CN110473288A - Dam model method for reconstructing, device and electronic equipment - Google Patents
Dam model method for reconstructing, device and electronic equipment Download PDFInfo
- Publication number
- CN110473288A CN110473288A CN201910782814.8A CN201910782814A CN110473288A CN 110473288 A CN110473288 A CN 110473288A CN 201910782814 A CN201910782814 A CN 201910782814A CN 110473288 A CN110473288 A CN 110473288A
- Authority
- CN
- China
- Prior art keywords
- semantic segmentation
- dykes
- dams
- binary map
- defect area
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/155—Segmentation; Edge detection involving morphological operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the present invention proposes a kind of dam model method for reconstructing, device and electronic equipment, is related to hydrotechnics field.Wherein, dam model method for reconstructing includes: and obtains multiple to carry acquisition position information and include the original data of dykes and dams；Every original data is handled using the semantic segmentation learning model of the defect area of dykes and dams for identification of training in advance, obtains the semantic segmentation binary map that can identify the dykes and dams defect area；Respectively every original data is merged to obtain image to be reconstructed with the corresponding semantic segmentation binary map；Three-dimensional reconstruction is carried out based on image to be reconstructed described in multiple and the corresponding acquisition position information, obtains the dam model that the dykes and dams defect area is marked.It realizes and accurately positions defect, be conducive to security risk existing for discovery dykes and dams in time, also save human cost.
Description
Technical field
The present invention relates to hydrotechnics fields, set in particular to a kind of dam model method for reconstructing, device and electronics
It is standby.
Background technique
Dykes and dams belong to key water control project infrastructure, people's livelihood quality assurance, expanding economy are all had very big
It influences.It is built as hydraulic engineering rapidly increases, carries out intelligent patrol detection for key water control project infrastructure, it has also become main trend.
Current intelligent patrol detection mainly passes through acquisition dykes and dams image, then carries out threedimensional model weight based on collected dykes and dams image
It builds, finally by the threedimensional model of manual inspection dykes and dams, to find the position of dykes and dams existing defects.Wherein, human cost it is high and
It is easy holiday defect, security risk existing for dykes and dams is made to be difficult to find in time.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of dam model method for reconstructing, device and electronic equipments.
To achieve the goals above, technical solution used in the embodiment of the present invention is as follows:
In a first aspect, the embodiment of the present invention provides a kind of dam model method for reconstructing, the dam model method for reconstructing packet
It includes: obtaining multiple and carry acquisition position information and include the original data of dykes and dams；Utilize the dike for identification of training in advance
The semantic segmentation learning model of dam defect area handles every original data, obtains that the dykes and dams defect can be identified
The semantic segmentation binary map in region；Every original data is carried out with the corresponding semantic segmentation binary map respectively
Fusion obtains image to be reconstructed；Three-dimensional Gravity is carried out based on image to be reconstructed described in multiple and the corresponding acquisition position information
It builds, obtains the dam model that the dykes and dams defect area is marked.
In alternative embodiments, the dam model method for reconstructing further include: to the obtained semantic segmentation two
Value figure carries out Morphological scale-space；Based on treated the semantic segmentation binary map, quantifying for the dykes and dams defect area is obtained
Change information；Described the step of carrying out three-dimensional reconstruction based on multiple described images to be reconstructed and corresponding acquisition position information includes:
Three-dimensional Gravity is carried out based on image to be reconstructed described in multiple, the corresponding acquisition position information and the corresponding quantification information
It builds.
In alternative embodiments, described respectively by every original data and the corresponding semantic segmentation
It includes: respectively by every original data and corresponding treated institute that binary map, which is merged to obtain image to be reconstructed,
Semantic segmentation binary map is stated to be merged to obtain the image to be reconstructed.
In alternative embodiments, described based on treated the semantic segmentation binary map, it obtains the dykes and dams and lacks
The step of falling into the quantification information in region includes: to extract institute from treated the semantic segmentation binary map using Framework Arithmetic
State the framework information of dykes and dams defect area；According to the framework information, the semantic segmentation binary map after combination processing is calculated
The image quantitative information of the dykes and dams defect area；Institute is obtained using preset transformational relation according to described image quantitative information
State quantification information；Wherein, the transformational relation is world coordinate system and the acquisition equipment for collecting the original data
Camera coordinates system between mapping relations.
In alternative embodiments, the Morphological scale-space include: out operation processing and closed operation processing one of or it
Between combination.
In alternative embodiments, the semantic segmentation using the defect area of dykes and dams for identification of training in advance
The step of practising every original data of model treatment includes: the pixel array for reading the original data；According to
According to the input size of the semantic segmentation learning model, the pixel array is divided into multiple subarrays to be processed；It utilizes
The semantic segmentation learning model successively handles each subarray to be processed, to obtain multiple bianry image arrays；Foundation
The multiple bianry image array obtains the semantic segmentation binary map.
