CN109977191A - Problem map detection method, device, electronic equipment and medium - Google Patents
Problem map detection method, device, electronic equipment and medium Download PDFInfo
- Publication number
- CN109977191A CN109977191A CN201910257029.0A CN201910257029A CN109977191A CN 109977191 A CN109977191 A CN 109977191A CN 201910257029 A CN201910257029 A CN 201910257029A CN 109977191 A CN109977191 A CN 109977191A
- Authority
- CN
- China
- Prior art keywords
- map
- map datum
- detection model
- test object
- training
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
Abstract
The embodiment of the invention discloses a kind of problem map detection method, device, electronic equipment and media, wherein this method comprises: obtaining target electronic map datum, wherein includes at least one test object in target electronic map datum；By target electronic map datum input map detection model trained in advance, the output of detection model determines target electronic map datum with the presence or absence of layout drawing mistake according to the map, wherein, map detection model is obtained based on the setting regions scale parameter training of each test object, and the output of map detection model includes the mark on the target electronic map datum there is no layout drawing mistake to test object.The embodiment of the present invention for the first time detects depth learning technology introducing problem map automatically, it solves the problems, such as the existing map detection method low efficiency for relying on artificial detection and human cost is high, the automatic mark for realizing the automatic detection and problem map of electronic map, improves detection efficiency.
Description
Technical field
The present embodiments relate to mapping technical field of geographic information more particularly to a kind of problem map detection methods, dress
It sets, electronic equipment and medium.
Background technique
In Ancient Times in China, " version " is the books for taking notes or keeping accounts for registering the registered permanent residence and soil, and " figure " refers to map, and " domain " represents household register
And map, and it is gradually evolved into the synonym in national territory.As time goes on, map and daily life cease manner of breathing
It closes, directly affects socio-economic development or even national security.However, map using security issues event takes place frequently in recent years, if
It is that " problem map " circulates, adverse effect, damage national sovereignty safety and interests, Geng Huicheng not only is generated to the public
For the handle of the attack of hostile force overseas, or even initiation international discord, must attract great attention to this.
It is reported that China's the Map Market Rapid Expansion at present, prosperity and development.Since 2012, the ground of annual public publication
About more than 2000 kinds of figure, nearly 300,000,000 width.Only 2016, map of navigation electronic just completed 6,600,000,000 yuan of total value of service, had interconnection entoilage
The mapping qualification unit of figure service qualification completes service total value up to 28,200,000,000 yuan.But at the same time, the exhibition of the Map Market and to map
Show that " problem maps " such as China's territorial sovereignty, safety and maritime rights and interests is damaged in use still to remain incessant after repeated prohibition.According to statistics, 2012
Since, the whole nation is carried out law enforcement more than 11000 times altogether, investigates and prosecutes Related Cases nearly 1000, is taken over all kinds of illegal maps and is produced
Product are cured more than 200,000 parts, and completing more than 1000 has the disposition of " map " website.During the Map Market checks within 2016, all kinds of
The Map Market, cultural goods market, Internet map service unit, exhibition (exhibition), memorial museum, museum etc., find and according to
Method has investigated and prosecuted large quantities of illegal, violation cases.These cases, some leakage pictures, the important island in the China Cuo Hua, state boundary etc., jeopardize
National sovereignty；Some uploads in internet, marks sensitive and classified information, it is open publish, illegal transaction concerning security matters map etc., danger
And national security；Some is not indicated concerned countries and area by China's political diplomacy opinion, and illegal mapping, volume
Figure, offer Map Services etc., damage national interests.That especially some internets publish, " problem map " from media releasing,
Distribution is wide, propagates fastly, harm is big.
Currently, relying primarily on artificial visual interpretation to " problem map " checking method." problem map " and correct map
It compares, existing difference may be very small, and testing staff needs skillfully to grasp the method for drafting of correct map, and detection speed is slow,
And large labor intensity.
Summary of the invention
The embodiment of the present invention provides a kind of problem map detection method, device, electronic equipment and medium, with improving problem
The detection efficiency of figure, and guarantee detection accuracy.
In a first aspect, the embodiment of the invention provides a kind of problem map detection methods, this method comprises:
Obtain target electronic map datum, wherein include at least one test object in the target electronic map datum；
By target electronic map datum input map detection model trained in advance, according to the map detection model
Output determine the target electronic map datum with the presence or absence of layout drawing mistake, wherein the map detection model is based on
The setting regions scale parameter training of each test object obtains, and the output of the map detection model is included in that there is no domains
Draw the mark on the target electronic map datum of mistake to test object.
Second aspect, the embodiment of the invention also provides a kind of problem map detection device, which includes:
Module is obtained, for obtaining target electronic map datum, wherein include at least in the target electronic map datum
One test object；
Detection module, for the map detection model that target electronic map datum input is trained in advance, according to institute
The output for stating map detection model determines the target electronic map datum with the presence or absence of layout drawing mistake, wherein describedly
Figure detection model is obtained based on the setting regions scale parameter training of each test object, the output packet of the map detection model
Include the mark on the target electronic map datum there is no layout drawing mistake to test object.
The third aspect, the embodiment of the invention also provides a kind of electronic equipment, comprising:
One or more processors；
Storage device, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processing
Device realizes the problem map detection method as described in any embodiment of the present invention.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer
Program realizes the problem map detection method as described in any embodiment of the present invention when the program is executed by processor.
The embodiment of the present invention passes through the target electronic map datum input that will acquire map detection model trained in advance, root
Determine target electronic map datum with the presence or absence of layout drawing mistake according to the output of map detection model, wherein map detects mould
Type is obtained based on the setting regions scale parameter training of each test object, and the output of map detection model is included in that there is no versions
Figure draws the mark on the target electronic map datum of mistake to test object, i.e., by for the first time asking depth learning technology introducing
Topic map detect automatically, solves the map detection method low efficiency of existing dependence artificial detection and human cost is high asks
Topic realizes the automatic mark of the automatic detection and problem map of electronic map, improves detection efficiency, and ensure that detection
Accuracy.
