arrangement information.
In a kind of feasible implementation, acquisition module is specifically used for through picture collection equipment, and acquisition is for above-mentioned more
The picture to be processed of the different angle of a single-item；
Module 503 is cut to be specifically used for cutting out each carry from the picture to be processed of the different angle of multiple single-items
The single-item picture of the different angle of single-item mark.
The annotation equipment of picture provided by the embodiments of the present application, by each detection block information for identifying by figure to be processed
Each single-item in piece, which outlines, to be come, and picture to be processed is cut based on each detection block, can quickly obtain a large amount of single-item figure
Piece greatly improves collecting efficiency, and corresponding based on each detection block without doing individual Image Acquisition to every kind of single-item
Single-item mark can directly obtain the single-item mark of single-item picture, a large amount of mark task be quickly completed, without cumbersome artificial
Mark saves human cost, while reducing mark error probability, improves annotating efficiency.
It is apparent to those skilled in the art that the annotation equipment of picture provided by the embodiments of the present application,
Its realization principle and the technical effect of generation are identical with preceding method embodiment, for convenience and simplicity of description, Installation practice
Part does not refer to place, can refer to corresponding contents in preceding method embodiment, details are not described herein.
The embodiment of the present application also provides a kind of electronic equipment (calculating equipment), as shown in fig. 6, electronics shown in fig. 6 is set
60 include: processor 601 and memory 602, memory 602 be stored at least one instruction, an at least Duan Chengxu, code set or
Instruction set, at least one instruction, an at least Duan Chengxu, code set or instruction set are loaded by processor 601 and are executed to realize
Corresponding contents in preceding method embodiment.
Optionally, electronic equipment 60 can also include transceiver 603.Processor 601 is connected with transceiver 603, such as passes through
Bus 604 is connected.It should be noted that transceiver 603 is not limited to one in practical application, the structure of the electronic equipment 60 is not
Constitute the restriction to the embodiment of the present application.
Wherein, processor 601 can be CPU, general processor, DSP, ASIC, FPGA or other programmable logic devices
Part, transistor logic, hardware component or any combination thereof.It, which may be implemented or executes, combines present disclosure institute
The various illustrative logic blocks of description, module and circuit.Processor 601 is also possible to realize the combination of computing function, example
It is such as combined comprising one or more microprocessors, DSP and the combination of microprocessor etc..
Bus 604 may include an access, and information is transmitted between said modules.Bus 604 can be pci bus or EISA
Bus etc..Bus 604 can be divided into address bus, data/address bus, control bus etc..For convenient for indicating, in Fig. 6 only with one slightly
Line indicates, it is not intended that an only bus or a type of bus.
Memory 602 can be ROM or can store the other kinds of static storage device of static information and instruction, RAM
Or the other kinds of dynamic memory of information and instruction can be stored, it is also possible to EEPROM, CD-ROM or other CDs
Storage, optical disc storage (including compression optical disc, laser disc, optical disc, Digital Versatile Disc, Blu-ray Disc etc.), magnetic disk storage medium
Or other magnetic storage apparatus or can be used in carry or store have instruction or data structure form desired program generation
Code and can by any other medium of computer access, but not limited to this.
Electronic equipment provided by the embodiments of the present application, will be in picture to be processed by each detection block information identified
Each single-item, which outlines, to be come, and picture to be processed is cut based on each detection block, a large amount of single-item picture can be quickly obtained, be not necessarily to
Individual Image Acquisition is done to every kind of single-item, greatly improves collecting efficiency, and be based on the corresponding single-item mark of each detection block
Know, the single-item mark of single-item picture can be directly obtained, quickly complete a large amount of mark task, without cumbersome artificial mark,
Human cost is saved, while reducing mark error probability, improves annotating efficiency.
The embodiment of the present application also provides a kind of computer readable storage mediums, and the computer storage medium is based on storing
The instruction of calculation machine, when run on a computer, allows computer to execute corresponding contents in preceding method embodiment.
It should be understood that although each step in the flow chart of attached drawing is successively shown according to the instruction of arrow,
These steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly stating otherwise herein, these steps
Execution there is no stringent sequences to limit, can execute in the other order.Moreover, at least one in the flow chart of attached drawing
Part steps may include that perhaps these sub-steps of multiple stages or stage are not necessarily in synchronization to multiple sub-steps
Completion is executed, but can be executed at different times, execution sequence, which is also not necessarily, successively to be carried out, but can be with other
At least part of the sub-step or stage of step or other steps executes in turn or alternately.
The above is only some embodiments of the application, it is noted that for the ordinary skill people of the art
For member, under the premise of not departing from the application principle, several improvements and modifications can also be made, these improvements and modifications are also answered
It is considered as the protection scope of the application.
Claims (14)
1. a kind of mask method of picture characterized by comprising
Picture to be processed is identified, determines the corresponding detection block information of each single-item in the picture to be processed；
Each corresponding detection block of detection block information is ranked up, and arrangement information is identified based on preset single-item, really
The corresponding single-item mark of each detection block after fixed sequence；
According to each detection block information, each single-item picture for carrying single-item mark is cut out from the picture to be processed.