Second aspect, the embodiment of the present invention provide a kind of dam model reconstructing device, the dam model reconstructing device packet
Include: obtain module, for obtain multiple carry acquisition position information and include dykes and dams original data；Processing module is used
Every initial pictures number is handled in the semantic segmentation learning model using the defect area of dykes and dams for identification of training in advance
According to obtaining the semantic segmentation binary map that can identify the dykes and dams defect area；Fusion Module, for described first by every respectively
Beginning image data is merged to obtain image to be reconstructed with the corresponding semantic segmentation binary map；Module is rebuild, for being based on
Multiple described images to be reconstructed and the corresponding acquisition position information carry out three-dimensional reconstruction, obtain that the dykes and dams defect is marked
The dam model in region.
In alternative embodiments, the processing module includes: reading submodule, for reading the initial pictures number
According to pixel array；Submodule is divided, for the input size according to the semantic segmentation learning model, by the pixel
Array is divided into multiple subarrays to be processed；Submodule is handled, for successively handling often using the semantic segmentation learning model
A subarray to be processed, to obtain multiple bianry image arrays；Submodule is obtained, for according to the multiple bianry image
Array obtains the semantic segmentation binary map.
The third aspect, the embodiment of the present invention provide a kind of electronic equipment, including processor and memory, the memory are deposited
Contain the machine-executable instruction that can be executed by the processor, the processor can be performed the machine-executable instruction with
Realize any method of aforementioned embodiments.
Fourth aspect, the embodiment of the present invention provide a kind of computer readable storage medium, are stored thereon with computer program,
The method as described in any one of aforementioned embodiments is realized when the computer program is executed by processor.
Dam model method for reconstructing provided in an embodiment of the present invention, first with the semanteme point that can identify dykes and dams defect area
It cuts learning model and handles every original data, obtain the semantic segmentation binary map that can identify dykes and dams defect area.That is, logical
The identification to dykes and dams pixel scale is crossed, goes out to belong to the pixel of dykes and dams defect respectively.Again respectively by every original data with
Corresponding semantic segmentation binary map is merged to obtain the image to be reconstructed for indicating dykes and dams defect.Finally according to image to be reconstructed
And the threedimensional model of corresponding acquisition position information reconstruction dykes and dams, the threedimensional model enable are straight while showing dykes and dams
Sight accurately shows defect present on dykes and dams and position.It realizes and accurately positions defect, be conducive to discovery dykes and dams in time and exist
Security risk, also save human cost.
To enable the above objects, features and advantages of the present invention to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only certain embodiments of the present invention, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 shows application scenario diagram provided in an embodiment of the present invention.
Fig. 2 shows the schematic diagrames of electronic equipment provided in an embodiment of the present invention.
Fig. 3 shows the step flow chart of dam model method for reconstructing provided in an embodiment of the present invention.
Fig. 4 is the sub-step flow chart of step S102 in Fig. 3.
Fig. 5 shows another part of the step flow chart of dam model method for reconstructing provided in an embodiment of the present invention.
Fig. 6 shows the schematic diagram of dam model reconstructing device provided in an embodiment of the present invention.
Icon: 100- electronic equipment；200- acquires equipment；110- memory；120- processor；130- communication module；
300- dam model reconstructing device；301- obtains module；302- processing module；303- Fusion Module；304- rebuilds module；305-
Morphological scale-space module；306- computing module.
Specific embodiment
Below in conjunction with attached drawing in the embodiment of the present invention, technical solution in the embodiment of the present invention carries out clear, complete
Ground description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Usually exist
The component of the embodiment of the present invention described and illustrated in attached drawing can be arranged and be designed with a variety of different configurations herein.
Therefore, the detailed description of the embodiment of the present invention provided in the accompanying drawings is not intended to limit below claimed
The scope of the present invention, but be merely representative of selected embodiment of the invention.Based on the embodiment of the present invention, those skilled in the art
Member's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
It should be noted that the relational terms of term " first " and " second " or the like be used merely to an entity or
Operation is distinguished with another entity or operation, and without necessarily requiring or implying between these entities or operation, there are any
This actual relationship or sequence.Moreover, the terms "include", "comprise" or its any other variant be intended to it is non-exclusive
Property include so that include a series of elements process, method, article or equipment not only include those elements, but also
Further include other elements that are not explicitly listed, or further include for this process, method, article or equipment it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including described
There is also other identical elements in the process, method, article or equipment of element.
Referring to FIG. 1, Fig. 1 shows the application scenarios of the embodiment of the present invention.As shown in Figure 1, acquisition equipment 200 and electronics
It is communicated to connect between equipment 100.
Above-mentioned acquisition equipment 200 is that when carrying out Image Acquisition, can synchronize the equipment for positioning own location information.It is optional
Ground, above-mentioned acquisition equipment 200 can be unmanned plane, and unmanned plane can determine the location information of itself.When unmanned plane carries out image
It, can synchronous acquisition location information when acquisition.Above-mentioned acquisition position information can be GPS information.It can be with according to acquisition position information
Determine position of the unmanned plane when carrying out Image Acquisition in world coordinate system.Certainly, above-mentioned to adopt in some other embodiments
Collection equipment 200 can also be other acquisition equipment 200 that can be positioned, for example, camera, camera, mobile phone etc..
