CN111382645A - Method and system for identifying expired buildings in electronic map - Google Patents
Method and system for identifying expired buildings in electronic map Download PDFInfo
- Publication number
- CN111382645A CN111382645A CN201811646297.3A CN201811646297A CN111382645A CN 111382645 A CN111382645 A CN 111382645A CN 201811646297 A CN201811646297 A CN 201811646297A CN 111382645 A CN111382645 A CN 111382645A
- Authority
- CN
- China
- Prior art keywords
- picture
- building
- buildings
- target area
- identifying
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 74
- 238000003709 image segmentation Methods 0.000 claims abstract description 51
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 38
- 238000012545 processing Methods 0.000 claims abstract description 29
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 238000013528 artificial neural network Methods 0.000 claims abstract description 9
- 238000005520 cutting process Methods 0.000 claims description 17
- 230000011218 segmentation Effects 0.000 claims description 10
- 238000012216 screening Methods 0.000 claims description 8
- 230000004931 aggregating effect Effects 0.000 claims description 5
- 238000004220 aggregation Methods 0.000 claims description 3
- 230000002776 aggregation Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 238000007689 inspection Methods 0.000 description 3
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A30/00—Adapting or protecting infrastructure or their operation
- Y02A30/60—Planning or developing urban green infrastructure
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for identifying outdated buildings in an electronic map, wherein the method for identifying the outdated buildings in the electronic map comprises the following steps: acquiring a training data set and a label data set; preprocessing the training data set and the label data set to obtain a training picture set and a label picture set; training a deep convolutional neural network image segmentation model according to the training picture set and the label picture set and obtaining model parameters; inputting the target area picture set into a trained deep convolution neural network image segmentation model, and performing building identification through trained model parameters to obtain a target area building probability image map; and processing the target area building probability image map to identify an overdue building. The method disclosed by the invention improves the updating speed of the electronic navigation map, can control and modify the operation style and the operation quality of the electronic map, reduces the number of overdue buildings processed manually and reduces the labor cost.
Description
Technical Field
The invention belongs to the technical field of image processing, particularly relates to an electronic map overdue building processing method based on deep convolutional neural network satellite image segmentation, and particularly relates to a method and a system for identifying overdue buildings in an electronic map.
Background
The high-precision electronic map is one of basic data of the unmanned system, and the data for the unmanned system must have the capability of being updated in time. The existing consumer-grade electronic map generally adopts a map operator to manually delete the expired data, so that the updating speed of the electronic map is between 3 months and 1 year, and the data requirement of an unmanned system on the electronic map cannot be met. In addition, the existing consumer-grade electronic map has low efficiency of quality inspection at the later stage of map operation due to the inconsistency of proficiency level, operation style and the like of operators.
On the other hand, the real-time updating of the satellite pictures is fast at present, but in the prior art, elements on the satellite pictures, including buildings, are manually traced, and the method of manual operation cannot adapt to the fast-updating satellite pictures.
Therefore, there is an urgent need to develop a new electronic map processing method, which can improve or solve the above technical problems caused by manual updating and quality inspection of the electronic map by operators and adapt to the rapidly updated satellite images.
Disclosure of Invention
In view of the above-mentioned drawbacks and deficiencies of the prior art, it is an object of the present invention to provide a method and system for identifying expired buildings in an electronic map. The method and the system for identifying the overdue buildings in the electronic map can solve the technical problems caused by manual updating and quality inspection of the electronic map at present, and adapt to the satellite picture which is updated quickly through automatic judgment of the model.
In order to achieve the above object, in one aspect of the present invention, a method for identifying an expired building in an electronic map is provided, which includes:
acquiring a training data set and a label data set;
preprocessing the training data set and the label data set to obtain a training picture set and a label picture set;
training a deep convolutional neural network image segmentation model according to the training picture set and the label picture set and obtaining model parameters;
inputting the target area picture set into a trained deep convolution neural network image segmentation model, carrying out building identification through trained model parameters, and taking an S-shaped growth curve function as an output layer to obtain a target area building probability image map;
and processing the building probability image map of the target area, screening probability values in the building probability image map of the target area through a preset threshold value, and identifying overdue buildings.
Further, the training data set is a satellite picture and/or the label data set is city building data, and the building data at least comprises a building outline picture and building outline vector boundary data.
Further, the preprocessing the training data set and the label data set to obtain a training picture set and a label picture set, including:
unifying the projection coordinates of the satellite picture and the projection coordinates of the building contour vector boundary;
converting the building outline picture into a grid according to the resolution of the satellite picture;
cutting the boundary of the satellite picture and the boundary of the building outline picture according to a uniform frame;
and respectively cutting the satellite picture and the building outline picture according to a unified geographical grid to obtain a training picture set and a label picture set.