2. mask method according to claim 1, which is characterized in that it is described that picture to be processed is identified, determine institute
State the corresponding detection block information of each single-item in picture to be processed, comprising:
Picture to be processed is identified by the detector of pre-training, determines each single-item difference in the picture to be processed
Corresponding detection block information.
3. mask method according to claim 2, which is characterized in that the detector by pre-training is to figure to be processed
Before piece is identified, further includes:
The detection block of training sample picture based on acquisition and each single-item in the correspondence training sample picture marked
Information, the training detector.
4. mask method according to claim 1, which is characterized in that the detection block information, comprising:
The dimension information of location information of the detection block in picture to be processed, detection block.
5. mask method according to claim 4, which is characterized in that described to each corresponding detection block of detection block information
It is ranked up, comprising:
According to location information of each detection block in picture to be processed, each detection block is ranked up.
6. mask method according to claim 1, which is characterized in that it is described picture to be processed is identified before, also
Include:
The picture to be processed including multiple single-items is acquired by picture collection equipment, the multiple single-item is identified according to the single-item
Arrangement information is arranged.
7. mask method according to claim 6, which is characterized in that described acquired by picture collection equipment includes multiple
The picture to be processed of single-item, comprising:
Pass through picture collection equipment, to be processed picture of the acquisition for the different angle of the multiple single-item；
It is described that each single-item picture for carrying single-item mark is cut out from the picture to be processed, comprising:
Each different angle for carrying single-item mark is cut out from the picture to be processed of the different angle of the multiple single-item
Single-item picture.
8. a kind of annotation equipment of picture characterized by comprising
Identification module determines that each single-item in the picture to be processed respectively corresponds for identifying to picture to be processed
Detection block information；
Sorting module for being ranked up to each corresponding detection block of detection block information, and is based on preset single-item mark
Know arrangement information, the corresponding single-item mark of each detection block after determining sequence；
Module is cut, each carries single-item mark for being cut out from the picture to be processed according to each detection block information
The single-item picture of knowledge.
9. annotation equipment according to claim 8, which is characterized in that the detection block information, comprising:
The dimension information of location information of the detection block in picture to be processed, detection block.
10. annotation equipment according to claim 9, which is characterized in that the sorting module is specifically used for according to each inspection
Location information of the frame in picture to be processed is surveyed, each detection block is ranked up.
11. annotation equipment according to claim 8, which is characterized in that further include:
Acquisition module, for by picture collection equipment acquisition include multiple single-items picture to be processed, the multiple single-item according to
It is arranged according to single-item mark arrangement information.
12. annotation equipment according to claim 11, which is characterized in that the acquisition module is specifically used for adopting by picture
Collect equipment, to be processed picture of the acquisition for the different angle of the multiple single-item；
The cutting module is specifically used for cutting out each carrying from the picture to be processed of the different angle of the multiple single-item
The single-item picture for the different angle for thering is single-item to identify.
13. a kind of electronic equipment characterized by comprising
Processor and memory, the memory are stored at least one instruction, at least a Duan Chengxu, code set or instruction set,
At least one instruction, an at least Duan Chengxu, the code set or instruction set loaded by the processor and executed with
Realize the method according to claim 1 to 7.
14. a kind of computer readable storage medium, which is characterized in that the computer storage medium refers to for storing computer
It enables, program, code set or instruction set, when run on a computer, so that computer is executed such as any one of claim 1-7
The method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811623619.2A CN109657681A (en) | 2018-12-28 | 2018-12-28 | Mask method, device, electronic equipment and the computer readable storage medium of picture |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811623619.2A CN109657681A (en) | 2018-12-28 | 2018-12-28 | Mask method, device, electronic equipment and the computer readable storage medium of picture |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109657681A true CN109657681A (en) | 2019-04-19 |
Family
ID=66117900
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811623619.2A Pending CN109657681A (en) | 2018-12-28 | 2018-12-28 | Mask method, device, electronic equipment and the computer readable storage medium of picture |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109657681A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110162649A (en) * | 2019-05-24 | 2019-08-23 | 北京百度网讯科技有限公司 | Sample data acquisition methods obtain system, server and computer-readable medium |