It referring to figure 2., is the block diagram of electronic equipment 100.The electronic equipment 100 includes memory 110, processing
Device 120 and communication module 130.The memory 110, processor 120 and each element of communication module 130 between each other directly or
It is electrically connected indirectly, to realize the transmission or interaction of data.For example, these elements between each other can be logical by one or more
It interrogates bus or signal wire is realized and is electrically connected.
Wherein, memory 110 is for storing program or data.The memory 110 may be, but not limited to, at random
It accesses memory (Random Access Memory, RAM), read-only memory (Read Only Memory, ROM), may be programmed
Read-only memory (Programmable Read-Only Memory, PROM), erasable read-only memory (Erasable
Programmable Read-Only Memory, EPROM), electricallyerasable ROM (EEROM) (Electric Erasable
Programmable Read-Only Memory, EEPROM) etc..
Data or program of the processor 120 for being stored in read/writable memory device 110, and execute correspondingly function.
Communication module 130 is used to establish by the network logical between the electronic equipment 100 and other communication terminals
Letter connection, and for passing through the network sending and receiving data.
It should be understood that structure shown in Fig. 2 is only the structural schematic diagram of electronic equipment 100, the electronic equipment 100
It may also include than shown in Fig. 2 more perhaps less component or with the configuration different from shown in Fig. 2.It is shown in Fig. 2
Each component can be realized using hardware, software, or its combination.
First embodiment
Referring to FIG. 3, Fig. 3 shows a kind of dam model method for reconstructing provided in an embodiment of the present invention.Above-mentioned dykes and dams mould
Type method for reconstructing is applied to electronic equipment 100.As shown in figure 3, above-mentioned dam model method for reconstructing may comprise steps of:
Step S101 obtains multiple and carries acquisition position information and include the original data of dykes and dams.
In embodiments of the present invention, receive unmanned plane from multiple station acquisitions to the multiple angles of dykes and dams initial pictures number
According to.It should be noted that the GPS of unmanned plane at this time is believed when unmanned plane often collects an image data comprising dykes and dams
Breath is written as acquisition position information in the image attributes of the collected image data, to obtain original data.
Step S102 handles every using the semantic segmentation learning model of the defect area of dykes and dams for identification of training in advance
The original data obtains the semantic segmentation binary map that can identify the dykes and dams defect area.
The semantic segmentation learning model of the above-mentioned defect area of dykes and dams for identification can first pass through training in advance and obtain.It needs
It is bright, semantic segmentation learning model can based on full convolutional neural networks (Fully Convolutional Network,
FCN) training obtains.FCN is (the Convolutional Neural of convolutional neural networks end to end based on semantic segmentation
Network, CNN).Specifically, FCN is obtained and the full link sort layer of CNN is replaced with warp lamination, can be realized picture
The detection of plain rank, it can realize the identification to each pixel.It is, FCN also can be regarded as an extension
CNN, prediction object can be converted to a semantic segmentation image, i.e. dense prediction by it.FCN is mainly by down-sampling and up-sampling
Two parts composition, wherein down-sampling part includes convolutional layer, pond layer and dropout layers, and up-sampling layer is mainly by warp lamination
It constitutes.Down-sampling part is mainly a very deep convolutional network, is replaced using the preferable deep neural network of detection effect
The training time can significantly be shortened, such as VGG16, VGG19, GoogLeNet, ResNet.VGG19 network is commonly used in extensive
The target recognition and classification of image, while extensive application industrially demonstrates the validity of VGG19 network.Therefore, this hair
It is bright to use VGG19 network as down-sampling part, realize the feature extraction to defective data.Upper sampling process and down-sampling process
On the contrary, having stronger predictive ability, a class prediction can be generated to each pixel.Upper sampling process is not only able to achieve pair
The accurate positioning of target, moreover it is possible to guarantee that the Output Size of whole network is equal to input size, realize Target Segmentation end to end.
Further, selected FCN is trained to the step that can identify the semantic segmentation learning model of dykes and dams defect area
It suddenly include: to obtain the image with dykes and dams.It is the consistent sample image of input size of multiple sizes and FCN by image cropping.
The dykes and dams defect area occurred in sample image (for example, dykes and dams crack area) is labeled.Utilize the sample graph after mark
As being trained to FCN, to obtain the semantic segmentation learning model of dykes and dams defect area for identification.
Semantic segmentation binary map is obtained after handling original data by above-mentioned semantic segmentation learning model.Above-mentioned semanteme
Divide binary map by the image-region of the dykes and dams defect presented in original data each pixel assign the first value and
Assign the pixel of other image-regions to second value.
In embodiments of the present invention, as shown in figure 4, the above-mentioned language using the defect area of dykes and dams for identification of training in advance
Justice segmentation learning model the step of handling every original data includes:
Sub-step S1021 reads the pixel array of original data.