Further, the deep convolutional neural network image segmentation model adopts a medical image segmentation model.
Further, screening the probability values in the probability image map of the buildings in the target area through a preset threshold value, and deleting overdue buildings after identifying the overdue buildings.
Further, building identification is carried out through trained model parameters, and an S-shaped growth curve function is used as an output layer to obtain a building probability image map of a target area, wherein the building probability image map comprises the following steps:
dividing the target area picture set to obtain a plurality of sub-area picture sets;
aggregating the plurality of sub-region picture sets;
respectively processing the sub-region image sets through a trained deep convolution neural network image segmentation model to obtain a plurality of sub-region building probability image maps;
and splicing the sub-area building probability image maps to obtain a target area building probability image map.
In another aspect of the present invention, a system for identifying an expired building in an electronic map is further provided, where the system includes:
a data acquisition unit configured to acquire a training dataset and a label dataset;
the picture acquisition unit is configured to preprocess the training data set and the label data set to obtain a training picture set and a label picture set;
the model training unit is configured to train a deep convolutional neural network image segmentation model according to the training picture set and the label picture set and obtain model parameters;
the building probability image map acquisition unit is configured for inputting the target area image set into a trained deep convolutional neural network image segmentation model, performing building identification through trained model parameters, and taking an S-shaped growth curve function as an output layer to acquire a building probability image map of the target area;
and the overdue building processing unit is configured for processing the target area building probability image, screening the probability values in the target area building probability image through a preset threshold value and identifying overdue buildings.
Further, the training data set is a satellite picture and/or the label data set is city building data, and the building data at least comprises a building outline picture and building outline vector boundary data.
Further, the picture acquiring unit includes:
the projection coordinate unifying subunit is configured to unify the projection coordinates of the satellite picture and the projection coordinates of the building contour vector boundary;
the picture rasterizing subunit is configured to convert the building outline picture into a grid according to the resolution of the satellite picture;
a picture boundary clipping subunit configured to clip the boundary of the satellite picture and the boundary of the building outline picture according to a unified frame;
and the picture grid cutting subunit is configured to respectively cut the satellite picture and the building outline picture according to a unified geographic grid to obtain a training picture set and a label picture set.
Further, the deep convolutional neural network image segmentation model adopts a medical image segmentation model.
Further, the overdue building processing unit is configured to screen the probability value in the target area building probability image map through a preset threshold, and delete the overdue building after identifying the overdue building.
Further, the building probability image acquisition unit includes:
the target area picture set segmentation subunit is configured to segment the target area picture set to obtain a plurality of sub-area picture sets;
a sub-region picture set aggregation subunit configured to aggregate the plurality of sub-region picture sets;
the sub-region building probability image map acquisition subunit is configured to process the sub-region image sets respectively through a trained deep convolutional neural network image segmentation model to acquire a plurality of sub-region building probability image maps;
and the target area building probability image acquisition subunit is configured for splicing the sub-area building probability image to acquire a target area building probability image.
In another aspect of the present invention, there is also provided an apparatus for identifying an expired building in an electronic map, the apparatus including:
one or more processors;
a data store for storing data and one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method of any of the above.
In another aspect of the invention, there is also provided a computer readable storage medium storing a computer program which, when executed by a processor, implements the method as defined in any one of the above.
Compared with the prior art, the invention has the beneficial effects that:
1. the method for identifying the overdue buildings in the electronic map improves the updating speed of the electronic navigation map, has higher speed and precision of identifying the buildings by the deep convolutional neural network model compared with the existing manual processing mode, can control and modify the operation style and the operation quality of the electronic map, reduces the number of the overdue buildings manually processed, and reduces the labor cost.
2. According to the system for identifying the overdue buildings in the electronic map, the updating speed of the electronic navigation map is improved by arranging the picture acquisition unit, the model training unit, the building probability image acquisition unit, the overdue building processing unit and the like, compared with the existing manual processing mode, the speed and the accuracy of identifying the buildings by the deep convolutional neural network model are higher, the operation style and the operation quality of the electronic map can be controlled and modified, the number of the overdue buildings processed manually is reduced, and the labor cost is reduced.