| CN111951554A (en) * | 2020-08-20 | 2020-11-17 | 北京嘀嘀无限科技发展有限公司 | Illegal parking road information acquisition method and system |
| CN112465691A (en) * | 2020-11-25 | 2021-03-09 | 北京旷视科技有限公司 | Image processing method, image processing device, electronic equipment and computer readable medium |
| WO2022266996A1 (en) * | 2021-06-25 | 2022-12-29 | 烟台创迹软件有限公司 | Object detection method and object detection device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105843816A (en) * | 2015-01-15 | 2016-08-10 | 阿里巴巴集团控股有限公司 | Method and device for determining display information of picture |
| CN108256494A (en) * | 2018-01-30 | 2018-07-06 | 厦门华联电子股份有限公司 | A kind of Intelligent storage cabinet and its intelligent interlayer board and intelligent cargo management system |
| CN108647553A (en) * | 2018-05-10 | 2018-10-12 | 上海扩博智能技术有限公司 | Rapid expansion method, system, equipment and the storage medium of model training image |
| CN108900733A (en) * | 2018-07-04 | 2018-11-27 | 百度在线网络技术（北京）有限公司 | Capture apparatus, sync pulse jamming system and method |
| CN108985214A (en) * | 2018-07-09 | 2018-12-11 | 上海斐讯数据通信技术有限公司 | The mask method and device of image data |
| CN108985199A (en) * | 2018-07-02 | 2018-12-11 | 百度在线网络技术（北京）有限公司 | Detection method, device and the storage medium of commodity loading or unloading operation |
-
2018
- 2018-12-28 CN CN201811623619.2A patent/CN109657681A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105843816A (en) * | 2015-01-15 | 2016-08-10 | 阿里巴巴集团控股有限公司 | Method and device for determining display information of picture |
| CN108256494A (en) * | 2018-01-30 | 2018-07-06 | 厦门华联电子股份有限公司 | A kind of Intelligent storage cabinet and its intelligent interlayer board and intelligent cargo management system |
| CN108647553A (en) * | 2018-05-10 | 2018-10-12 | 上海扩博智能技术有限公司 | Rapid expansion method, system, equipment and the storage medium of model training image |
| CN108985199A (en) * | 2018-07-02 | 2018-12-11 | 百度在线网络技术（北京）有限公司 | Detection method, device and the storage medium of commodity loading or unloading operation |
| CN108900733A (en) * | 2018-07-04 | 2018-11-27 | 百度在线网络技术（北京）有限公司 | Capture apparatus, sync pulse jamming system and method |
| CN108985214A (en) * | 2018-07-09 | 2018-12-11 | 上海斐讯数据通信技术有限公司 | The mask method and device of image data |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110162649A (en) * | 2019-05-24 | 2019-08-23 | 北京百度网讯科技有限公司 | Sample data acquisition methods obtain system, server and computer-readable medium |
| CN110162649B (en) * | 2019-05-24 | 2021-06-18 | 北京百度网讯科技有限公司 | Sample data acquisition method, acquisition system, server and computer readable medium |
| CN111951554A (en) * | 2020-08-20 | 2020-11-17 | 北京嘀嘀无限科技发展有限公司 | Illegal parking road information acquisition method and system |
| CN112465691A (en) * | 2020-11-25 | 2021-03-09 | 北京旷视科技有限公司 | Image processing method, image processing device, electronic equipment and computer readable medium |
| WO2022266996A1 (en) * | 2021-06-25 | 2022-12-29 | 烟台创迹软件有限公司 | Object detection method and object detection device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109657681A (en) | Mask method, device, electronic equipment and the computer readable storage medium of picture | |
| CN108320404B (en) | Commodity identification method and device based on neural network and self-service cash register | |
| US11587195B2 (en) | Image processing methods and arrangements useful in automated store shelf inspections | |
| CN106952402A (en) | A kind of data processing method and device | |
| US20190244055A1 (en) | Method and apparatus for checkout based on image identification technique of convolutional neural network | |
| CN110378420A (en) | A kind of image detecting method, device and computer readable storage medium | |
| CN109446883A (en) | Condition of merchandise recognition methods, device, electronic equipment and readable storage medium storing program for executing | |
| CN107477971A (en) | A kind of management method and equipment to food in refrigerator | |
| CN110705424B (en) | Method and device for positioning commodity display position and storage medium | |
| CN108345912A (en) | Commodity rapid settlement system based on RGBD information and deep learning | |
| CN110175590A (en) | A kind of commodity recognition method and device | |
| JP2009187482A (en) | Shelf allocation reproducing method, shelf allocation reproduction program, shelf allocation evaluating method, shelf allocation evaluation program, and recording medium | |
| WO2019075911A1 (en) | Merchandise sorting system and sorting method | |
| CN108830147A (en) | A kind of commodity on shelf price recognition methods based on image recognition, device and system | |
| CN102982332A (en) | Retail terminal goods shelf image intelligent analyzing system based on cloud processing method | |
| AU2020386867A1 (en) | Item identification and tracking system | |
| CN109271935A (en) | The matching method of article and electronic tag, apparatus and system | |
| US11514665B2 (en) | Mapping optical-code images to an overview image | |
| CN108596137A (en) | A kind of commodity scanning record method based on image recognition algorithm | |
| CN109410211A (en) | The dividing method and device of target object in a kind of image | |
| CN109977983A (en) | Obtain the method and device of training image | |
| CN108960238A (en) | Commodity identifying processing method and apparatus based on multi-cam | |
| US20140086508A1 (en) | Semantic Theme Based Shape Collage Representation for an Image Collection | |
| CN111881894A (en) | Method, system, equipment and storage medium for collecting goods selling information of container | |
| US11776673B2 (en) | System and method for augmented reality detection of loose pharmacy items |
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 |