The size of above-mentioned pixel array is identical as the size of original data, each element exists in pixel array
A pixel is corresponded in original data.The value of each element is the pixel value of corresponding pixel.For example, reading
The original data of 5472*3648 obtains the pixel array of a 5472*3648.It is located at (1,1) in above-mentioned pixel array
Element be original data in (1,1) pixel pixel value.
Pixel array is divided into multiple wait locate by sub-step S1022 according to the input size of semantic segmentation learning model
Manage subarray.
It is to be appreciated that there is limitation in the picture size that semantic segmentation learning model can be handled each time.In order to meet language
The input size requirement of justice segmentation learning model, in the related technology by the way of cutting input picture.However, to input picture
Cutting, both time-consuming, also occupying system resources.
In embodiments of the present invention, meet semantic segmentation learning model input size by the way that pixel array to be divided into
Multiple subarrays to be processed, then the subarray to be processed obtained after division is sequentially input into semantic segmentation learning model and is known
Not, whole process cuts initial pictures without practical.Pixel array is divided using above-mentioned, both meets semantic segmentation
Requirement of the model to input data size is practised, the process for cutting original data is in turn avoided, plays time-saving effect.
Example is connected, semantic segmentation learning model is inputted having a size of 912*608, then drawn the pixel array of 5472*3648
It is divided into the subarray to be processed of 36 912*608.
Sub-step S1023 successively handles each subarray to be processed using semantic segmentation learning model, to obtain multiple two
It is worth pattern matrix.
In embodiments of the present invention, the corresponding pixel of each subarray to be processed is clicked through using semantic segmentation learning model
Row identification tells the pixel for belonging to dykes and dams defect area, and obtains corresponding bianry image array based on distinguishing results.
Sub-step S1024 obtains semantic segmentation binary map according to multiple bianry image arrays.
In embodiments of the present invention, obtained multiple bianry image arrays are integrated, obtains semantic segmentation binary map.
Further, the semantic segmentation binary map obtained using semantic segmentation learning model is certainly existed defect information and not connected
The problems such as continuous, containing noise.In order to improve the above problem, the integrality of the dykes and dams defect area improved and detection can
By property.In some embodiments, on the basis of dam model method for reconstructing shown in Fig. 3, as shown in figure 5, above-mentioned dykes and dams mould
Type method for reconstructing can with comprising steps of
Step S201 carries out Morphological scale-space to obtained semantic segmentation binary map.
In embodiments of the present invention, above-mentioned Morphological scale-space can be out operation processing and closed operation processing one of or between
Combination.Above-mentioned operation processing of opening is first to carry out corrosion treatment to semantic segmentation binary map, then carry out expansion process.By opening behaviour
Noise can be removed by dealing with.Above-mentioned closed operation processing is first to carry out expansion process to semantic segmentation binary map, then corroded
Processing.The dykes and dams defect area identified can be made continuous by closed operation processing.In some embodiments, if it is decided that semantic
The noise for dividing binary map is more, then uses and open operation processing.In some embodiments, if it is decided that semantic segmentation binary map
Without noise, then handled using closed operation.Certainly, in some embodiments, it is also possible to which operation processing and closed operation processing will be opened
It enables.
Above-mentioned corrosion treatment may is that definition structure member, and the origin of the structural elements of definition is successively moved to semantic segmentation
On each pixel of binary map.Corrosion results after moving each time are by semantic segmentation binary map and structural elements overlapping region
Minimum value on image determines.Small and meaningless object can be eliminated by etching operation, is suitable for noise reduction process, expression
Formula is as follows:
Wherein, f represents semantic segmentation binary map, and b representative structure member, (x, y) represents pixel in semantic segmentation binary map
Coordinate position.
Above-mentioned expansion process may is that definition structure member, and the origin of the structural elements of definition is successively moved to semantic segmentation
On each pixel of binary map.Expansion results after moving each time are by semantic segmentation binary map and structural elements overlapping region
Maximum value on image determines.Certain cavities and elimination in target area can be used to fill up by expansive working and be included in mesh
The grain noise that disappears in region is marked, expression formula is as follows:
Wherein, f represents semantic segmentation binary map, and b representative structure member, (x, y) represents pixel in semantic segmentation binary map
Coordinate position.
Step S202 obtains the quantification of the dykes and dams defect area based on treated the semantic segmentation binary map
Information.
In embodiments of the present invention, quantification information can be the dimension information of dykes and dams defect area.For example, quantification is believed
Breath includes the morphological features such as length, area, width.
It is alternatively possible to first extract dykes and dams defect area from treated semantic segmentation binary map using Framework Arithmetic
Framework information.The framework information of the dykes and dams defect area extracted may include skeleton curve, and above-mentioned skeleton curve is with single pixel
Point is width, and but the center line with dykes and dams defect area essentially coincides.Above-mentioned skeleton curve can sketch the contours of dykes and dams defect area
Extending direction.
Then, according to framework information, semantic segmentation binary map after combination processing, the image for calculating dykes and dams defect area is fixed
Measure information.Above-mentioned image quantitative information can be the morphological features such as length, area, the width of dykes and dams defect area in the picture.