3. The device for identifying the overdue buildings in the electronic map provided by the embodiment is convenient for popularization of the technology for identifying the overdue buildings in the electronic map through the computer readable medium stored with the computer program.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments made with reference to the following drawings:
fig. 1 is a flowchart illustrating an exemplary method for identifying an expired building in an electronic map according to an embodiment of the present invention;
FIG. 2 is a natural image of an exemplary method for identifying expired buildings in an electronic map according to an embodiment of the present invention;
fig. 3 is a probability map obtained from a natural image picture through a trained deep convolutional neural network image segmentation model in an exemplary method for identifying an expired building in an electronic map according to an embodiment of the present invention;
fig. 4 is a probability chart of a method for identifying an expired building in an electronic map according to a second embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the relevant invention and not restrictive of the invention. It should be noted that, for convenience of description, only the portions related to the present invention are shown in the drawings.
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict. The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.
As shown in fig. 1, the present embodiment provides a method for identifying an expired building in an electronic map, which includes the following steps:
s1: acquiring a training data set and a label data set, wherein the training data set is a satellite picture, the label data set is urban building data, and the building data at least comprises a building outline picture and building outline vector boundary data.
S2: preprocessing a training data set and a label data set to obtain a training picture set and a label picture set, wherein the training picture set and the label picture set are used as training input of a deep convolutional neural network image segmentation model, and the method specifically comprises the following steps:
s2-1: unifying the projection coordinates of the satellite picture and the projection coordinates of the building contour vector boundary;
s2-2: converting the building outline picture into a grid according to the resolution of the satellite picture;
s2-3: cutting the boundary of the satellite picture and the boundary of the building outline picture according to a uniform frame;
s2-4: and respectively cutting the satellite picture and the building outline picture according to a unified geographical grid to obtain a training picture set and a label picture set.
S3: according to the training picture set and the label picture set, training a deep convolutional neural network image segmentation model and obtaining model parameters, wherein the deep convolutional neural network image segmentation model adopts a medical image segmentation model.
S4: inputting a target area picture set into a trained deep convolution neural network image segmentation model, carrying out building identification through trained model parameters, and taking an S-shaped growth curve function (Si gmo i d function) as an output layer to obtain a target area building probability image map;
s4-1: dividing the target area picture set to obtain a plurality of sub-area picture sets;
s4-2: aggregating the plurality of sub-region picture sets;
s4-3: respectively processing the sub-region image sets through a trained deep convolution neural network image segmentation model to obtain a plurality of sub-region building probability image maps;
s4-4: and splicing the sub-area building probability image maps to obtain a target area building probability image map.
And S5, processing the building probability image map of the target area, screening the probability value in the building probability image map of the target area through a preset threshold value, and deleting the overdue building.
In this embodiment, the acquired training data set is a satellite picture, and the tag data set is city building data, where the building data at least includes a building contour picture and building contour vector boundary data.
In this embodiment, preprocessing the training data set and the tag data set to obtain a training picture set and a tag picture set, including the following steps:
unifying the projection coordinates of the satellite picture and the projection coordinates of the building contour vector boundary;
converting the building outline picture into a grid according to the resolution of the satellite picture;
cutting the boundary of the satellite picture and the boundary of the building outline picture according to the unified frame;
and respectively cutting the satellite picture and the building outline picture according to the unified geographic grid to obtain a training picture set and a label picture set.
In the embodiment, a medical Image Segmentation model (volumetric Networks for biological Image Segmentation) is adopted as a deep Convolutional neural network Image Segmentation model, and the deep Convolutional neural network Image (CNN) Segmentation model has excellent automatic feature extraction capability, so that the accuracy of the CNN-based semantic Segmentation model is 20% -30% higher than that of the conventional Image Segmentation method. The medical Image Segmentation model (volumetric Networks for biological Image Segmentation) is a CNN Segmentation network architecture applied to cell Image Segmentation, and has the advantages that information can skip an intermediate full connection layer from a low feature layer to be directly connected to a high feature layer, so that uncertainty of results caused by sampling is avoided, and finally a high-quality pixel-level Segmentation effect is achieved.
In this embodiment, the preset threshold is 50%, a mark with a probability value lower than 50% in the target area building probability image map is identified as an expired building, and the expired building is deleted. In principle, the threshold value can be preset to 50%, i.e. within the range of the building, at least half of the pixels represent the probability that it is not a building. However, in actual production, the false deletion should be avoided as much as possible, so a more strict threshold range can be preset, and the threshold range can be 20-50%.
