CN110059750A - House type shape recognition process, device and equipment - Google Patents
House type shape recognition process, device and equipment Download PDFInfo
- Publication number
- CN110059750A CN110059750A CN201910317211.0A CN201910317211A CN110059750A CN 110059750 A CN110059750 A CN 110059750A CN 201910317211 A CN201910317211 A CN 201910317211A CN 110059750 A CN110059750 A CN 110059750A
- Authority
- CN
- China
- Prior art keywords
- floor plan
- network
- house type
- type shape
- sub
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of house type shape recognition process, device and equipment, are related to the technical field of image procossing, this method comprises: obtaining floor plan to be identified；Floor plan is input in the house type shape recognition model pre-established, exports the recognition result of floor plan；Wherein, which is established by following manner: obtaining the floor plan sample and corresponding identification information of setting quantity；Establish the network structure of convolutional neural networks；Floor plan sample, house type shape information and the structural information being labeled in floor plan sample are input to convolutional neural networks to be trained；The neural network that training is completed carries out tuning processing, obtain the house type shape recognition model of training completion, it solves existing in the prior art when floor plan quantity is larger, using the low technical problem of existing manual type recognition efficiency and accuracy, the efficiency and accuracy of house type shape recognition are improved.
Description
Technical field
The present invention relates to technical field of image processing, more particularly, to a kind of house type shape recognition process, device and equipment.
Background technique
House ornamentation design refers to that house decoration designs, and the planning carried out in function pattern before family formally decoratees and goes into operation is set
Meter, the Art Design at each space interface.It is in installing meter, needs to comprehensively consider floor plan room shape, floor plan room shape
Shape specifically includes that square, rectangle, L shape, polygon, spill, convex etc..
Before house ornamentation design, it is necessary first to the geometric vector data of floor plan room shape image are obtained, it is presently mainly logical
It crosses manual identified and manually generates these data, when floor plan quantity is larger, using the recognition efficiency and standard of existing manual type
True property is low.
Summary of the invention
The purpose of the present invention is to provide a kind of house type shape recognition process and devices, to improve the effect of house type shape recognition
Rate and accuracy.
A kind of house type shape recognition process provided by the invention, which comprises obtain floor plan to be identified；By institute
It states floor plan to be input in the house type shape recognition model pre-established, exports the recognition result of the floor plan；The house type
Shape recognition model is established by following manner: obtaining the floor plan sample and corresponding identification information of setting quantity；The mark
Knowing information includes house type shape information, and the structural information being labeled in the floor plan sample；Establish convolutional neural networks
Network structure；The network structure includes the first sub-network and the second sub-network interconnected；By the house type pattern
This and the structural information being labeled in the floor plan sample are input in first sub-network and are trained；By the house type
Pattern sheet, house type shape information and the structural information being labeled in the floor plan sample are input to second sub-network
In be trained；First sub-network and second sub-network that training is completed carry out tuning processing, obtain having trained
At house type shape recognition model.
Further, the first sub-network in the house type shape recognition model includes two layers of convolutional layer, for extracting family
The structural information of type figure；The structural information includes dot position information and side location information；Second sub-network includes recurrence
The tensor is mapped as including classification information by sub-network and convolutional layer for the structural information to be encoded to tensor
Vector.
Further, the floor plan is input to before the step in the house type shape recognition model pre-established, institute
State method further include: sampling processing is carried out to the floor plan, the floor plan after being sampled；To the family after sampling
Type figure is normalized, the floor plan after obtaining normalized.
Further, after the step of obtaining the house type shape recognition model of training completion, the method also includes: it calculates
The corresponding recognition result of the floor plan sample of the house type shape recognition model output；It is corresponding according to the floor plan sample
Identification information and the recognition result, calculate the recognition accuracy of the floor plan sample；Judging the recognition accuracy is
It is no to meet preset threshold；If not, the step of continuing to execute the floor plan sample and corresponding identification information that obtain setting quantity,
Until the recognition accuracy meets preset threshold.
A kind of house type shape recognition device provided by the invention, described device includes: acquisition module, to be identified for obtaining
Floor plan；As a result output module is exported for being input to the floor plan in the house type shape recognition model pre-established
The recognition result of the floor plan；The house type shape recognition model is established by following manner: obtaining the house type of setting quantity
Pattern sheet and corresponding identification information；The identification information includes house type shape information, and is labeled in the floor plan sample
In structural information；Establish the network structure of convolutional neural networks；The network structure includes the first subnet interconnected
Network and the second sub-network；The floor plan sample and the structural information that is labeled in the floor plan sample are input to described the
It is trained in one sub-network；By the floor plan sample, house type shape information and it is labeled in the floor plan sample
Structural information is input in second sub-network and is trained；First sub-network that training is completed and second son
Network carries out tuning processing, obtains the house type shape recognition model of training completion.
Further, the first sub-network in the house type shape recognition model includes two layers of convolutional layer, for extracting family
The structural information of type figure；The structural information includes dot position information and side location information；Second sub-network includes recurrence
The tensor is mapped as including classification information by sub-network and convolutional layer for the structural information to be encoded to tensor
Vector.
Further, described device further include: sampling processing module is obtained for carrying out sampling processing to the floor plan
The floor plan after to sampling；Normalized module is obtained for the floor plan after sampling to be normalized
The floor plan after to normalized.