Optionally, according to framework information, formula is utilized:
Calculate the length information of dykes and dams defect area in the picture.Wherein, f (x, y) indicates geometric calibration index, can root
It is determined according to the distortion situation of image, usually in the case where image does not have geometric distortion, f (x, y) is 1.Dl indicates framework information
Length infinitesimal, C represents the corresponding skeleton curve of framework information.
It is to be appreciated that carrying out skeletal extraction, obtained framework information to dykes and dams defect area using medial axis transformation algorithm
It further include dykes and dams defect area axis at a distance from background other than skeleton curve.Public affairs are utilized according to framework information based on this
Formula:
Obtain the area of dykes and dams defect area in the picture.Wherein, f (x, y) indicates that geometric calibration index, dS indicate skeleton
The area element of information, S, which is represented, utilizes medial axis transformation algorithm treated the corresponding curved surface of dykes and dams defect area.
According to framework information, formula is utilized:
Obtain the width of dykes and dams defect area in the picture.
Finally, using transformational relation, obtaining quantification information according to described image quantitative information.Above-mentioned transformational relation is
World coordinate system and collect the original data acquisition equipment 200 camera coordinates system between mapping relations.It can be with
Understand ground, the object of one known dimensions is shot using same acquisition equipment 200, identical object distance and focal length, can be obtained
The corresponding relationship of (i.e. actual size) and (i.e. Pixel Dimensions) under camera coordinates system under its world coordinate system.Therefore it can lead in advance
It crosses test to obtain above-mentioned transformational relation and store, to use.
Every original data is merged to obtain to weight by step S103 with corresponding semantic segmentation binary map respectively
Build image.
It is to be appreciated that each original data corresponding one is semantic by the processing of semantic segmentation learning model
Divide binary map.Above-mentioned semantic segmentation binary map will belong to dykes and dams defect area using the first value and second value in original data
The pixel in domain and the pixel for being not belonging to dykes and dams defect area distinguish.
In the embodiment of the present invention, after every original data and corresponding semantic segmentation binary map are merged, obtain
To image to be reconstructed.For above-mentioned image to be reconstructed is compared to original data, marks in original data and existed
Dykes and dams defect area.Meanwhile also losing the acquisition position information in image attributes.Therefore in some embodiments, by every
Before original data is merged with semantic segmentation binary map, the acquisition position information extraction of the original data is gone out
Come, designated storage area is written.After original data is merged with corresponding semantic segmentation binary map, from specified storage
The acquisition position information of original data is taken out in region, and is written in the image attributes of image to be reconstructed.
In some embodiments, for example, in dam model method for reconstructing corresponding to Fig. 5, it is above-mentioned respectively will be at the beginning of every
Beginning image data and corresponding semantic segmentation binary map, which are merged to obtain image to be reconstructed, may is that every initial graph respectively
As data with corresponding treated that semantic segmentation binary map is merged to obtain image to be reconstructed.After obtaining image to be reconstructed,
The quantification information of the dykes and dams defect area of image to be reconstructed can be also written in the image attributes of image to be reconstructed.
Step S104 carries out three-dimensional reconstruction based on multiple images to be reconstructed and corresponding acquisition position information, is marked
The dam models of dykes and dams defect areas.
In embodiments of the present invention, the corresponding acquisition position information of each image to be reconstructed.According to image to be reconstructed
Acquisition position information, it may be determined that in multiple images to be reconstructed characterize dykes and dams on same position region, and be based on this, carry out three
Dimension modeling, obtains the model of dykes and dams.Dykes and dams defect area, relevant people can be accurately identified on the dam model of three-dimensional reconstruction
Member's monitoring.
It in some embodiments, can also be by dykes and dams defect area each on dam model after dam model reconstruction
Quantification information automatic marking and corresponding position, to facilitate related personnel to learn.To realize to dykes and dams defect area
Precise positioning and determination.
In order to execute the corresponding steps in above-described embodiment and each possible mode, a kind of dam model weight is given below
The implementation of device 300 is built, optionally, which can use above-mentioned electronic equipment shown in Fig. 2
100 device architecture.Further, referring to Fig. 6, Fig. 6 is a kind of dam model reconstructing device provided in an embodiment of the present invention
300 functional block diagram.It should be noted that dam model reconstructing device 300 provided by the present embodiment, basic principle and
The technical effect of generation is identical with above-described embodiment, and to briefly describe, the present embodiment part does not refer to place, can refer to above-mentioned
Corresponding contents in embodiment.The dam model reconstructing device 300 includes: to obtain module 301, processing module 302, Fusion Module
303 and rebuild module 304.
Above-mentioned acquisition module 301, for obtain multiple carry acquisition position information and include dykes and dams original data.
In embodiments of the present invention, above-mentioned steps S101 can be executed by acquisition module 301.
Above-mentioned processing module 302, for being learnt using the semantic segmentation of the defect area of dykes and dams for identification of training in advance
The model treatment every original data obtains the semantic segmentation binary map that can identify the dykes and dams defect area.