In order to facilitate understanding of the present invention, the method for identifying the expired buildings in the electronic map provided in this embodiment is further described below according to the principle of the method for identifying the expired buildings in the electronic map of the present invention:
the first step is as follows: acquiring a training dataset and a label dataset
The method comprises the steps of obtaining a training data set and a label data set, using the training data set as an alternative training data set to obtain public sub-meter satellite picture data provided by Google Earth, setting up the number of pictures according to training requirements, wherein the pictures generally need thousands of satellite pictures, and downloading vector boundary data of urban building blocks provided by an existing map development interface by using the label data set.
The second step is that: data pre-processing
Aligning the satellite picture and the map building acquired in the first step on a spatial position through data preprocessing, and cutting the satellite picture and the map building into an input data set required by a model, wherein the method specifically comprises the following steps of:
step 1: and unifying the projection coordinates of the satellite picture with the projection coordinates of the building outline vector boundary.
And 2, converting the vector building outline picture into a grid according to the resolution of the satellite picture.
And 3, cutting the boundary of the satellite picture and the boundary of the building outline picture according to the unified frame.
And 4, cutting the satellite picture and the building outline picture according to the unified geographic grid to obtain a training picture set and a label picture set, and taking the training picture set and the label picture set as training input of the deep convolutional neural network image segmentation model.
The third step: training a deep convolutional neural network image segmentation model and obtaining model parameters
The medical image segmentation model is applied to satellite image segmentation, parameters used when the medical image segmentation model is constructed cannot be directly applied due to different application scenes of the model in the image processing technology, the parameters of the model need to be retrained by using the data set obtained in the first step, namely, the training picture set and the label picture set obtained in the second step are input into the deep convolution neural network image segmentation model, and the model is trained to obtain model parameters.
The fourth step: identifying buildings in satellite imagery
As shown in fig. 2 and 3, after the satellite picture of the target operation area is segmented in the way of satellite picture preprocessing in step one, building identification is performed through trained model parameters, and a Sigmoid growth curve function (Sigmoid function) is used as an output layer to obtain a building probability image map of each area in the target operation area. Fig. 2 is a natural image picture, and after the natural image picture shown in fig. 2 is put into a model and the model is segmented by a trained deep convolutional neural network image, a target area building probability image shown in fig. 3 is obtained. The probability image identifies the probability through the model, and is influenced by the distinguishing degree of the features in the previous training sample. Inputting the target area picture set into a trained deep convolutional neural network image segmentation model, extracting corresponding features of the target area picture set by the model, and comparing the features with previous training data, for example, 99% of the training data have shadow places in the pictures which are buildings, and shadow features are extracted from a certain picture of the target area picture set, so that the model can judge that the probability of the region being a building is higher, and a higher probability value is given to obtain a target area building probability shadow map.
In the method for identifying the expired buildings in the electronic map provided by this embodiment, the training image set and the tag image set are input into the deep convolutional neural network image segmentation model, and the model can calculate some common features of the buildings in the images from the training data, such as differences between the material of the roof and the land, for example, a certain amount of shadows may be generated along with the buildings. An advantage of deep learning by means of a deep convolutional neural network image segmentation model is that it can extract many such features, up to some of which are not intuitively perceptible. By means of the characteristics, the model can map pixel values 0-255 of the satellite picture into probabilities 0-1, and forms the correspondence between the pixel values and the probabilities, for example, the probability value of 0.5 is corresponding to the pixel value of 128 gray.
The fifth step: screening and deleting the expired buildings
After the building is identified through the model, a high-resolution building probability image of the whole target area is obtained. The building probability image map cannot be directly applied to processing vector overdue buildings in the map, so that the building probability image map needs to be further processed, and the processing steps are as follows:
alternatively, the preset threshold value is 45%, and the buildings with the probability values lower than 45% in the probability image map of the buildings in the target area are identified as overdue buildings, and the overdue buildings are deleted.
The method for identifying the overdue buildings in the electronic map improves the updating speed of the electronic navigation map, has higher speed and precision for identifying the buildings by the deep convolutional neural network model compared with the existing manual processing mode, can control and modify the operation style and the operation quality of the electronic map, reduces the number of the overdue buildings processed manually, and reduces the labor cost.
The embodiment also provides a system for identifying an expired building in an electronic map, which includes:
a data acquisition unit configured to acquire a training dataset and a label dataset; in this embodiment, the acquired training data set is a satellite picture, and the tag data set is city building data, where the building data at least includes a building contour picture and building contour vector boundary data.