Further, described device further include: the first computing module, for calculating the house type shape recognition model output
The corresponding recognition result of the floor plan sample；Second computing module, for according to the corresponding mark of the floor plan sample
Information and the recognition result calculate the recognition accuracy of the floor plan sample；Judgment module, for judging that the identification is quasi-
Whether true rate meets preset threshold；Execution module, for obtaining the setting floor plan sample of quantity and right if not, continuing to execute
The step of identification information answered, until the recognition accuracy meets preset threshold.
A kind of house type shape recognition equipment provided by the invention, including memory, processor are stored in the memory
The computer program that can be run on the processor, the processor realize above-mentioned house type shape when executing the computer program
Shape recognition methods.
A kind of machine readable storage medium provided by the invention, it is executable that the machine readable storage medium is stored with machine
Instruction, for the machine-executable instruction when being called and being executed by processor, the machine-executable instruction promotes the processing
Device realizes above-mentioned house type shape recognition process.
House type shape recognition process, device, equipment and machine readable storage medium provided by the invention, the house type shape
Identification model is established by convolutional neural networks training, and the network structure includes the first sub-network interconnected and the
Two sub-networks export the house type by being input to floor plan to be identified in the house type shape recognition model pre-established
The recognition result of figure, which is used to be identified comprising the convolutional neural networks of the first sub-network interconnected and the second sub-network
House type shape can identify different house type shapes, and manual intervention is not necessarily in identification process, improve house type shape recognition
Efficiency and accuracy.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of flow chart of house type shape recognition process provided in an embodiment of the present invention；
Fig. 2 is the specific flow chart provided in an embodiment of the present invention for establishing house type shape recognition model；
Fig. 3 is the network structure of convolutional neural networks provided in an embodiment of the present invention；
Fig. 4 is the flow chart of another house type shape recognition process provided in an embodiment of the present invention；
Fig. 5 is the training method flow chart of another specific network structure provided in an embodiment of the present invention；
Fig. 6 is a kind of structural schematic diagram of house type shape recognition device provided in an embodiment of the present invention；
Fig. 7 is a kind of structural schematic diagram of house type shape recognition equipment provided in an embodiment of the present invention.
Specific embodiment
Technical solution of the present invention is clearly and completely described below in conjunction with embodiment, it is clear that described reality
Applying example is a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, the common skill in this field
Art personnel every other embodiment obtained without making creative work belongs to the model that the present invention protects
It encloses.
It is in installing meter, floor plan room shape specifically includes that square, rectangle, L shape, polygon, spill, convex
Shape；Each shape is different for house ornamentation design；Before doing house ornamentation design, need to obtain the geometry of house type image
Vector data, presently mainly by manually generating these data, the recognition efficiency and accuracy of the manual type are difficult to full
Sufficient user demand.
Based on this, the embodiment of the invention provides a kind of house type shape recognition process, device and equipment；The technology can answer
For that can also be applied in the terminal devices such as mobile phone, tablet computer, for identification house type shape in server.The technology can
To use relevant software or hardware realization, it is described below by embodiment.
The embodiment of the present invention provides a kind of house type shape recognition process, as shown in Figure 1, method includes the following steps:
Step S101 obtains floor plan to be identified.
Specifically, floor plan to be identified is usually the floor plan etc. for being provided by client or being extracted from database.
Floor plan is input in the house type shape recognition model pre-established, exports the identification of floor plan by step S102
As a result.
Further, floor plan is input in the house type shape recognition model pre-established, can identifies to obtain house type
The information such as the shape and geometric vector parameter of figure, wherein the shape of floor plan includes: bedroom, objective dining room, study, toilet, sun
The shape of the difference house type classification such as platform；Geometric vector parameter includes: the position on each side in floor plan, point position, turning position
Set etc. and different shape corresponding to the geometric parameters such as size.
The specific flow chart for establishing house type shape recognition model shown in Figure 2, the house type shape recognition model pass through
Following manner is established:
Step S201 obtains the floor plan sample and corresponding identification information of setting quantity；The identification information includes house type
Shape information, and the structural information being labeled in floor plan sample.
In the present embodiment, according to house type classification, such as: bedroom, objective dining room, study, toilet, balcony, uniform sampling converge
Always go out floor plan sample of different shapes.The quantity of floor plan sample can be set according to user demand, in the present embodiment, from family
The above-mentioned floor plan sample of different shapes of specified quantity is extracted in type chart database, such as 120,000, while obtaining floor plan
The corresponding identification information of sample, comprising: house type shape information, and the structural information etc. being labeled in floor plan sample.Wherein,
House type shape information is specifically as follows square, rectangle, L shape, polygon, spill, convex etc.；Structural information is specifically as follows
Position, point position, the corner location on each side in floor plan etc..
Step S202 establishes the network structure of convolutional neural networks；The network structure includes the first son interconnected
Network and the second sub-network.
Wherein, the first sub-network and the second sub-network can be made of one or more layers convolutional layer, and each convolutional layer can be with
Corresponding convolution kernel is set, to extract the feature of corresponding type or corresponding granularity.For example, convolutional layer in the first sub-network can be with
For extracting the feature of more details, such as position, the point position, corner location on each side in floor plan, these details are special
Sign is referred to as the structural information of floor plan；What the convolutional layer in the second sub-network can be used for extracting the first sub-network
Minutia carries out further induction-arrangement, and carries out classification processing based on shape of these minutias to floor plan.
Floor plan sample and the structural information being labeled in floor plan sample are input in the first sub-network by step S203
It is trained.
As described above, the first sub-network is used to extract the minutia of floor plan, and then in the first sub-network of training, need
By above-mentioned floor plan sample and the structural information being labeled in floor plan sample, so that the first sub-network trained can be quasi-
Really extract the structural information of floor plan.
Step S204, floor plan sample, house type shape information and the structural information that is labeled in floor plan sample is defeated
Enter into the second sub-network and is trained.
As described above, the second sub-network is whole for further being concluded to the minutia that the first sub-network extracts
Reason, and classified finishing is carried out based on shape of these minutias to floor plan, and then in the second sub-network of training, need
Floor plan sample, house type shape information and the structural information being labeled in floor plan sample are stated, so that the second son trained
Network can the shape to floor plan accurately sorted out.
Step S205, the first sub-network and the second sub-network that training is completed carry out tuning processing, obtain training completion
House type shape recognition model.
Floor plan is input to the first sub-network of training completion and the second sub-network continues to train, with what is completed to training
First sub-network and the second sub-network are finely adjusted, so that each layer of network weight weight values are attained by optimal value, are finally obtained
The house type shape recognition model that training is completed.
The embodiment of the invention provides a kind of house type shape recognition process, device and equipment, firstly, obtaining family to be identified
Then type figure establishes house type shape recognition model, know finally, floor plan to be identified is input to established house type shape
In other model, the recognition result of floor plan is exported, which, which uses, includes the first sub-network interconnected and the second sub-network
Convolutional neural networks identify house type shape, can identify different house type shapes, and in identification process be not necessarily to manual intervention, mention
The high efficiency and accuracy of house type shape recognition.
Based on above-mentioned house type shape recognition process, the present embodiment also provides a kind of network knot of specific convolutional neural networks
Structure is used for as shown in figure 3, the first two structure sheaf is the first sub-network, including convolutional layer 1 and convolutional layer 2 in the network structure
The structural information of the floor plan of input is extracted, which includes dot position information and side location information；Second sub-network packet
Recurrence subnet network layers and convolutional layer 3 are included, for structural information to be encoded to tensor, tensor is passed through into the full connection of full articulamentum again
Operation, be finally mapped as include classification information vector, recurrence subnet network layers realization the output of recurrence sub-network node is returned
The input terminal for returning to recurrence sub-network node moves in circles to carry out depth training, the above convolutional layer 1, convolutional layer 2, convolutional layer
3, include excitation function to assist expression complex characteristic in recurrence subnet network layers and full articulamentum, be similar to other deep learnings
Algorithm, convolutional neural networks are usually using line rectification unit (Rectified Linear Unit, abbreviation ReLU) etc..
Based on the network structure of above-mentioned convolutional neural networks, the present embodiment also provides a kind of training of specific network structure
Method, this method include the following steps:
Step 1,5 parts will be divided by the random uniform sampling of extracted 120,000 floor plans from house type chart database, chosen
It is wherein used as training sample for 4 parts, is in addition used as test sample for 1 part, per individually a about 2.4 ten thousand floor plans.
Step 2, neural network is trained by training sample, including two steps of pre-training and tuning, firstly, will
The structural information of floor plan sample and floor plan, comprising: dot position information and side location information are input in neural network structure
The first sub-network, two layers of convolutional layer of the first sub-network is trained, then, by floor plan sample, house type shape information
And the structural information cooperation picture classification markup information of floor plan is input to the second sub-network in neural network structure, to second
Sub-network is trained, finally, two parts by separated training combine again, is carried out last tuning training, is trained
The house type shape recognition model of completion.
Step 3, test sample is input in trained house type shape recognition model and carries out identification test, determine identification
As a result.
Another house type shape recognition process that the embodiment of the present invention also provides, as shown in figure 4, this method includes following step
It is rapid:
Step S401 obtains floor plan to be identified.
In the present embodiment, floor plan to be identified is the floor plan etc. extracted from database, wherein every floor plan
Pixel is 128*128.
Step S402 carries out sampling processing to floor plan, the floor plan after being sampled.
It is adopted for example, carrying out drop to all floor plans that original pixels extracted from house type chart database are 128*128
Sample obtains the image pattern of three kinds of different resolution sizes of 16*16,32*32,64*64 to reduce the size of floor plan,
The format of these image patterns can be png format.
Step S403, is normalized the floor plan after sampling, the floor plan after obtaining normalized.
Using obtained every image pattern in opencv parsing previous step, the picture of R, G, B triple channel is obtained, benefit
With gray processing algorithm, image pattern is converted into single channel image, and the pixel value of single channel image is normalized,
Variance after normalization is 1, mean value 0.
Floor plan is input in the house type shape recognition model pre-established, exports the identification of floor plan by step S404
As a result.
Further, floor plan is input in the house type shape recognition model pre-established, can identifies to obtain difference
Corresponding to the shape of house type classification, the position on each side in floor plan, point position, corner location etc. and different shape
The geometric parameters such as size.
The embodiment of the invention provides another house type shape recognition process, firstly, floor plan to be identified is obtained, to institute
The floor plan of acquisition carries out sampling and then normalized establishes house type shape recognition model, finally, house type to be identified
Figure is input in the house type shape recognition model after over-sampling and normalized, can obtain house type shape recognition result,
Which, which is used, identifies house type shape comprising the convolutional neural networks of the first sub-network interconnected and the second sub-network, can be with
It identifies different house type shapes, and is not necessarily to manual intervention in identification process, improve the efficiency and accuracy of house type shape recognition.