In embodiments of the present invention, above-mentioned steps S102 can be executed by processing module 302.Preferably, above-mentioned processing mould
Block 302 includes: reading submodule, divides submodule, handles submodule and obtain submodule.
Reading submodule, for reading the pixel array of the original data.
In embodiments of the present invention, above-mentioned sub-step S1021 can be executed by reading submodule.
Submodule is divided, for the input size according to the semantic segmentation learning model, the pixel array is drawn
It is divided into multiple subarrays to be processed.
In embodiments of the present invention, above-mentioned sub-step S1022 can be executed by division submodule.
Submodule is handled, for successively handling each subarray to be processed using the semantic segmentation learning model,
To obtain multiple bianry image arrays.
In embodiments of the present invention, above-mentioned sub-step S1023 can be executed by processing submodule.
Submodule is obtained, for obtaining the semantic segmentation binary map according to the multiple bianry image array.
In embodiments of the present invention, above-mentioned sub-step S1024 can be executed by obtaining submodule.
Above-mentioned Fusion Module 303, for respectively by every original data and the corresponding semantic segmentation two
Value figure is merged to obtain image to be reconstructed.
In embodiments of the present invention, above-mentioned steps S103 can be executed by Fusion Module 303.
Above-mentioned reconstruction module 304, for based on multiple described images to be reconstructed and the corresponding acquisition position information into
Row three-dimensional reconstruction obtains the dam model that the dykes and dams defect area is marked.
In embodiments of the present invention, above-mentioned steps S104 can be executed by reconstruction module 304.
Further, in some embodiments, above-mentioned dam model reconstructing device 300 further include: Morphological scale-space module
305 and computing module 306.
Above-mentioned Morphological scale-space module 305, for carrying out Morphological scale-space to the obtained semantic segmentation binary map.
In embodiments of the present invention, above-mentioned steps S201 can be executed by Morphological scale-space module 305.
Above-mentioned computing module 306, for obtaining the dykes and dams defect area based on treated the semantic segmentation binary map
The quantification information in domain.
In embodiments of the present invention, above-mentioned steps S202 can be executed by computing module 306.
Optionally, above-mentioned module can be stored in memory 110 shown in Fig. 2 in the form of software or firmware (Firmware)
In or solidify in the operating system (Operating System, OS) of the electronic equipment 100, and can be by the processor in Fig. 2
120 execute.Meanwhile the code etc. of data needed for executing above-mentioned module, program can store in the memory 110.
In conclusion the embodiment of the invention provides a kind of dam model method for reconstructing, device and electronic equipments.Wherein,
Above-mentioned dam model method for reconstructing includes: to obtain multiple to carry acquisition position information and include the original data of dykes and dams；Benefit
Every original data is handled with the semantic segmentation learning model of the defect area of dykes and dams for identification of training in advance, is obtained
To the semantic segmentation binary map that can identify the dykes and dams defect area；Obtained semantic segmentation binary map is carried out at morphology
Reason, improves the integrality and reliability of the dykes and dams defect area identified.Respectively by every original data with it is corresponding
Semantic segmentation binary map is merged to obtain image to be reconstructed.From extracting dykes and dams defect area in treated semantic segmentation binary map
The quantification information in domain.Finally, being carried out based on multiple image, quantification information and corresponding acquisition position information to be reconstructed three-dimensional
It rebuilds, obtains the dam model that the dykes and dams defect area is marked.It realizes and accurately positions defect, be conducive to discovery dike in time
Security risk existing for dam, also saves human cost.
In several embodiments provided herein, it should be understood that disclosed device and method can also pass through
Other modes are realized.The apparatus embodiments described above are merely exemplary, for example, flow chart and block diagram in attached drawing
Show the device of multiple embodiments according to the present invention, the architectural framework in the cards of method and computer program product,
Function and operation.In this regard, each box in flowchart or block diagram can represent the one of a module, section or code
Part, a part of the module, section or code, which includes that one or more is for implementing the specified logical function, to be held
Row instruction.It should also be noted that function marked in the box can also be to be different from some implementations as replacement
The sequence marked in attached drawing occurs.For example, two continuous boxes can actually be basically executed in parallel, they are sometimes
It can execute in the opposite order, this depends on the function involved.It is also noted that every in block diagram and or flow chart
The combination of box in a box and block diagram and or flow chart can use the dedicated base for executing defined function or movement
It realizes, or can realize using a combination of dedicated hardware and computer instructions in the system of hardware.
In addition, each functional module in each embodiment of the present invention can integrate one independent portion of formation together
Point, it is also possible to modules individualism, an independent part can also be integrated to form with two or more modules.
It, can be with if the function is realized and when sold or used as an independent product in the form of software function module
It is stored in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be a
People's computer, server or network equipment etc.) it performs all or part of the steps of the method described in the various embodiments of the present invention.