The image acquisition unit is configured for preprocessing the training data set and the label data set to obtain a training image set and a label image set; the method specifically comprises the following steps:
the projection coordinate unification subunit is configured to unify the projection coordinates of the satellite picture and the projection coordinates of the building contour vector boundary;
the picture rasterizing subunit is configured to convert the building outline picture into a grid according to the resolution of the satellite picture;
the picture boundary cutting subunit is configured for cutting the boundary of the satellite picture and the boundary of the building outline picture according to a uniform frame;
and the picture grid cutting subunit is configured for respectively cutting the satellite picture and the building outline picture according to the unified geographic grid to obtain a training picture set and a label picture set.
A medical image segmentation model (U-Net for short) is adopted as the deep Convolutional neural network image segmentation model.
The model training unit is configured to train a deep convolutional neural network image segmentation model according to a training picture set and the label picture set and obtain model parameters;
the building probability image map acquisition unit is configured for inputting the target area image set into a trained deep convolutional neural network image segmentation model, performing building identification through trained model parameters, and taking an S-shaped growth curve function as an output layer to acquire a building probability image map of the target area;
and the overdue building processing unit is configured for processing the building probability map of the target area, screening the probability values in the building probability map of the target area through a preset threshold value and deleting the overdue buildings. The method specifically comprises the following steps:
a threshold setting subunit configured to preset a threshold;
the overdue building identification subunit is configured for identifying a mark with a probability value lower than a preset threshold value in the target area building probability image map as an overdue building;
and the overdue building deleting subunit is configured to delete the overdue building.
It should be understood that the steps of the method for identifying the expired buildings in the electronic map correspond to sub-units recorded in the system for identifying the expired buildings in the electronic map. Thus, the operations and features described above for the system and the units included therein are equally applicable to the above method and will not be described again here.
According to the system for identifying the overdue buildings in the electronic map, the updating speed of the electronic navigation map is improved by arranging the picture acquisition unit, the model training unit, the building probability image acquisition unit, the overdue building processing unit and the like, compared with the existing manual processing mode, the speed and the accuracy of identifying the buildings by the deep convolutional neural network model are higher, the operation style and the operation quality of the electronic map can be controlled and modified, the number of the overdue buildings manually processed is reduced, and the labor cost is reduced.
The embodiment also provides a device for identifying expired buildings in an electronic map, which includes:
one or more processors;
a data store for storing data and one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method of any of the above.
The present embodiments also provide a computer readable storage medium storing a computer program which, when executed by a processor, implements a method as in any one of the above.
The device for identifying the overdue buildings in the electronic map provided by the embodiment is convenient for popularization of the technology for identifying the overdue buildings in the electronic map through the computer readable medium stored with the computer program.
Example two:
the same features of this embodiment and the first embodiment are not described again, and the different features of this embodiment and the first embodiment are:
the method for identifying the overdue buildings in the electronic map provided by the embodiment further comprises the following steps:
dividing the target area picture set to obtain a plurality of sub-area picture sets;
aggregating a plurality of sub-region picture sets;
respectively processing the sub-region image sets through a trained deep convolution neural network image segmentation model to obtain a plurality of sub-region building probability image maps;
and splicing the sub-region building probability image maps to obtain a target region building probability image map.
In order to facilitate understanding of the present invention, the method for identifying the expired buildings in the electronic map provided in this embodiment is further described below according to the principle of the method for identifying the expired buildings in the electronic map of the present invention:
image segmentation
If the target area is a city, the range of a single city may exceed the size of a computable image, so that the Map-Reduce method is used for dividing the building probability image recognition result of the single city into a plurality of sub-areas and performing parallel processing, namely the Map-Reduce method is used for distributing the probability image recognition work into a plurality of parts of data which can be processed by a computer, and then the data are processed respectively and finally the result is synthesized; for example, the size of a city in grade is about 50km-500km, the satellite picture for identification is an image with the resolution of 0.5m, if the whole process is carried out, the identification is a picture which is possibly 100000 pixels by 100000 pixels and is easy to exceed the software loading size, the identification result of a single city is divided into a plurality of sub-regions by adopting a Map-Reduce method for parallel processing, and the result is finally synthesized.
Two, aggregate image
And dividing building outlines in the region according to the range of the sub-regions, and aggregating the identified picture slices according to the geographic positions.
When the segmentation is performed, 200 pixels and 1000 pixels are theoretically enough, but the geographic length corresponding to 200 pixels is 100 meters and is likely to be smaller than a building, so that the extracted features are incomplete, and the model runs very slowly due to 1000 pixels. As an alternative, the picture is sliced into 400 × 400 pixels one picture (geographical area 200m × 200 m).
Thirdly, acquiring a sub-area building probability image map
As shown in fig. 4, the recognition images in the sub-regions are subjected to region statistics and averaged to obtain the probability that the building is recognized as a building. Suppose that there are 5 buildings, labeled 1-5, on the original electronic map. Whether the 5 buildings are overdue buildings or not needs to be judged according to the identification result, the identification result corresponding to each building is counted, and then the average is taken as the probability that the building is the building. Namely, the identification result in the range of the existing building is used as the evidence for judging whether the building exists. The final result is that the probability value corresponding to the buildings 1-4 is probably above 0.8, and the probability value corresponding to the building 5 is probably 0.1, and then the building 5 is directly identified as the overdue building through the probability value.
Fourthly, splicing the sub-region building probability image map
And after all the sub-regions are processed, fusing the vector building outlines in all the sub-regions into one image layer.
Fifthly, processing overdue buildings
In principle, the threshold value can be preset to 50%, i.e. within the range of the building, at least half of the pixels represent the probability that it is not a building. However, in actual production, the false deletion should be avoided as much as possible, so a more strict threshold range can be preset, and the threshold range can be 20-50%. Alternatively, the preset threshold is 50%, the buildings with the probability values lower than 50% in the probability image map of the buildings in the target area are identified as overdue buildings, and the overdue buildings are deleted.
The embodiment also provides a system for identifying an expired building in an electronic map, which specifically includes:
the target area picture set segmentation subunit is configured to segment the target area picture set to obtain a plurality of sub-area picture sets;
a sub-region picture set aggregation subunit configured to aggregate the plurality of sub-region picture sets;
the sub-region building probability image map acquisition subunit is configured to process the sub-region image sets respectively through a trained deep convolutional neural network image segmentation model to acquire a plurality of sub-region building probability image maps;
and the target area building probability image acquisition subunit is configured for splicing the sub-area building probability image to acquire a target area building probability image.
According to the method and the system for identifying the overdue buildings in the electronic map, the pictures of the target area are divided and aggregated, then the images of the sub-areas are processed through the trained deep convolutional neural network image division model, finally the probability image maps of all the sub-area buildings are spliced, and the overdue buildings are deleted, so that the overdue buildings in the electronic map can be processed for the target area with a city level or a larger area, and the popularization of the technology for identifying the overdue buildings in the electronic map is facilitated.
The foregoing description is only exemplary of the preferred embodiments of the invention and is illustrative of the principles of the technology employed. It will be appreciated by those skilled in the art that the scope of the invention herein disclosed is not limited to the particular combination of features described above, but also encompasses other arrangements formed by any combination of the above features or their equivalents without departing from the spirit of the invention. For example, the above features and (but not limited to) features having similar functions disclosed in the present invention are mutually replaced to form the technical solution.
Claims (12)
1. A method for identifying expired buildings in an electronic map, comprising:
acquiring a training data set and a label data set;
preprocessing the training data set and the label data set to obtain a training picture set and a label picture set;
training a deep convolutional neural network image segmentation model according to the training picture set and the label picture set and obtaining model parameters;
inputting the target area picture set into a trained deep convolution neural network image segmentation model, carrying out building identification through trained model parameters, and taking an S-shaped growth curve function as an output layer to obtain a target area building probability image map;
and processing the building probability image map of the target area, screening probability values in the building probability image map of the target area through a preset threshold value, and identifying overdue buildings.
2. The method for identifying the expired buildings in the electronic map as claimed in claim 1, wherein the training data set is a satellite picture and/or the tag data set is city building data, and the building data at least comprises a building outline picture and building outline vector boundary data.
3. The method for identifying the expired buildings according to claim 2, wherein the preprocessing the training data set and the tag data set to obtain a training picture set and a tag picture set comprises:
unifying the projection coordinates of the satellite picture and the projection coordinates of the building contour vector boundary;
converting the building outline picture into a grid according to the resolution of the satellite picture;
cutting the boundary of the satellite picture and the boundary of the building outline picture according to a uniform frame;
and respectively cutting the satellite picture and the building outline picture according to a unified geographical grid to obtain a training picture set and a label picture set.
4. The method for identifying the expired buildings in the electronic map as claimed in claim 1, wherein the deep convolutional neural network image segmentation model adopts a U-Net segmentation model.
5. The method for identifying the overdue buildings in the electronic map as claimed in claim 1, wherein the probability values in the probability image map of the buildings in the target area are filtered through a preset threshold, and after identifying the overdue buildings, the method further comprises deleting the overdue buildings.
6. The method for identifying the overdue buildings according to claim 1, wherein the buildings are identified through trained model parameters, and the S-shaped growth curve function is used as an output layer to obtain a probability image of the buildings in the target area, which comprises:
dividing the target area picture set to obtain a plurality of sub-area picture sets;
aggregating the plurality of sub-region picture sets;
respectively processing the sub-region image sets through a trained deep convolution neural network image segmentation model to obtain a plurality of sub-region building probability image maps;
and splicing the sub-area building probability image maps to obtain a target area building probability image map.
7. A system for identifying expired buildings in an electronic map, comprising:
a data acquisition unit configured to acquire a training dataset and a label dataset;
the picture acquisition unit is configured to preprocess the training data set and the label data set to obtain a training picture set and a label picture set;
the model training unit is configured to train a deep convolutional neural network image segmentation model according to the training picture set and the label picture set and obtain model parameters;
the building probability image map acquisition unit is configured for inputting the target area image set into a trained deep convolutional neural network image segmentation model, performing building identification through trained model parameters, and taking an S-shaped growth curve function as an output layer to acquire a building probability image map of the target area;
and the overdue building processing unit is configured for processing the target area building probability image, screening the probability values in the target area building probability image through a preset threshold value and identifying overdue buildings.
8. The system for identifying expired buildings according to claim 7, wherein the training data set is a satellite picture and/or the tag data set is city building data, and the building data at least comprises a building outline picture and building outline vector boundary data.
9. The system for identifying the expired buildings in the electronic map as claimed in claim 8, wherein the picture acquiring unit comprises:
the projection coordinate unifying subunit is configured to unify the projection coordinates of the satellite picture and the projection coordinates of the building contour vector boundary;
the picture rasterizing subunit is configured to convert the building outline picture into a grid according to the resolution of the satellite picture;
a picture boundary clipping subunit configured to clip the boundary of the satellite picture and the boundary of the building outline picture according to a unified frame;
and the picture grid cutting subunit is configured to respectively cut the satellite picture and the building outline picture according to a unified geographic grid to obtain a training picture set and a label picture set.
10. The system for identifying the expired buildings in the electronic map as claimed in claim 9, wherein the deep convolutional neural network image segmentation model adopts a U-Net segmentation model.
11. The system for identifying the overdue buildings according to claim 7, wherein the overdue building processing unit is further configured to filter the probability values in the probability image map of the buildings in the target area through a preset threshold, and delete the overdue buildings after identifying the overdue buildings.
12. The system for identifying the expired buildings in the electronic map as claimed in claim 7, wherein the building probability image map acquiring unit comprises:
the target area picture set segmentation subunit is configured to segment the target area picture set to obtain a plurality of sub-area picture sets;
a sub-region picture set aggregation subunit configured to aggregate the plurality of sub-region picture sets;
the sub-region building probability image map acquisition subunit is configured to process the sub-region image sets respectively through a trained deep convolutional neural network image segmentation model to acquire a plurality of sub-region building probability image maps;
and the target area building probability image acquisition subunit is configured for splicing the sub-area building probability image to acquire a target area building probability image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811646297.3A CN111382645B (en) | 2018-12-29 | 2018-12-29 | Method and system for identifying overdue building in electronic map |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811646297.3A CN111382645B (en) | 2018-12-29 | 2018-12-29 | Method and system for identifying overdue building in electronic map |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111382645A true CN111382645A (en) | 2020-07-07 |
| CN111382645B CN111382645B (en) | 2023-11-28 |
Family
ID=71218333
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811646297.3A Active CN111382645B (en) | 2018-12-29 | 2018-12-29 | Method and system for identifying overdue building in electronic map |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111382645B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115795070A (en) * | 2023-01-30 | 2023-03-14 | 智道网联科技(北京)有限公司 | Method and device for detecting map overdue guideboard, electronic equipment and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107516096A (en) * | 2016-06-15 | 2017-12-26 | 阿里巴巴集团控股有限公司 | A kind of character identifying method and device |
| CN108229364A (en) * | 2017-12-28 | 2018-06-29 | 百度在线网络技术（北京）有限公司 | Contour of building generation method, device, computer equipment and storage medium |
| CN108346428A (en) * | 2017-09-13 | 2018-07-31 | 腾讯科技（深圳）有限公司 | Voice activity detection and its method for establishing model, device, equipment and storage medium |
| CN108376232A (en) * | 2018-01-04 | 2018-08-07 | 北京星衡科技有限公司 | A kind of method and apparatus of automatic interpretation for remote sensing image |
| CN108764226A (en) * | 2018-04-13 | 2018-11-06 | 顺丰科技有限公司 | Image text recognition methods, device, equipment and its storage medium |
| CN108776772A (en) * | 2018-05-02 | 2018-11-09 | 北京佳格天地科技有限公司 | Across the time building variation detection modeling method of one kind and detection device, method and storage medium |
| CN108898603A (en) * | 2018-05-29 | 2018-11-27 | 北京佳格天地科技有限公司 | Plot segmenting system and method on satellite image |
| CN109034073A (en) * | 2018-07-30 | 2018-12-18 | 深圳大学 | Predict method, system, equipment and the medium of building demolition waste yield |
-
2018
- 2018-12-29 CN CN201811646297.3A patent/CN111382645B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107516096A (en) * | 2016-06-15 | 2017-12-26 | 阿里巴巴集团控股有限公司 | A kind of character identifying method and device |
| CN108346428A (en) * | 2017-09-13 | 2018-07-31 | 腾讯科技（深圳）有限公司 | Voice activity detection and its method for establishing model, device, equipment and storage medium |
| CN108229364A (en) * | 2017-12-28 | 2018-06-29 | 百度在线网络技术（北京）有限公司 | Contour of building generation method, device, computer equipment and storage medium |
| CN108376232A (en) * | 2018-01-04 | 2018-08-07 | 北京星衡科技有限公司 | A kind of method and apparatus of automatic interpretation for remote sensing image |
| CN108764226A (en) * | 2018-04-13 | 2018-11-06 | 顺丰科技有限公司 | Image text recognition methods, device, equipment and its storage medium |
| CN108776772A (en) * | 2018-05-02 | 2018-11-09 | 北京佳格天地科技有限公司 | Across the time building variation detection modeling method of one kind and detection device, method and storage medium |
| CN108898603A (en) * | 2018-05-29 | 2018-11-27 | 北京佳格天地科技有限公司 | Plot segmenting system and method on satellite image |
| CN109034073A (en) * | 2018-07-30 | 2018-12-18 | 深圳大学 | Predict method, system, equipment and the medium of building demolition waste yield |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115795070A (en) * | 2023-01-30 | 2023-03-14 | 智道网联科技(北京)有限公司 | Method and device for detecting map overdue guideboard, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111382645B (en) | 2023-11-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Rüther et al. | Application of snakes and dynamic programming optimisation technique in modeling of buildings in informal settlement areas | |
| CN113240691A (en) | Medical image segmentation method based on U-shaped network | |
| Rakhmadi et al. | Connected component labeling using components neighbors-scan labeling approach | |
| CN110992384B (en) | Semi-automatic image data labeling method, electronic device and storage medium | |
| WO2021077947A1 (en) | Image processing method, apparatus and device, and storage medium | |
| CN112883900B (en) | Method and device for bare-ground inversion of visible images of remote sensing images | |
| WO2021093345A1 (en) | High-precision semi-automatic image data labeling method, electronic apparatus, and storage medium | |
| CN113284154B (en) | Steel coil end face image segmentation method and device and electronic equipment | |
| CN110599453A (en) | Panel defect detection method and device based on image fusion and equipment terminal | |
| CN111415364A (en) | Method, system and storage medium for converting image segmentation samples in computer vision | |
| CN110837839A (en) | High-precision unmanned aerial vehicle orthoimage manufacturing and data acquisition method | |
| CN116597270A (en) | Road damage target detection method based on attention mechanism integrated learning network | |
| CN111914596B (en) | Lane line detection method, device, system and storage medium | |
| CN113033386B (en) | High-resolution remote sensing image-based transmission line channel hidden danger identification method and system | |
| CN109657728B (en) | Sample production method and model training method | |
| CN115705685A (en) | Image data set labeling method and device and electronic equipment | |
| CN112990237A (en) | Subway tunnel image leakage detection method based on deep learning | |
| CN111382645B (en) | Method and system for identifying overdue building in electronic map | |
| CN112435274A (en) | Remote sensing image planar ground object extraction method based on object-oriented segmentation | |
| CN111523543A (en) | Tunnel surface defect positioning method based on learning | |
| CN115546824B (en) | Taboo picture identification method, apparatus and storage medium | |
| CN115132370A (en) | Flow adjustment auxiliary method and device based on machine vision and deep learning | |
| CN115359091A (en) | Armor plate detection tracking method for mobile robot | |
| CN113536928A (en) | High-efficiency unsupervised pedestrian re-identification method and device | |
| CN112215205A (en) | Target identification method and device, computer 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 |