Detailed description of the invention
Fig. 1 is the flow chart of the problem of embodiment of the present invention one provides map detection method；
Fig. 2 is the flow chart of problem map detection method provided by Embodiment 2 of the present invention；
Fig. 3 is the structural schematic diagram of the problem of embodiment of the present invention three provides map detection device；
Fig. 4 is the structural schematic diagram for a kind of electronic equipment that the embodiment of the present invention four provides.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
Embodiment one
Fig. 1 is the flow chart of the problem of embodiment of the present invention one provides map detection method, and the present embodiment is applicable to pair
The case where the problems in electronic map map is detected, this method can be executed by problem map detection device, the device
It can be realized, and can be integrated on an electronic device by the way of software and/or hardware, which includes server.
As shown in Figure 1, problem map detection method provided in this embodiment may include:
S110, target electronic map datum is obtained, wherein include at least one detection pair in target electronic map datum
As.
Target electronic map datum refers to arbitrary electronic map data to be detected, including world map, country map and city
Any types such as city's map.Specific region in test object feeling the pulse with the finger-tip mark electronic map data, it is true according to current detection demand
It is fixed.For example, current detection demand is whether the detection country map middle finger corresponding area information of part of inquiring after the health of one's parents correct, then test object
Subregion of the province in country map is specified for this.
S120, by target electronic map datum input map detection model trained in advance, detection model according to the map
It exports and determines that target electronic map datum whether there is layout drawing mistake, wherein map detection model is based on each detection pair
The setting regions scale parameter training of elephant obtains, and the output of map detection model includes in the target that layout drawing mistake is not present
To the mark of test object on electronic map data.
Consider that electronic map data and common image data are different, after geographic area determines, electronic map
The chamfered shape for each sub-regions for including in data be usually it is determining, will not format with electronic map data, size
Variation with the factors such as resolution ratio and change.Therefore, in the training process of map detection model, setting and each test object one
One corresponding regional percentage parameter, the frame region which determines can be will test just in object is enclosed in,
I.e. in the frame region other than the region area that test object occupies, including interference region area it is minimum.Electronically
In the detection process of figure, map outline is the whether correct key factor of determining map, therefore, the side that regional percentage parameter determines
The matching degree of frame region and test object directly affects the accuracy of map detection.The interference region that frame region includes is smaller,
The region contour of test object is more clear, higher for the detection accuracy of test object.
The determination process of the setting regions scale parameter of each test object comprises determining that the target ratio of electronic map data
Example ruler, frequency of use highest scale bar when which currently can draw or show electronic map；Using target
In the electronic map data that scale bar indicates, the ratio between the corresponding geometry side length of test object is determined；By the geometry side length
Between ratio-dependent be the setting regions scale parameter of test object, for example, will test the ratio of width to height of object as its setting
Regional percentage parameter.
In the present embodiment, map detection model is obtained using the training of deep learning method.In the nerve net according to building
When network structured training model, in addition to being arranged common network model hyper parameter, such as learning rate, the number of iterations, initial method,
The training structure of model, the threshold value of training process and gradient Pruning strategy etc. will be with test objects also directed to each test object
Corresponding setting regions scale parameter is set as training parameter, simultaneously participates in model training.Have benefited from the only of electronic map data
The chamfered shape of characteristic, different test objects is variant, therefore, the mutual not phase of the setting regions scale parameter of each test object
Together.It is configured, is may be implemented for difference using the corresponding setting regions scale parameter of each test object as training parameter
Test object carry out accurately selected effect, be the accuracy and targetedly important foundation for guaranteeing model training.Work as model
During training is completed for electronic map detection, setting regions scale parameter is also used as screening conditions, to target electricity
Region on sub- map datum is screened, to quickly orient test object.
Illustratively, during model training, for country map class training data, test object is subregion respectively
A, the setting area of four sub-regions is arranged according to the ratio of width to height of four sub-regions on the map under target proportion ruler in B, C and D
Domain scale parameter is respectively 1:3.61,3.62:1,2.80:1,1:1.28 and 1:1.53.
Target electronic map datum is detected using map detection model, the difference of result can be exported by model
It is different, determine that target electronic map datum with the presence or absence of layout drawing mistake, realizes the intellectualized detection to electronic map data.Example
Such as, if layout drawing mistake is not present in target electronic map datum, model output result, which may is that, will test object at this
Visual frame mark is carried out on target electronic map datum；If target electronic map datum there are layout drawing mistake,
I.e. the target electronic map is problem map, then model output result may is that directly with text prompt or voice prompting etc.
Alarm mode exports the target electronic map datum and belongs to problem map, without carrying out frame mark to test object, that is, passes through
The difference that model exports result realizes the automatic mark to problem map.When there are multiple test objects on target electronic map
When, as long as mistake is drawn in the presence of one of test object, which belongs to problem map, Duo Gejian
When survey object draws correct, which belongs to correct map.It should be noted that in electronic map data
The detection zone of limited quantity that is usually fixed of test object, the correct drafting situation of test object is unique, and is examined
The mistake of survey object draws situation, and there are a variety of possibilities, and therefore, only study identifies in the training process of map detection model
The test object correctly drawn, when model is applied to the detection of problem map, model is exported in result also only to there is no draw
Test object in the electronic map data of mistake carries out frame mark, is based on exclusive method, does not export frame mark detection pair
The electronic map data of elephant then belongs to problem map.It is of course also possible to the position for showing zone errors on problem map of adaptability
It sets, in order to carry out subsequent amendment to the problem map.In addition, the test object for including in target electronic map datum is
Trained test object is participated in map detection model training process.
The technical solution of the present embodiment passes through the target electronic map datum input that will acquire map detection trained in advance
Model, the output of detection model determines target electronic map datum with the presence or absence of layout drawing mistake according to the map, wherein map
Detection model is obtained based on the setting regions scale parameter training of each test object, and the output of map detection model is included in not
There are, to the mark of test object, solve existing dependence artificial detection on the target electronic map datum of layout drawing mistake
Map detection method low efficiency and the high problem of human cost, realize the automatic detection of electronic map and problem map from
Dynamic mark, improves detection efficiency, and ensure that the accuracy of detection.