And storage medium above-mentioned includes: that USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited
The various media that can store program code such as reservoir (RAM, Random Access Memory), magnetic or disk.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made any to repair
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of dam model method for reconstructing, which is characterized in that the dam model method for reconstructing includes:
Multiple are obtained to carry acquisition position information and include the original data of dykes and dams；
Every initial pictures are handled using the semantic segmentation learning model of the defect area of dykes and dams for identification of training in advance
Data obtain the semantic segmentation binary map that can identify the dykes and dams defect area；
Respectively every original data is merged to obtain figure to be reconstructed with the corresponding semantic segmentation binary map
Picture；
Three-dimensional reconstruction is carried out based on image to be reconstructed described in multiple and the corresponding acquisition position information, obtains being marked described
The dam model of dykes and dams defect area.
2. dam model method for reconstructing according to claim 1, which is characterized in that the dam model method for reconstructing also wraps
It includes:
Morphological scale-space is carried out to the obtained semantic segmentation binary map；
Based on treated the semantic segmentation binary map, the quantification information of the dykes and dams defect area is obtained；
Described the step of carrying out three-dimensional reconstruction based on multiple described images to be reconstructed and corresponding acquisition position information includes: to be based on
Multiple described image, the corresponding acquisition position information and corresponding described quantification information to be reconstructed carry out three-dimensional reconstruction.
3. dam model method for reconstructing according to claim 2, which is characterized in that described respectively by every initial graph
As data merged to obtain image to be reconstructed with the corresponding semantic segmentation binary map include:
Respectively by every original data with corresponding treated that the semantic segmentation binary map is merged to obtain
The image to be reconstructed.
4. dam model method for reconstructing according to claim 2, which is characterized in that described based on treated the semanteme
Divide binary map, the step of obtaining the quantification information of the dykes and dams defect area includes:
The framework information of the dykes and dams defect area is extracted from treated the semantic segmentation binary map using Framework Arithmetic；
According to the framework information, the semantic segmentation binary map after combination processing calculates the figure of the dykes and dams defect area
As quantitative information；
The quantification information is obtained using preset transformational relation according to described image quantitative information；Wherein, the conversion
Mapping relations of the relationship between world coordinate system and the camera coordinates system for the acquisition equipment for collecting the original data.
5. dam model method for reconstructing according to claim 2, which is characterized in that the Morphological scale-space includes: out behaviour
It deals with and one or both of closed operation processing combination.
6. dam model method for reconstructing according to claim 1, which is characterized in that described to utilize that trains in advance to be used to know
The step of semantic segmentation learning model of other dykes and dams defect area handles every original data include:
Read the pixel array of the original data；
According to the input size of the semantic segmentation learning model, the pixel array is divided into multiple submatrixs to be processed
Column；
Each subarray to be processed is successively handled using the semantic segmentation learning model, to obtain multiple bianry image battle arrays
Column；
According to the multiple bianry image array, the semantic segmentation binary map is obtained.
7. a kind of dam model reconstructing device, which is characterized in that the dam model reconstructing device includes:
Obtain module, for obtain multiple carry acquisition position information and include dykes and dams original data；
Processing module, for handling every using the semantic segmentation learning model of the defect area of dykes and dams for identification of training in advance
The original data obtains the semantic segmentation binary map that can identify the dykes and dams defect area；
Fusion Module, for respectively merging every original data with the corresponding semantic segmentation binary map
Obtain image to be reconstructed；
Module is rebuild, for carrying out three-dimensional reconstruction based on multiple described images to be reconstructed and the corresponding acquisition position information,
Obtain the dam model that the dykes and dams defect area is marked.
8. dam model reconstructing device according to claim 7, which is characterized in that the processing module includes:
Reading submodule, for reading the pixel array of the original data；
Submodule is divided to be divided into the pixel array for the input size according to the semantic segmentation learning model
Multiple subarrays to be processed；
Submodule is handled, for successively handling each subarray to be processed using the semantic segmentation learning model, with
To multiple bianry image arrays；
Submodule is obtained, for obtaining the semantic segmentation binary map according to the multiple bianry image array.