The present embodiment also provides the training method of another specific network structure, as shown in figure 5, this method includes following
Step:
Step S501 obtains the floor plan sample and corresponding identification information of setting quantity；Identification information includes house type shape
Shape information, and the structural information being labeled in floor plan sample.
During the training house type shape recognition model, what is extracted from house type chart database is square, rectangular
Shape, L shape, polygon, spill, the floor plan sample of six kinds of shapes of convex are 120,000 total, each has about 20,000 images.
Step S502 establishes the network structure of convolutional neural networks；Network structure includes the first subnet interconnected
Network and the second sub-network.
Wherein, the first sub-network and the second sub-network can be made of one or more layers convolutional layer, and each convolutional layer can be with
Corresponding convolution kernel is set, to extract the feature of corresponding type or corresponding granularity, for example, convolutional layer in the first sub-network can be with
Convolutional layer for extracting the feature of the more details such as each side in floor plan, point position, corner location, in the second sub-network
Can minutia that further the first sub-network of induction-arrangement is extracted, and classification processing is carried out to the shape of floor plan.
Floor plan sample and the structural information being labeled in floor plan sample are input in the first sub-network by step S503
It is trained.
As described above, the first sub-network can extract the more details such as each side, point position, corner location in floor plan
Feature, it is therefore desirable to above-mentioned minutia is input to the first sub-network, so that the first sub-network trained can be accurate
Extract the structural information of floor plan.
Step S504, floor plan sample, house type shape information and the structural information that is labeled in floor plan sample is defeated
Enter into the second sub-network and is trained.
As described above, convolutional layer in the second sub-network can details that further the first sub-network of induction-arrangement is extracted it is special
Sign, and classification processing is carried out to the shape of floor plan, it is therefore desirable to above-mentioned minutia is input to the second sub-network, so that instruction
The second sub-network practised can the shape to floor plan accurately sorted out.
5 parts will be divided by the random uniform sampling of extracted 120,000 floor plans from house type chart database, and choose wherein 4 parts
As training sample, in addition test sample is used as 1 part, per individually a about 2.4 ten thousand floor plans.Using sample average
Sampling distribution be sampled, the sampling distribution using sample average is the distribution that all sample average is formed, the i.e. probability of μ
Distribution, the sampling distribution of sample average are symmetrically, with the increase of sample size n, no matter original is generally in shape
No Normal Distribution, the sampling distribution of sample average will all tend to normal distribution, and the mathematic expectaion of distribution is population mean
μ, variance are the 1/n of population variance.
Based on deep learning Vgg16 network, house type diagram shape classification is learnt using deep neural network disaggregated model, will be instructed
Practice sample for the training of depth convolutional neural networks, the structure of pre-training model can be used, be first randomized all weights,
Then it being trained according to the data set of oneself, another kind is that the training of part is carried out to it using the method for pre-training model,
Specific way is to remain unchanged some layers of the weight that model originates, the subsequent layer of re -training obtains new weight.
In the process, it can repeatedly be attempted, so as to find frozen crust frozen layers and retraining layer according to result
Best collocation between retrain layers.
Step S505, the first sub-network and the second sub-network that training is completed carry out tuning processing, obtain training completion
House type shape recognition model.
The training of above-mentioned house type shape recognition model is divided into two processes, is matched first by the training sample that above-mentioned steps generate
Close the point manually marked, side information training the first two structure sheaf, the figure with the graph structure information that the point that manually marks, side are constituted
Piece classification markup information latter two structure sheaf of training and full articulamentum, then two parts of separated training are combined, it carries out
Last tuning training.
Step S506 calculates the corresponding recognition result of floor plan sample of house type shape recognition model output.
It is similar with training sample is imported, test sample is read in into memory first, is then recycled with for successively by test data
Afferent nerve network calls query () function, and the corresponding label of output array maximum value is compared and recorded with data label
Test result.
Step S507 calculates the identification of floor plan sample according to the corresponding identification information of floor plan sample and recognition result
Accuracy rate.
The test result according to obtained in above-mentioned steps, finally calculates accuracy rate.This time parameters of test are as follows: study
Rate 0.3, test result: accuracy rate 0.9853, it is 50 seconds time-consuming.
Step S508, judges whether recognition accuracy meets preset threshold.If so, terminating；If not, executing step
S509。
Step S509, if not, continuing to execute the step of the floor plan sample and corresponding identification information that obtain setting quantity
Suddenly, until recognition accuracy meets preset threshold.
The depth convolutional neural networks that test sample is input to training completion are tested, using cross validation mode,
Extracted whole random uniform samplings of sample image are divided into 5 parts, use 4 parts of data therein as training sample every time
This, is in addition used as test sample user test sample for 1 part, and such identification experiment is repeated 5 times, and 5 average value is finally taken to make
For recognition result, for evaluating whether reasonability and the house type shape recognition process of designed network reach beyond artificial judgment
Standard.
The training method of another specific network structure provided in an embodiment of the present invention, firstly, obtaining setting quantity
Then floor plan sample and corresponding identification information establish the network structure of convolutional neural networks, and by floor plan sample, family
Type shape information and the structural information etc. being labeled in floor plan sample, which are input in neural network, is trained simultaneously tuning,
The house type shape recognition model of training completion is obtained, finally, the floor plan sample for calculating the output of house type shape recognition model is corresponding
Recognition result, judge whether the recognition accuracy of established convolutional neural networks meets preset threshold, by network knot
The continuous training of structure and tuning make the recognition accuracy for the convolutional neural networks finally established and recognition efficiency meet setting
Standard.
The structural schematic diagram of a kind of house type shape recognition device provided in an embodiment of the present invention, as shown in fig. 6, device packet
It includes:
Module 10 is obtained, for obtaining floor plan to be identified.
As a result output module 20 export family for being input to floor plan in the house type shape recognition model pre-established
The recognition result of type figure；Wherein, house type shape recognition model is established by following manner: obtaining the floor plan sample of setting quantity
With corresponding identification information；The identification information includes house type shape information, and the structural information being labeled in floor plan sample；
Establish the network structure of convolutional neural networks；Network structure includes the first sub-network and the second sub-network interconnected；It will
Floor plan sample and the structural information being labeled in floor plan sample, which are input in the first sub-network, to be trained；By house type pattern
Originally, house type shape information and the structural information being labeled in floor plan sample are input in the second sub-network and are trained；It will
The first sub-network and the second sub-network that training is completed carry out tuning processing, obtain the house type shape recognition model of training completion.
The first sub-network in house type shape recognition model includes two layers of convolutional layer, and the structure for extracting floor plan is believed
Breath；Structural information includes dot position information and side location information；Second sub-network includes recurrence sub-network and convolutional layer, and being used for will
Structural information is encoded to tensor, tensor is mapped as include classification information vector.
Further, device further include: sampling processing module is sampled for carrying out sampling processing to floor plan
Floor plan afterwards；Normalized module, for the floor plan after sampling to be normalized, after obtaining normalized
Floor plan.
In the present embodiment, the device further include: the first computing module, for calculating the family of house type shape recognition model output
The corresponding recognition result of type pattern sheet；Second computing module, for being tied according to the corresponding identification information of floor plan sample and identification
Fruit calculates the recognition accuracy of floor plan sample；Judgment module, for judging whether recognition accuracy meets preset threshold；It holds
Row module, for if not, continue to execute obtain setting quantity floor plan sample and corresponding identification information the step of, until
Recognition accuracy meets preset threshold.
The embodiment of the invention provides a kind of structural schematic diagrams of house type shape recognition equipment, as shown in fig. 7, the equipment packet
Memory 31, processor 32 are included, the computer program that can be run on the processor, the place are stored in the memory
The step of reason device realizes above-mentioned house type shape recognition process when executing the computer program.
Referring to Fig. 7, electronic equipment further include: bus 33 and communication interface 34, processor 32, communication interface 34 and memory
31 are connected by bus 33；Processor 32 is for executing the executable module stored in memory 31, such as computer program.
Wherein, memory 31 may include high-speed random access memory (RAM, Random Access Memory),
It may further include nonvolatile memory (non-volatile memory), for example, at least a magnetic disk storage.By at least
One communication interface 34 (can be wired or wireless) realizes the communication between the system network element and at least one other network element
Connection, can be used internet, wide area network, local network, Metropolitan Area Network (MAN) etc..
Bus 33 can be isa bus, pci bus or eisa bus etc..The bus can be divided into address bus, data
Bus, control bus etc..Only to be indicated with a four-headed arrow convenient for indicating, in Fig. 7, it is not intended that an only bus or
A type of bus.
Wherein, memory 31 is for storing program, and the processor 32 executes the journey after receiving and executing instruction
Sequence, method performed by the device that the process that aforementioned any embodiment of the present invention discloses defines can be applied in processor 32,
Or it is realized by processor 32.
Processor 32 may be a kind of IC chip, the processing capacity with signal.During realization, above-mentioned side
Each step of method can be completed by the integrated logic circuit of the hardware in processor 32 or the instruction of software form.Above-mentioned
Processor 32 can be general processor, including central processing unit (Central Processing Unit, abbreviation CPU), network
Processor (Network Processor, abbreviation NP) etc.；It can also be digital signal processor (Digital Signal
Processing, abbreviation DSP), specific integrated circuit (Application Specific Integrated Circuit, referred to as
ASIC), ready-made programmable gate array (Field-Programmable Gate Array, abbreviation FPGA) or other are programmable
Logical device, discrete gate or transistor logic, discrete hardware components.It may be implemented or execute in the embodiment of the present invention
Disclosed each method, step and logic diagram.General processor can be microprocessor or the processor is also possible to appoint
What conventional processor etc..The step of method in conjunction with disclosed in the embodiment of the present invention, can be embodied directly in hardware decoding processing
Device executes completion, or in decoding processor hardware and software module combination execute completion.Software module can be located at
Machine memory, flash memory, read-only memory, programmable read only memory or electrically erasable programmable memory, register etc. are originally
In the storage medium of field maturation.The storage medium is located at memory 31, and processor 32 reads the information in memory 31, in conjunction with
Its hardware completes the step of above method.
The embodiment of the invention also provides a kind of machine readable storage medium, which is stored with machine
Executable instruction, for the machine-executable instruction when being called and being executed by processor, machine-executable instruction promotes processor real
Existing above-mentioned house type shape recognition process, specific implementation can be found in method implementation, and details are not described herein.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (10)
1. a kind of house type shape recognition process, which is characterized in that the described method includes:
Obtain floor plan to be identified；
The floor plan is input in the house type shape recognition model pre-established, exports the recognition result of the floor plan；
The house type shape recognition model is established by following manner:
Obtain the floor plan sample and corresponding identification information of setting quantity；The identification information includes house type shape information, with
And it is labeled in the structural information in the floor plan sample；
Establish the network structure of convolutional neural networks；The network structure includes the first sub-network interconnected and the second son
Network；
The floor plan sample and the structural information being labeled in the floor plan sample are input in first sub-network
It is trained；
The floor plan sample, house type shape information and the structural information being labeled in the floor plan sample are input to
It is trained in second sub-network；
First sub-network and second sub-network that training is completed carry out tuning processing, obtain the house type of training completion
Shape recognition model.
2. the method according to claim 1, wherein the first sub-network packet in the house type shape recognition model
Two layers of convolutional layer is included, for extracting the structural information of floor plan；The structural information includes dot position information and side location information；
Second sub-network includes recurrence sub-network and convolutional layer, will be described for the structural information to be encoded to tensor
Tensor be mapped as include classification information vector.
3. the method according to claim 1, wherein the floor plan to be input to the house type shape pre-established
Before step in identification model, the method also includes:
Sampling processing is carried out to the floor plan, the floor plan after being sampled；
The floor plan after sampling is normalized, the floor plan after obtaining normalized.
4. the method according to claim 1, wherein the step of obtaining the house type shape recognition model of training completion
Later, the method also includes:
Calculate the corresponding recognition result of the floor plan sample of the house type shape recognition model output；
According to the corresponding identification information of the floor plan sample and the recognition result, the identification for calculating the floor plan sample is quasi-
True rate；
Judge whether the recognition accuracy meets preset threshold；
If not, the step of continuing to execute the floor plan sample and corresponding identification information that obtain setting quantity, until the knowledge
Other accuracy rate meets preset threshold.
5. a kind of house type shape recognition device, which is characterized in that described device includes:
Module is obtained, for obtaining floor plan to be identified；
As a result output module, for being input to the floor plan in the house type shape recognition model pre-established, described in output
The recognition result of floor plan；
The house type shape recognition model is established by following manner:
Obtain the floor plan sample and corresponding identification information of setting quantity；The identification information includes house type shape information, with
And it is labeled in the structural information in the floor plan sample；
Establish the network structure of convolutional neural networks；The network structure includes the first sub-network interconnected and the second son
Network；
The floor plan sample and the structural information being labeled in the floor plan sample are input in first sub-network
It is trained；
The floor plan sample, house type shape information and the structural information being labeled in the floor plan sample are input to
It is trained in second sub-network；
First sub-network and second sub-network that training is completed carry out tuning processing, obtain the house type of training completion
Shape recognition model.
6. device according to claim 5, which is characterized in that the first sub-network packet in the house type shape recognition model
Two layers of convolutional layer is included, for extracting the structural information of floor plan；The structural information includes dot position information and side location information；
Second sub-network includes recurrence sub-network and convolutional layer, will be described for the structural information to be encoded to tensor
Tensor be mapped as include classification information vector.
7. device according to claim 5, which is characterized in that described device further include:
Sampling processing module, for carrying out sampling processing to the floor plan, the floor plan after being sampled；
Normalized module, for the floor plan after sampling to be normalized, after obtaining normalized
The floor plan.
8. device according to claim 5, which is characterized in that described device further include:
First computing module, the corresponding identification knot of the floor plan sample for calculating the house type shape recognition model output
Fruit；
Second computing module, for according to the corresponding identification information of the floor plan sample and the recognition result, described in calculating
The recognition accuracy of floor plan sample；
Judgment module, for judging whether the recognition accuracy meets preset threshold；
Execution module, for if not, continuing to execute the step of the floor plan sample and corresponding identification information that obtain setting quantity
Suddenly, until the recognition accuracy meets preset threshold.
9. a kind of house type shape recognition equipment, including memory, processor, being stored in the memory can be in the processor
The computer program of upper operation, which is characterized in that the processor realizes the claims 1 when executing the computer program
The step of to 4 described in any item methods.
10. a kind of machine readable storage medium, which is characterized in that the machine readable storage medium is stored with the executable finger of machine
It enables, for the machine-executable instruction when being called and being executed by processor, the machine-executable instruction promotes the processor
Realize the described in any item methods of Claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910317211.0A CN110059750A (en) | 2019-04-17 | 2019-04-17 | House type shape recognition process, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910317211.0A CN110059750A (en) | 2019-04-17 | 2019-04-17 | House type shape recognition process, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110059750A true CN110059750A (en) | 2019-07-26 |
Family
ID=67319748
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910317211.0A Pending CN110059750A (en) | 2019-04-17 | 2019-04-17 | House type shape recognition process, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110059750A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108664860A (en) * | 2017-04-01 | 2018-10-16 | 中山市琪朗灯饰厂有限公司 | The recognition methods of room floor plan and device |
| CN110633553A (en) * | 2019-10-09 | 2019-12-31 | 郑豪 | Automatic generation method and system for residential house type plane graph |
| CN110728235A (en) * | 2019-10-11 | 2020-01-24 | 广东三维家信息科技有限公司 | House type area marking method and device |
| CN110765291A (en) * | 2019-10-28 | 2020-02-07 | 广东三维家信息科技有限公司 | Retrieval method and device and electronic equipment |
| CN110795858A (en) * | 2019-11-06 | 2020-02-14 | 广东博智林机器人有限公司 | Method and device for generating home decoration design drawing |
| CN111144243A (en) * | 2019-12-13 | 2020-05-12 | 江苏艾佳家居用品有限公司 | Household pattern recognition method and device based on counterstudy |
| CN111177821A (en) * | 2019-12-02 | 2020-05-19 | 江苏艾佳家居用品有限公司 | Automatic layout method and system based on house type graph splicing |
| CN111259734A (en) * | 2020-01-08 | 2020-06-09 | 深圳市彬讯科技有限公司 | Spatial house type identification method and device, computer equipment and storage medium |
| CN111340954A (en) * | 2020-02-18 | 2020-06-26 | 广东三维家信息科技有限公司 | House type wall drawing method and model training method and device thereof |
| CN111339336A (en) * | 2020-03-11 | 2020-06-26 | 贝壳技术有限公司 | Analysis method and system for house type data |
| CN111582358A (en) * | 2020-04-30 | 2020-08-25 | 上海添玑网络服务有限公司 | Training method and device for house type recognition model and house type weight judging method and device |
| CN112116613A (en) * | 2020-09-25 | 2020-12-22 | 贝壳技术有限公司 | Model training method, image segmentation method, image vectorization method and system thereof |
| CN113591929A (en) * | 2021-07-05 | 2021-11-02 | 华南师范大学 | Family pattern recognition method combining direction sensing kernel cluster |
| WO2021217340A1 (en) * | 2020-04-27 | 2021-11-04 | Li Jianjun | Ai-based automatic design method and apparatus for universal smart home scheme |
| CN114401520A (en) * | 2021-12-28 | 2022-04-26 | 广西壮族自治区公众信息产业有限公司 | Method and system for detecting thermodynamic diagram of wireless network signal |
| CN114973297A (en) * | 2022-06-17 | 2022-08-30 | 广州市圆方计算机软件工程有限公司 | Wall area identification method, system, equipment and medium for planar house type graph |
| CN116502546A (en) * | 2023-06-29 | 2023-07-28 | 深圳市明源云科技有限公司 | House type design method, house type design device, electronic equipment and readable storage medium |
| CN116578915A (en) * | 2023-04-11 | 2023-08-11 | 广州极点三维信息科技有限公司 | Structured house type analysis method and system based on graphic neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844614A (en) * | 2017-01-18 | 2017-06-13 | 天津中科智能识别产业技术研究院有限公司 | A kind of floor plan functional area system for rapidly identifying |
| CN108009598A (en) * | 2017-12-27 | 2018-05-08 | 北京诸葛找房信息技术有限公司 | Floor plan recognition methods based on deep learning |
| CN108764022A (en) * | 2018-04-04 | 2018-11-06 | 链家网（北京）科技有限公司 | A kind of image-recognizing method and system |
| US20180365496A1 (en) * | 2017-06-17 | 2018-12-20 | Matterport, Inc. | Automated classification based on photo-realistic image/model mappings |
-
2019
- 2019-04-17 CN CN201910317211.0A patent/CN110059750A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844614A (en) * | 2017-01-18 | 2017-06-13 | 天津中科智能识别产业技术研究院有限公司 | A kind of floor plan functional area system for rapidly identifying |
| US20180365496A1 (en) * | 2017-06-17 | 2018-12-20 | Matterport, Inc. | Automated classification based on photo-realistic image/model mappings |
| CN108009598A (en) * | 2017-12-27 | 2018-05-08 | 北京诸葛找房信息技术有限公司 | Floor plan recognition methods based on deep learning |
| CN108764022A (en) * | 2018-04-04 | 2018-11-06 | 链家网（北京）科技有限公司 | A kind of image-recognizing method and system |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108664860B (en) * | 2017-04-01 | 2022-02-01 | 中山市琪朗灯饰厂有限公司 | Method and device for identifying room plan |
| CN108664860A (en) * | 2017-04-01 | 2018-10-16 | 中山市琪朗灯饰厂有限公司 | The recognition methods of room floor plan and device |
| CN110633553A (en) * | 2019-10-09 | 2019-12-31 | 郑豪 | Automatic generation method and system for residential house type plane graph |
| CN110728235A (en) * | 2019-10-11 | 2020-01-24 | 广东三维家信息科技有限公司 | House type area marking method and device |
| CN110765291A (en) * | 2019-10-28 | 2020-02-07 | 广东三维家信息科技有限公司 | Retrieval method and device and electronic equipment |
| CN110795858A (en) * | 2019-11-06 | 2020-02-14 | 广东博智林机器人有限公司 | Method and device for generating home decoration design drawing |
| CN110795858B (en) * | 2019-11-06 | 2023-04-07 | 广东博智林机器人有限公司 | Method and device for generating home decoration design drawing |
| CN111177821A (en) * | 2019-12-02 | 2020-05-19 | 江苏艾佳家居用品有限公司 | Automatic layout method and system based on house type graph splicing |
| CN111177821B (en) * | 2019-12-02 | 2022-07-08 | 江苏艾佳家居用品有限公司 | Automatic layout method and system based on house type graph splicing |
| CN111144243A (en) * | 2019-12-13 | 2020-05-12 | 江苏艾佳家居用品有限公司 | Household pattern recognition method and device based on counterstudy |
| CN111144243B (en) * | 2019-12-13 | 2022-07-08 | 江苏艾佳家居用品有限公司 | Household pattern recognition method and device based on counterstudy |
| CN111259734B (en) * | 2020-01-08 | 2023-06-16 | 深圳市彬讯科技有限公司 | Space household identification method, device, computer equipment and storage medium |
| CN111259734A (en) * | 2020-01-08 | 2020-06-09 | 深圳市彬讯科技有限公司 | Spatial house type identification method and device, computer equipment and storage medium |
| CN111340954B (en) * | 2020-02-18 | 2023-11-14 | 广东三维家信息科技有限公司 | House type wall drawing method and model training method and device thereof |
| CN111340954A (en) * | 2020-02-18 | 2020-06-26 | 广东三维家信息科技有限公司 | House type wall drawing method and model training method and device thereof |
| CN111339336B (en) * | 2020-03-11 | 2021-04-30 | 贝壳找房(北京)科技有限公司 | Analysis method and system for house type data |
| CN111339336A (en) * | 2020-03-11 | 2020-06-26 | 贝壳技术有限公司 | Analysis method and system for house type data |
| WO2021217340A1 (en) * | 2020-04-27 | 2021-11-04 | Li Jianjun | Ai-based automatic design method and apparatus for universal smart home scheme |
| CN111582358A (en) * | 2020-04-30 | 2020-08-25 | 上海添玑网络服务有限公司 | Training method and device for house type recognition model and house type weight judging method and device |
| CN111582358B (en) * | 2020-04-30 | 2024-05-10 | 上海添玑网络服务有限公司 | Training method and device for house type recognition model, and house type weight judging method and device |
| CN112116613A (en) * | 2020-09-25 | 2020-12-22 | 贝壳技术有限公司 | Model training method, image segmentation method, image vectorization method and system thereof |
| CN113591929A (en) * | 2021-07-05 | 2021-11-02 | 华南师范大学 | Family pattern recognition method combining direction sensing kernel cluster |
| CN113591929B (en) * | 2021-07-05 | 2023-07-25 | 华南师范大学 | House pattern recognition method combining direction perception kernel cluster |
| CN114401520A (en) * | 2021-12-28 | 2022-04-26 | 广西壮族自治区公众信息产业有限公司 | Method and system for detecting thermodynamic diagram of wireless network signal |
| CN114401520B (en) * | 2021-12-28 | 2024-02-09 | 广西壮族自治区公众信息产业有限公司 | Method and system for detecting thermodynamic diagram of wireless network signal |
| CN114973297A (en) * | 2022-06-17 | 2022-08-30 | 广州市圆方计算机软件工程有限公司 | Wall area identification method, system, equipment and medium for planar house type graph |
| CN116578915A (en) * | 2023-04-11 | 2023-08-11 | 广州极点三维信息科技有限公司 | Structured house type analysis method and system based on graphic neural network |
| CN116578915B (en) * | 2023-04-11 | 2024-06-11 | 广州极点三维信息科技有限公司 | Structured house type analysis method and system based on graphic neural network |
| CN116502546B (en) * | 2023-06-29 | 2023-10-10 | 深圳市明源云科技有限公司 | House type design method, house type design device, electronic equipment and readable storage medium |
| CN116502546A (en) * | 2023-06-29 | 2023-07-28 | 深圳市明源云科技有限公司 | House type design method, house type design device, electronic equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059750A (en) | House type shape recognition process, device and equipment | |
| CN108229509B (en) | Method and device for identifying object class and electronic equipment | |
| CN107679466B (en) | Information output method and device | |
| CN108229341A (en) | Sorting technique and device, electronic equipment, computer storage media, program | |
| CN106599925A (en) | Plant leaf identification system and method based on deep learning | |
| CN108960404B (en) | Image-based crowd counting method and device | |
| CN105320945A (en) | Image classification method and apparatus | |
| CN104915972A (en) | Image processing apparatus, image processing method and program | |
| WO2020143316A1 (en) | Certificate image extraction method and terminal device | |
| CN109657582A (en) | Recognition methods, device, computer equipment and the storage medium of face mood | |
| CN110781976B (en) | Extension method of training image, training method and related device | |
| WO2021232670A1 (en) | Pcb component identification method and device | |
| WO2023116565A1 (en) | Method for intelligently designing network security architecture diagram | |
| CN110457677A (en) | Entity-relationship recognition method and device, storage medium, computer equipment | |
| CN108319888A (en) | The recognition methods of video type and device, terminal | |
| CN110705225A (en) | Contract marking method and device | |
| CN111340022A (en) | Identity card information identification method and device, computer equipment and storage medium | |
| CN112365007B (en) | Model parameter determining method, device, equipment and storage medium | |
| CN104462526A (en) | Multi-user online collaboration rapid vectorization method for high-resolution remote sensing images | |
| CN110399760A (en) | A kind of batch two dimensional code localization method, device, electronic equipment and storage medium | |
| CN110414586A (en) | Antifalsification label based on deep learning tests fake method, device, equipment and medium | |
| CN109460767A (en) | Rule-based convex print bank card number segmentation and recognition methods | |
| CN105678790A (en) | High-resolution remote sensing image supervised segmentation method based on variable Gaussian hybrid model | |
| CN112487853A (en) | Handwriting comparison method and system, electronic equipment and storage medium | |
| CN106663212A (en) | Character recognition device, character recognition method, and program |
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: 20190726 |