Embodiment two
Fig. 2 is the flow chart of problem map detection method provided by Embodiment 2 of the present invention, and the present embodiment is in above-mentioned reality
Further progress optimizes on the basis of applying example.As shown in Fig. 2, this method may include:
S210, an at least simple electric map datum is obtained, as training set, wherein every simple electric map number
According to upper, setting regions scale parameter corresponding with each test object is used to mark each test object.
For example, enough electronic map datas can be searched for from internet as training set, and electronically to every
Diagram data carries out frame mark, to determine the test object of model training process.In order to guarantee that inspection is accurately positioned in training set
Object is surveyed, frame mark is realized by the way of manually marking.Illustratively, user is according to determining for each test object in advance
Setting regions scale parameter, using the regional center of test object as frame center, making corresponding rectangle frame will test object
In adaptability is enclosed in.
Optionally, an at least simple electric map datum is obtained, as training set, comprising:
An at least simple electric map datum is cleaned, using the simple electric map datum after cleaning as instruction
Practice collection.Wherein, data cleansing refers to the data that mistake is drawn in Rejection of samples electronic map data.Use the correct electronics of drafting
Map datum is as training set, the accuracy of the training pattern guaranteed.
Optionally, model training will be being carried out in the default neural network structure of training set input, is obtaining map detection model
Before, the training process of map detection model further include: every simple electric map datum in training set is increased in real time
Strength reason, to be based on obtaining training set progress model training after enhancing processing.Wherein, real time enhancing processing includes but is not limited to:
Data decentralization, data normalization, data ZCA (Zero-phase Component Analysis) albefaction, data gray
Change, data rotate at random, data level offset, data offset of vertical, data size scale at random, data random channel deviates,
The overturning, data binaryzation and data random cropping vertically at random of the overturning of data Random Level, data.By to simple electric map
Data carry out real time enhancing processing, and real-time transform and the update of simple electric map datum may be implemented, increase sample data
Diversity；And the transformation simple electric map datum handled enhancing is stored in a manner of caching, and works as sample
After data are used for model training, which discharges at random, the memory space of occupancy electronic equipment that can't be excessive, thus
It realizes and alleviates storage pressure on the basis of enriching sample data.
S220, model training will be carried out in the default neural network structure of training set input, obtains map detection model,
In, presetting in neural network structure includes that network is suggested in region.
Default neural network structure in the present embodiment may include Faster-RCNN and Mask-RCNN network structure, this
Include that network (Region Proposal Network, RPN) is suggested in region in a little neural network structures, is generated using RPN network
ROI (Region Of Interest, region of interest).During model training, user is by the setting area of each test object
Domain scale parameter is preset as training parameter, and region suggests that network just utilizes the setting regions scale parameter, will be right
The test object answered is determined in the form of frame, then proceedes to carry out Map recognition based on default neural network structure, most
Training obtains map detection model eventually.
For example, being based on Faster-RCNN network structure, when being trained to RPN network, for each simple electric
Diagram data is classified simultaneously using multitask loss function (multi-task loss) and returns calculatings, to class probability with
Bezel locations carry out joint training.Multitask loss function includes that Classification Loss function (Softmax Loss) and frame return damage
It loses function (Smooth L1Loss), is defined as follows:
Wherein: NclsIndicate input sample number in mini-batch, NregIndicate the quantity of anchor (anchor point) position, i
It is the index of anchor in mini-batch；piIt is the probability that the anchor is predicted to be a target；It is in marker samples
Positive sample probability, andIt is a vector, represents the position that RPN predicts the anchor
It sets；For actual position；Positive sample therein refers to test object, negative sample middle finger image background.
It is the Classification Loss of foreground/background:
It is bezel locations loss:
The error situation of map of considering a problem has countless possibility, such as color mistake, contour line mistake, region missing
Deng, therefore, during model training, suggest that network only marks positive sample in the original image of electronic map data using region,
Rather than label negative sample.For example, setting threshold value T=0.7, target of the probability P greater than T is positive sample, and probability P is less than or equal to
The target of T is negative sample, and the target that P is greater than T is only marked in the original image of electronic map data.
The corresponding loss function of map detection model that final training obtains includes generating the loss of ROI and ROI classifying
It for the loss of specific category, that is, determines the corresponding loss in test object region position, and will test object classification and be positive
True map area or the corresponding loss of wrong map area:
Loss=Lrpn+Lspe
Wherein:
Lrpn=Lrpn_class+Lrpn_bbox,Lrpn_classIt is the Classification Loss that network is suggested in region, i.e. network by the inspection of proposition
Survey the loss that frame is divided into foreground/background；Lrpn_bboxIt is the bezel locations loss that network is suggested in region；Lspe=Lspe_class+
Lspe_bbox, Lspe_classIt is the Classification Loss that the region of interest of proposition is classified as to specific category；Lspe_bboxIt is to suggest region
The detection block and true frame (frame of the test object marked in training set) that network proposes compare and the bezel locations of calculating
Loss.
S230, target electronic map datum is obtained, wherein include at least one detection pair in target electronic map datum
As.
In the present embodiment, the test object for including in target electronic map datum belong to during model training by with
In the test object of training pattern.
S240, by target electronic map datum input map detection model trained in advance, detection model according to the map
It exports and determines that target electronic map datum whether there is layout drawing mistake.
Based on the above technical solution, optionally, training set is inputted in default neural network structure and carries out model
Training, obtains map detection model, comprising:
Using the feature extraction network of default neural network structure, the feature of every simple electric map datum is extracted, is obtained
To the characteristic image of every simple electric map datum；
Network, the setting area based on each test object are suggested into the region that characteristic image inputs default neural network structure
Domain scale parameter determines the suggestion areas of each test object on every simple electric map datum；
Every characteristic image for carrying suggestion areas is inputted in default neural network structure, and is adjusted according to area differentiation
The training parameter of default neural network structure, trained processing obtain map detection model, wherein area differentiation refers to every sample
Difference on electronic map data, between the suggestion areas position of each test object and its tab area position.
It is below that Faster-RCNN network structure illustrates with default neural network structure:
Simple electric map datum is inputted in Faster-RCNN network structure, first with the feature of Faster-RCNN
The feature that the shared convolutional layer in network extracts simple electric map datum is extracted, the characteristic pattern of simple electric map datum is obtained
Picture；The region of the characteristic image input Faster-RCNN obtained through shared convolutional layer is suggested in network, in the every of sliding window
9 rectangular windows (3 kinds of length-width ratios, 3 kinds of scales), an a rectangular window i.e. anchor are set in the corresponding original picture of a pixel
Point；Suggest in network in region, continues to input characteristic image into convolutional layer progress convolutional calculation, then according to convolution results and anchor
Point carries out recurrence and classified calculating, exports the suggestion areas of test object；The characteristic image for carrying suggestion areas is inputted into the pond ROI
Change layer, generate fixed-size characteristic pattern again, the fixed characteristic pattern of size is inputted whether full articulamentum determines test object
There are layout drawing mistakes, and position the test object to make mistake.Wherein, when the suggestion of network output test object is suggested in region
Behind region, the position of the bezel locations of the test object manually marked in simple electric map datum and suggestion areas is carried out
Compare, the training parameter of neural network structure is preset according to the discrepancy adjustment of the two, for example, being cut according to the discrepancy adjustment gradient
Policing parameter, to guarantee to train the accuracy of obtained map detection model.
Further, the feature extraction network for presetting neural network structure includes feature pyramid network；Correspondingly, utilizing
The feature extraction network of default neural network structure, extracts the feature of every simple electric map datum, obtains every sample electricity
The characteristic image of sub- map datum, comprising:
Analysis On Multi-scale Features processing is carried out to every simple electric map datum using feature pyramid network；
Based on Analysis On Multi-scale Features treated simple electric map datum, feature extraction is carried out, to obtain characteristic image.For
Solve the problems, such as, introduced feature pyramid network (Feature lower for the electronic map data recognition correct rate under multiple dimensioned
Pyramid Networks, FPN), to reinforce ability to express of the obtained map detection model of training on multiple dimensioned, reach comprehensive
The map feature under multiple scales is closed, the effect for the influence that mutative scale effect detects electronic map is ignored, to accurately detect
Multiple dimensioned electronic map data realizes the self-adapting detecting to multiple dimensioned problem map datum.For example, utilizing feature gold word
Tower network can be automatic to the progress of the electronic map data of ultrahigh resolution (such as 30000*30000) down-sampled, to solve to pass
The image processing tool of system can not directly handle the problem of electronic map data of ultrahigh resolution.
Based on the above technical solution, optionally, the training process of map detection model further include:
Obtain electronic map data verifying collection；
The electronics that electronic map data verifying is concentrated using at least one map detection model obtained in training process
Map datum is detected, and is adopted according to the current detection result of electronic map data verifying collection to current map detection model
Default neural network structure is assessed；
The default neural network structure that assessed value is greater than preset threshold is determined as target nerve network structure, and will be based on
The map detection model that the training of target nerve network structure obtains is determined as target map detection model, to be examined using target map
Survey model inspection target electronic map datum.
Illustratively, in data preparation stage, from the connection a large amount of electronic map data of online collection, according to a certain percentage,
Construct training set and verifying collection.Training set be used for model training, verifying collection for during model training to trained mould
Type is verified.
Base net network is a part of default neural network structure, chooses different base net networks, presets neural network structure
Performance is different, and then causes the performance of the map detection model obtained based on the training of default neural network structure different.In model
In training process, optimal base network can be determined first, is then based on the optimal base network, determine optimal default neural network
Structure.
Illustratively, firstly, network hyper parameter is arranged, based on same default neural network structure to different base net networks
It is screened, selectable base net network includes but is not limited to: ResNet50, ResNet101, VGG19, ResNext101-
64x4d, ResNext101-32x8d, ResNext152-32x8d-IN5k etc..Specifically, during model training, using prison
Visual organ assesses the model obtained based on different base network trainings, is weighed in terms of model calculating speed and detection accuracy
Weighing apparatus determines that the optimal base network in identical hyper parameter, such as optimal base network can be ResNet101.Then, it utilizes
The model inspection result for verifying collection is preset neural network structure to the difference for including the optimal base network and is assessed.Wherein, may be used
The default neural network structure of selection includes Faster-RCNN and Mask-RCNN etc..Specifically, can be in the whole of model training
In a period, the map detection model obtained based on different default neural network structure training is carried out using monitor comprehensive
Monitoring and evaluation saves optimal model and the corresponding default neural network structure of ideal model in entire cycle of training,
Default neural network structure, that is, target nerve the network structure.In default neural network structure or the evaluation process of training pattern
In, it can be using monitoring object as assessment factor, such as the corresponding value of monitoring object smaller, training pattern and corresponding nerve
The assessment result of network structure is higher.Monitoring object may include: training loss, the loss of region candidate network training, verifying damage
Region candidate network verification of becoming estranged loss etc..The assessed value of neural network structure is higher, and the detection accuracy of map detection model is got over
It is high.
In addition, data preparation stage can also create test set simultaneously, for carrying out to trained map detection model
Test.Training set, verifying collection and test set keep certain mathematical distribution, for example, proportions are distributed as 8:1:1.According to survey
The model inspection of collection is tried as a result, different confidence threshold values is adjusted, so that model inspection precision meets the requirements.
The technical solution of the present embodiment by obtaining training set, and every simple electric map datum in training set first
Upper use setting regions scale parameter corresponding with each test object marks each test object, then inputs training set pre-
If carrying out model training in neural network structure, map detection model is obtained, the target electronic map datum that finally will acquire is defeated
Enter map detection model, the output of detection model determines that target electronic map datum is wrong with the presence or absence of layout drawing according to the map
Accidentally, it solves the problems, such as the existing map detection method low efficiency for relying on artificial detection and human cost is high, realize electronics
The automatic mark of the automatic detection and problem map of map, improves detection efficiency, and ensure that the accuracy of detection；Also,
Suggest network based on region in default neural network structure, it is accurately fixed according to the setting regions scale parameter of test object to realize
The effect of position test object；In addition, being solved multiple dimensioned by the introduced feature pyramid network in default neural network structure
Under the lower problem of problem map datum recognition correct rate, realize the adaptive inspection to multiple dimensioned problem map datum
It surveys.
Embodiment three
Fig. 3 is the structural schematic diagram of the problem of embodiment of the present invention three provides map detection device, and the present embodiment is applicable
In detected to the problems in electronic map map the case where.The device can realize by the way of software and/or hardware,
And can integrate on an electronic device, which includes server.
As shown in figure 3, problem map detection device provided in this embodiment may include obtaining module 310 and detection module
320, in which:
Module 310 is obtained, for obtaining target electronic map datum, wherein include at least in target electronic map datum
One test object；
Detection module 320, for the map detection model that the input of target electronic map datum is trained in advance, according to the map
The output of detection model determines target electronic map datum with the presence or absence of layout drawing mistake, wherein map detection model is based on
The setting regions scale parameter training of each test object obtains, and the output of map detection model is included in that there is no layout drawings
To the mark of test object on the target electronic map datum of mistake.
Optionally, which further includes map detection model training module, which includes:
Training set acquiring unit, for obtaining an at least simple electric map datum, as training set, wherein every
On simple electric map datum, each test object is marked using setting regions scale parameter corresponding with each test object；
Model training unit carries out model training for inputting training set in default neural network structure, obtains map
Detection model, wherein include that network is suggested in region in default neural network structure.
Optionally, model training unit includes:
Characteristic image determines subelement, for the feature extraction network using default neural network structure, extracts every sample
The feature of this electronic map data obtains the characteristic image of every simple electric map datum；
Suggestion areas determines subelement, and network is suggested in the region for characteristic image to be inputted default neural network structure,
Setting regions scale parameter based on each test object determines building for each test object on every simple electric map datum
Discuss region；
Model training subelement, every characteristic image for that will carry suggestion areas input default neural network structure
In, and the training parameter for presetting neural network structure is adjusted according to area differentiation, trained processing obtains map detection model,
In, area differentiation refers on every simple electric map datum, the suggestion areas position of each test object and its marked area
Difference between the position of domain.
Optionally, feature extraction network includes feature pyramid network；
Correspondingly, characteristic image determines that subelement is used for:
Analysis On Multi-scale Features processing is carried out to every simple electric map datum using feature pyramid network；
Based on Analysis On Multi-scale Features treated simple electric map datum, feature extraction is carried out, to obtain characteristic image.
Optionally, map detection model training module further include:
Training set is being inputted default neural network for executing in model training unit by data real time enhancing processing unit
Model training is carried out in structure, before obtaining the operation of map detection model, to every simple electric map number in training set
According to progress real time enhancing processing.
Optionally, training set acquiring unit is used for:
An at least simple electric map datum is cleaned, using the simple electric map datum after cleaning as instruction
Practice collection.
Optionally, map detection model training module further include:
Verifying collection acquiring unit, for obtaining electronic map data verifying collection；
Neural network structure assessment unit, for utilizing at least one map detection model obtained in training process to electricity
The electronic map data that sub- map datum verifying is concentrated is detected, and according to the current detection knot of electronic map data verifying collection
Fruit assesses default neural network structure used by current map detection model；
Target nerve network structure determination unit, the default neural network structure for assessed value to be greater than preset threshold are true
It is set to target nerve network structure, and the map detection model obtained based on the training of target nerve network structure is determined as target
Map detection model, to detect target electronic map datum using target map detection model.
Provided by any embodiment of the invention ask can be performed in problem map detection device provided by the embodiment of the present invention
Map detection method is inscribed, has the corresponding functional module of execution method and beneficial effect.Not detailed description is interior in the present embodiment
Holding can be with reference to the description in embodiment of the present invention method.
Example IV
Fig. 4 is the structural schematic diagram for a kind of electronic equipment that the embodiment of the present invention four provides.Fig. 4, which is shown, to be suitable for being used in fact
The block diagram of the example electronic device 412 of existing embodiment of the present invention.The electronic equipment 412 that Fig. 4 is shown is only an example,
Should not function to the embodiment of the present invention and use scope bring any restrictions.
As shown in figure 4, electronic equipment 412 is showed in the form of universal electronic device.The component of electronic equipment 412 can wrap
Include but be not limited to: one or more processor 416, storage device 428 connect different system components (including storage device 428
With processor 416) bus 418.
Bus 418 indicates one of a few class bus structures or a variety of, including storage device bus or storage device control
Device processed, peripheral bus, graphics acceleration port, processor or total using the local of any bus structures in a variety of bus structures
Line.For example, these architectures include but is not limited to industry standard architecture (Industry Subversive
Alliance, ISA) bus, microchannel architecture (Micro Channel Architecture, MAC) bus is enhanced
Isa bus, Video Electronics Standards Association (Video Electronics Standards Association, VESA) local are total
Line and peripheral component interconnection (Peripheral Component Interconnect, PCI) bus.
Electronic equipment 412 typically comprises a variety of computer system readable media.These media can be it is any can be by
The usable medium that electronic equipment 412 accesses, including volatile and non-volatile media, moveable and immovable medium.
Storage device 428 may include the computer system readable media of form of volatile memory, such as arbitrary access
Memory (Random Access Memory, RAM) 430 and/or cache memory 432.Electronic equipment 412 can be into one
Step includes other removable/nonremovable, volatile/non-volatile computer system storage mediums.Only as an example, it stores
System 434 can be used for reading and writing immovable, non-volatile magnetic media (Fig. 4 do not show, commonly referred to as " hard disk drive ").
Although not shown in fig 4, the disc driver for reading and writing to removable non-volatile magnetic disk (such as " floppy disk ") can be provided,
And to removable anonvolatile optical disk, such as CD-ROM (Compact Disc Read-Only Memory, CD-ROM),
Digital video disk (Digital Video Disc-Read Only Memory, DVD-ROM) or other optical mediums) read-write light
Disk drive.In these cases, each driver can pass through one or more data media interfaces and 418 phase of bus
Even.Storage device 428 may include at least one program product, which has one group of (for example, at least one) program mould
Block, these program modules are configured to perform the function of various embodiments of the present invention.
Program/utility 440 with one group of (at least one) program module 442 can store in such as storage dress
It sets in 428, such program module 442 includes but is not limited to operating system, one or more application program, other program moulds
It may include the realization of network environment in block and program data, each of these examples or certain combination.Program module
442 usually execute function and/or method in embodiment described in the invention.
Electronic equipment 412 (such as keyboard, can also be directed toward terminal, display 424 with one or more external equipments 414
Deng) communication, can also be enabled a user to one or more terminal interact with the electronic equipment 412 communicate, and/or with make
Any terminal that the electronic equipment 412 can be communicated with one or more of the other computing terminal (such as network interface card, modem
Etc.) communication.This communication can be carried out by input/output (I/O) interface 422.Also, electronic equipment 412 can also lead to
Cross network adapter 420 and one or more network (such as local area network (Local Area Network, LAN), wide area network
(Wide Area Network, WAN) and/or public network, such as internet) communication.As shown in figure 4, network adapter 420
It is communicated by bus 418 with other modules of electronic equipment 412.It should be understood that although not shown in the drawings, can be set in conjunction with electronics
Standby 412 use other hardware and/or software module, including but not limited to: microcode, terminal driver, redundant processor, outside
Disk drive array, disk array (Redundant Arrays of Independent Disks, RAID) system, tape drive
Dynamic device and data backup storage system etc..
The program that processor 416 is stored in storage device 428 by operation, thereby executing various function application and number
According to processing, such as realize problem map detection method provided by any embodiment of the invention, this method may include:
Obtain target electronic map datum, wherein include at least one test object in the target electronic map datum；
By target electronic map datum input map detection model trained in advance, according to the map detection model
Output determine the target electronic map datum with the presence or absence of layout drawing mistake, wherein the map detection model is based on
The setting regions scale parameter training of each test object obtains, and the output of map detection model is included in that there is no layout drawings
To the mark of test object on the target electronic map datum of mistake.
Embodiment five
The embodiment of the present invention five additionally provides a kind of computer readable storage medium, is stored thereon with computer program, should
Realize that such as problem map detection method provided by any embodiment of the invention, this method can wrap when program is executed by processor
It includes:
Obtain target electronic map datum, wherein include at least one test object in the target electronic map datum；
By target electronic map datum input map detection model trained in advance, according to the map detection model
Output determine the target electronic map datum with the presence or absence of layout drawing mistake, wherein the map detection model is based on
The setting regions scale parameter training of each test object obtains, and the output of map detection model is included in that there is no layout drawings
To the mark of test object on the target electronic map datum of mistake.
The computer storage medium of the embodiment of the present invention, can be using any of one or more computer-readable media
Combination.Computer-readable medium can be computer-readable signal media or computer readable storage medium.It is computer-readable
Storage medium for example may be-but not limited to-the system of electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, device or
Device, or any above combination.The more specific example (non exhaustive list) of computer readable storage medium includes: tool
There are electrical connection, the portable computer diskette, hard disk, random access memory (RAM), read-only memory of one or more conducting wires
(ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-
ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.In this document, computer-readable storage
Medium can be any tangible medium for including or store program, which can be commanded execution system, device or device
Using or it is in connection.
Computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including but unlimited
In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can
Any computer-readable medium other than storage medium is read, which can send, propagates or transmit and be used for
By the use of instruction execution system, device or device or program in connection.
The program code for including on computer-readable medium can transmit with any suitable medium, including --- but it is unlimited
In wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
The computer for executing operation of the present invention can be write with one or more programming languages or combinations thereof
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
It further include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
It fully executes, partly execute on the user computer on the user computer, being executed as an independent software package, portion
Divide and partially executes or executed on remote computer or terminal completely on the remote computer on the user computer.It is relating to
And in the situation of remote computer, remote computer can pass through the network of any kind --- including local area network (LAN) or extensively
Domain net (WAN)-be connected to subscriber computer, or, it may be connected to outer computer (such as provided using Internet service
Quotient is connected by internet).
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (10)
1. a kind of problem map detection method characterized by comprising
Obtain target electronic map datum, wherein include at least one test object in the target electronic map datum；
By target electronic map datum input map detection model trained in advance, according to the defeated of the map detection model
Determine the target electronic map datum with the presence or absence of layout drawing mistake out, wherein the map detection model is based on each
The setting regions scale parameter training of test object obtains, and the output of the map detection model is included in that there is no layout drawings
To the mark of test object on the target electronic map datum of mistake.
2. the method according to claim 1, wherein the training process of the map detection model includes:
An at least simple electric map datum is obtained, as training set, wherein on every simple electric map datum, use
Setting regions scale parameter corresponding with each test object marks each test object；
The training set is inputted in default neural network structure and carries out model training, obtains the map detection model, wherein
It include that network is suggested in region in the default neural network structure.
3. according to the method described in claim 2, it is characterized in that, by the training set input in default neural network structure into
Row model training obtains the map detection model, comprising:
Using the feature extraction network of the default neural network structure, the feature of every simple electric map datum is extracted, is obtained
To the characteristic image of every simple electric map datum；
Network is suggested into the region that the characteristic image inputs the default neural network structure, based on setting for each test object
Determine regional percentage parameter, determines the suggestion areas of each test object on every simple electric map datum；
Every characteristic image for carrying suggestion areas is inputted in the default neural network structure, and is adjusted according to area differentiation
The training parameter of the default neural network structure, trained processing obtain the map detection model, wherein the area difference
On every simple electric map datum of different finger, between the suggestion areas position of each test object and its tab area position
Difference.
4. according to the method described in claim 2, it is characterized in that, the feature extraction network includes feature pyramid network；
Correspondingly, extracting every simple electric map datum using the feature extraction network of the default neural network structure
Feature obtains the characteristic image of every simple electric map datum, comprising:
Analysis On Multi-scale Features processing is carried out to every simple electric map datum using the feature pyramid network；
Based on Analysis On Multi-scale Features treated simple electric map datum, feature extraction is carried out, to obtain the characteristic image.
5. according to the method described in claim 2, it is characterized in that, the training set is inputted in default neural network structure
Carry out model training, before obtaining the map detection model, the training process of the map detection model further include:
Real time enhancing processing is carried out to every simple electric map datum in the training set.
6. according to the method described in claim 2, it is characterized in that, an acquisition at least simple electric map datum, makees
For training set, comprising:
An at least simple electric map datum is cleaned, using the simple electric map datum after cleaning as institute
State training set.
7. according to the method any in claim 2-6, which is characterized in that the training process of the map detection model is also
Include:
Obtain electronic map data verifying collection；
The electronics that electronic map data verifying is concentrated using at least one map detection model obtained in training process
Map datum is detected, and according to the current detection result of electronic map data verifying collection to current map detection model
Used default neural network structure is assessed；
The default neural network structure that assessed value is greater than preset threshold is determined as target nerve network structure, and will be based on described
The map detection model that the training of target nerve network structure obtains is determined as target map detection model, with using the target
Figure detection model detects the target electronic map datum.
8. a kind of problem map detection device characterized by comprising
Module is obtained, for obtaining target electronic map datum, wherein include at least one in the target electronic map datum
Test object；
Detection module, for the map detection model that target electronic map datum input is trained in advance, according to describedly
The output of figure detection model determines the target electronic map datum with the presence or absence of layout drawing mistake, wherein the map inspection
It surveys model to obtain based on the setting regions scale parameter training of each test object, the output of the map detection model is included in
There is no on the target electronic map datum of layout drawing mistake to the mark of test object.
9. a kind of electronic equipment characterized by comprising
One or more processors；
Storage device, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
The now problem map detection method as described in any in claim 1-7.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The problem map detection method as described in any in claim 1-7 is realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910257029.0A CN109977191B (en) | 2019-04-01 | 2019-04-01 | Problem map detection method, device, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910257029.0A CN109977191B (en) | 2019-04-01 | 2019-04-01 | Problem map detection method, device, electronic equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109977191A true CN109977191A (en) | 2019-07-05 |
| CN109977191B CN109977191B (en) | 2021-04-30 |
Family
ID=67082158
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910257029.0A Active CN109977191B (en) | 2019-04-01 | 2019-04-01 | Problem map detection method, device, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109977191B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334228A (en) * | 2019-07-09 | 2019-10-15 | 广西壮族自治区基础地理信息中心 | A kind of Internet Problems map screening method based on deep learning |
| CN112288696A (en) * | 2020-10-20 | 2021-01-29 | 北京百度网讯科技有限公司 | Auxiliary quality inspection method and device, electronic equipment and storage medium |
| CN112414430A (en) * | 2019-08-22 | 2021-02-26 | 北京初速度科技有限公司 | Electronic navigation map quality detection method and device |
| CN112642158A (en) * | 2020-12-31 | 2021-04-13 | 完美世界(北京)软件科技发展有限公司 | Game resource map auditing method and device, storage medium and computer equipment |
| CN112785567A (en) * | 2021-01-15 | 2021-05-11 | 北京百度网讯科技有限公司 | Map detection method and device, electronic equipment and storage medium |
| CN112966609A (en) * | 2021-03-05 | 2021-06-15 | 北京百度网讯科技有限公司 | Target detection method and device |
| CN113326245A (en) * | 2020-02-28 | 2021-08-31 | 北京京东乾石科技有限公司 | Method, device, equipment and computer readable medium for switching map service |
| CN116541574A (en) * | 2023-07-07 | 2023-08-04 | 湖北珞珈实验室 | Intelligent extraction method, device, storage medium and equipment for map sensitive information |
| CN117275030A (en) * | 2023-09-27 | 2023-12-22 | 自然资源部地图技术审查中心 | Method and device for auditing map |
| CN117372510A (en) * | 2023-12-05 | 2024-01-09 | 中交天津港湾工程研究院有限公司 | Map annotation identification method, terminal and medium based on computer vision model |
| CN117275030B (en) * | 2023-09-27 | 2024-05-14 | 自然资源部地图技术审查中心 | Method and device for auditing map |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103823828A (en) * | 2013-10-31 | 2014-05-28 | 小米科技有限责任公司 | Map scale setting method, map scale setting device and terminal |
| CN105513071A (en) * | 2015-12-08 | 2016-04-20 | 江苏省基础地理信息中心 | Topographic map schema quality evaluation method |
| CN105893256A (en) * | 2016-03-30 | 2016-08-24 | 西北工业大学 | Software failure positioning method based on machine learning algorithm |
| CN107436906A (en) * | 2016-05-27 | 2017-12-05 | 高德信息技术有限公司 | A kind of information detecting method and device |
| CN109145171A (en) * | 2018-07-23 | 2019-01-04 | 广州市城市规划勘测设计研究院 | A kind of multiple dimensioned map data updating method |
| CN109215487A (en) * | 2018-08-24 | 2019-01-15 | 宽凳（北京）科技有限公司 | A kind of high-precision cartography method based on deep learning |
| CN109443382A (en) * | 2018-10-22 | 2019-03-08 | 北京工业大学 | Vision SLAM closed loop detection method based on feature extraction Yu dimensionality reduction neural network |
| US20190095521A1 (en) * | 2017-09-22 | 2019-03-28 | Accenture Global Solutions Limited | Virtual artificial intelligence based consultant |
-
2019
- 2019-04-01 CN CN201910257029.0A patent/CN109977191B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103823828A (en) * | 2013-10-31 | 2014-05-28 | 小米科技有限责任公司 | Map scale setting method, map scale setting device and terminal |
| CN105513071A (en) * | 2015-12-08 | 2016-04-20 | 江苏省基础地理信息中心 | Topographic map schema quality evaluation method |
| CN105893256A (en) * | 2016-03-30 | 2016-08-24 | 西北工业大学 | Software failure positioning method based on machine learning algorithm |
| CN107436906A (en) * | 2016-05-27 | 2017-12-05 | 高德信息技术有限公司 | A kind of information detecting method and device |
| US20190095521A1 (en) * | 2017-09-22 | 2019-03-28 | Accenture Global Solutions Limited | Virtual artificial intelligence based consultant |
| CN109145171A (en) * | 2018-07-23 | 2019-01-04 | 广州市城市规划勘测设计研究院 | A kind of multiple dimensioned map data updating method |
| CN109215487A (en) * | 2018-08-24 | 2019-01-15 | 宽凳（北京）科技有限公司 | A kind of high-precision cartography method based on deep learning |
| CN109443382A (en) * | 2018-10-22 | 2019-03-08 | 北京工业大学 | Vision SLAM closed loop detection method based on feature extraction Yu dimensionality reduction neural network |
Non-Patent Citations (3)
| Title |
|---|
| 包贺先等: "基于改进贝叶斯判别分析的地图修改检测", 《全国测绘科技信息网中南分网第二十八次学术信息交流会论文集》 * |
| 周佳薇等: ""问题地图"现状分析及对策", 《测绘技术装备》 * |
| 黄龙: ""一种快速挖掘互联网问题地图图片的方法"", 《测绘与空间地理信息》 * |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334228A (en) * | 2019-07-09 | 2019-10-15 | 广西壮族自治区基础地理信息中心 | A kind of Internet Problems map screening method based on deep learning |
| CN112414430B (en) * | 2019-08-22 | 2022-12-02 | 北京初速度科技有限公司 | Electronic navigation map quality detection method and device |
| CN112414430A (en) * | 2019-08-22 | 2021-02-26 | 北京初速度科技有限公司 | Electronic navigation map quality detection method and device |
| CN113326245A (en) * | 2020-02-28 | 2021-08-31 | 北京京东乾石科技有限公司 | Method, device, equipment and computer readable medium for switching map service |
| CN113326245B (en) * | 2020-02-28 | 2024-01-12 | 北京京东乾石科技有限公司 | Method, apparatus, device and computer readable medium for switching map service |
| CN112288696A (en) * | 2020-10-20 | 2021-01-29 | 北京百度网讯科技有限公司 | Auxiliary quality inspection method and device, electronic equipment and storage medium |
| CN112288696B (en) * | 2020-10-20 | 2024-03-08 | 北京百度网讯科技有限公司 | Auxiliary quality inspection method and device, electronic equipment and storage medium |
| CN112642158A (en) * | 2020-12-31 | 2021-04-13 | 完美世界(北京)软件科技发展有限公司 | Game resource map auditing method and device, storage medium and computer equipment |
| CN112642158B (en) * | 2020-12-31 | 2021-09-28 | 完美世界(北京)软件科技发展有限公司 | Game resource map auditing method and device, storage medium and computer equipment |
| CN113730918A (en) * | 2020-12-31 | 2021-12-03 | 完美世界(北京)软件科技发展有限公司 | Game resource map auditing method and device, storage medium and computer equipment |
| CN113730918B (en) * | 2020-12-31 | 2023-07-14 | 完美世界(北京)软件科技发展有限公司 | Game resource map auditing method and device, storage medium and computer equipment |
| CN112785567A (en) * | 2021-01-15 | 2021-05-11 | 北京百度网讯科技有限公司 | Map detection method and device, electronic equipment and storage medium |
| CN112785567B (en) * | 2021-01-15 | 2023-09-22 | 北京百度网讯科技有限公司 | Map detection method, map detection device, electronic equipment and storage medium |
| CN112966609A (en) * | 2021-03-05 | 2021-06-15 | 北京百度网讯科技有限公司 | Target detection method and device |
| CN112966609B (en) * | 2021-03-05 | 2023-08-11 | 北京百度网讯科技有限公司 | Target detection method and device |
| CN116541574B (en) * | 2023-07-07 | 2023-10-03 | 湖北珞珈实验室 | Intelligent extraction method, device, storage medium and equipment for map sensitive information |
| CN116541574A (en) * | 2023-07-07 | 2023-08-04 | 湖北珞珈实验室 | Intelligent extraction method, device, storage medium and equipment for map sensitive information |
| CN117275030A (en) * | 2023-09-27 | 2023-12-22 | 自然资源部地图技术审查中心 | Method and device for auditing map |
| CN117275030B (en) * | 2023-09-27 | 2024-05-14 | 自然资源部地图技术审查中心 | Method and device for auditing map |
| CN117372510A (en) * | 2023-12-05 | 2024-01-09 | 中交天津港湾工程研究院有限公司 | Map annotation identification method, terminal and medium based on computer vision model |
| CN117372510B (en) * | 2023-12-05 | 2024-03-01 | 中交天津港湾工程研究院有限公司 | Map annotation identification method, terminal and medium based on computer vision model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109977191B (en) | 2021-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109977191A (en) | Problem map detection method, device, electronic equipment and medium | |
| CN103528617B (en) | A kind of cockpit instrument identifies and detection method and device automatically | |
| CN104268603B (en) | Intelligent marking method and system for text objective questions | |
| CN105574550A (en) | Vehicle identification method and device | |
| CN104123534B (en) | Bio-identification | |
| CN108399386A (en) | Information extracting method in pie chart and device | |
| CN108920580A (en) | Image matching method, device, storage medium and terminal | |
| CN108305243A (en) | A kind of magnetic tile surface defect detection method based on deep learning | |
| CN107808358A (en) | Image watermark automatic testing method | |
| CN109376631A (en) | A kind of winding detection method and device neural network based | |
| CN105654066A (en) | Vehicle identification method and device | |
| CN110490238A (en) | A kind of image processing method, device and storage medium | |
| CN110362777A (en) | A kind of picture examination method and system of multi-information fusion | |
| CN103164692A (en) | Special vehicle instrument automatic identification system and algorithm based on computer vision | |
| CN109978872B (en) | White matter microstructure characteristic screening system and method based on white matter fiber tracts | |
| CN114419736A (en) | Experiment scoring method, system, equipment and readable storage medium | |
| CN110175519B (en) | Method and device for identifying separation and combination identification instrument of transformer substation and storage medium | |
| CN109766818A (en) | Pupil center's localization method and system, computer equipment and readable storage medium storing program for executing | |
| CN113409284B (en) | Circuit board fault detection method, device, equipment and storage medium | |
| CN108628993A (en) | electronic map adaptive classification method, device, equipment and storage medium | |
| CN112381099A (en) | Question recording system based on digital education resources | |
| CN109977832A (en) | A kind of image processing method, device and storage medium | |
| CN113807278A (en) | Deep learning-based land use classification and change prediction method | |
| CN110705619A (en) | Fog concentration grade judging method and device | |
| CN107533636A (en) | Pre-matching prediction for pattern test |
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 |