9. a kind of electronic equipment, which is characterized in that including processor and memory, the memory is stored with can be by the place
The machine-executable instruction that device executes is managed, the machine-executable instruction can be performed to realize claim 1-6 in the processor
Any method.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
Such as method of any of claims 1-6 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910782814.8A CN110473288B (en) | 2019-08-23 | 2019-08-23 | Dam model reconstruction method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910782814.8A CN110473288B (en) | 2019-08-23 | 2019-08-23 | Dam model reconstruction method and device and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110473288A true CN110473288A (en) | 2019-11-19 |
| CN110473288B CN110473288B (en) | 2022-08-05 |
Family
ID=68512104
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910782814.8A Active CN110473288B (en) | 2019-08-23 | 2019-08-23 | Dam model reconstruction method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110473288B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112288741A (en) * | 2020-11-23 | 2021-01-29 | 四川长虹电器股份有限公司 | Product surface defect detection method and system based on semantic segmentation |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103413288A (en) * | 2013-08-27 | 2013-11-27 | 南京大学 | LCD general defect detecting method |
| JP2016015638A (en) * | 2014-07-02 | 2016-01-28 | 株式会社デンソー | Fault detection device and fault detection program |
| CN106709568A (en) * | 2016-12-16 | 2017-05-24 | 北京工业大学 | RGB-D image object detection and semantic segmentation method based on deep convolution network |
| CN108876790A (en) * | 2017-09-14 | 2018-11-23 | 北京旷视科技有限公司 | Image, semantic dividing method and device, neural network training method and device |
| US20190064378A1 (en) * | 2017-08-25 | 2019-02-28 | Wei D. LIU | Automated Seismic Interpretation Using Fully Convolutional Neural Networks |
| CN109884067A (en) * | 2019-01-26 | 2019-06-14 | 北京中水科海利工程技术有限公司 | Dam face slab defect precise positioning and visual processing method |
-
2019
- 2019-08-23 CN CN201910782814.8A patent/CN110473288B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103413288A (en) * | 2013-08-27 | 2013-11-27 | 南京大学 | LCD general defect detecting method |
| JP2016015638A (en) * | 2014-07-02 | 2016-01-28 | 株式会社デンソー | Fault detection device and fault detection program |
| CN106709568A (en) * | 2016-12-16 | 2017-05-24 | 北京工业大学 | RGB-D image object detection and semantic segmentation method based on deep convolution network |
| US20190064378A1 (en) * | 2017-08-25 | 2019-02-28 | Wei D. LIU | Automated Seismic Interpretation Using Fully Convolutional Neural Networks |
| CN108876790A (en) * | 2017-09-14 | 2018-11-23 | 北京旷视科技有限公司 | Image, semantic dividing method and device, neural network training method and device |
| CN109884067A (en) * | 2019-01-26 | 2019-06-14 | 北京中水科海利工程技术有限公司 | Dam face slab defect precise positioning and visual processing method |
Non-Patent Citations (5)
| Title |
|---|
| GAO VU DUNG 等: "Autonomous concrete crack detection using deep fully convolutional neural network", 《AUTOMATION IN CONSTRUCTION》 * |
| YOUNG JIN CHA 等: "Deep Learning-Based Crack Damage Detection Using Convolutional Neural Networks", 《COMPUTER AIDED CIVIL AND INFRASTRUCTURE ENGINEERING》 * |
| ZHENG TONG 等: "Recognition, location, measurement, and 3D reconstruction of concealed cracks using convolutional neural networks", 《CONSTRUCTION AND BUILDING MATERIALS》 * |
| 王朝琴: "模块图像的三维重建与缺陷检测", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 王森 等: "基于深度学习的全卷积网络图像裂纹检测", 《计算机辅助设计与图形学学报》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112288741A (en) * | 2020-11-23 | 2021-01-29 | 四川长虹电器股份有限公司 | Product surface defect detection method and system based on semantic segmentation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110473288B (en) | 2022-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xu et al. | Detecting tiny objects in aerial images: A normalized Wasserstein distance and a new benchmark | |
| CN110009013A (en) | Encoder training and characterization information extracting method and device | |
| Fernandes et al. | Pavement pathologies classification using graph-based features | |
| CN111044570A (en) | Defect identification and early warning method and device for power equipment and computer equipment | |
| CN109859183B (en) | Edge calculation-based multi-element integrated water body intelligent identification method and ecological station | |
| CN111242144B (en) | Method and device for detecting abnormality of power grid equipment | |
| CN111091101B (en) | High-precision pedestrian detection method, system and device based on one-step method | |
| Rahman et al. | Semantic deep learning integrated with RGB feature-based rule optimization for facility surface corrosion detection and evaluation | |
| CN114092487A (en) | Target fruit instance segmentation method and system | |
| CN107688811A (en) | Licence plate recognition method and device | |
| CN109740609A (en) | A kind of gauge detection method and device | |
| CN112818849B (en) | Crowd density detection algorithm based on context attention convolutional neural network for countermeasure learning | |
| CN114332473A (en) | Object detection method, object detection device, computer equipment, storage medium and program product | |
| CN113610070A (en) | Landslide disaster identification method based on multi-source data fusion | |
| CN111723814A (en) | Cross-image association based weak supervision image semantic segmentation method, system and device | |
| CN113706481A (en) | Sperm quality detection method, sperm quality detection device, computer equipment and storage medium | |
| CN114037907A (en) | Detection method and device for power transmission line, computer equipment and storage medium | |
| CN111476307A (en) | Lithium battery surface defect detection method based on depth field adaptation | |
| CN114639102B (en) | Cell segmentation method and device based on key point and size regression | |
| CN110473288A (en) | Dam model method for reconstructing, device and electronic equipment | |
| Luo et al. | RBD-Net: robust breakage detection algorithm for industrial leather | |
| KR102416714B1 (en) | System and method for city-scale tree mapping using 3-channel images and multiple deep learning | |
| CN111079752A (en) | Method and device for identifying circuit breaker in infrared image and readable storage medium | |
| CN114463300A (en) | Steel surface defect detection method, electronic device, and storage medium | |
| CN115082416A (en) | Lens flaw detection method, device, equipment and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |