CN109977191B - Problem map detection method, device, electronic equipment and medium - Google Patents
Problem map detection method, device, electronic equipment and medium Download PDFInfo
- Publication number
- CN109977191B CN109977191B CN201910257029.0A CN201910257029A CN109977191B CN 109977191 B CN109977191 B CN 109977191B CN 201910257029 A CN201910257029 A CN 201910257029A CN 109977191 B CN109977191 B CN 109977191B
- Authority
- CN
- China
- Prior art keywords
- detection
- map data
- electronic map
- map
- 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.)
- Active
Links
Images
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 problem map detection method, a problem map detection device, electronic equipment and a problem map detection medium, wherein the method comprises the following steps: acquiring target electronic map data, wherein the target electronic map data comprises at least one detection object; inputting target electronic map data into a pre-trained map detection model, and determining whether the target electronic map data has layout drawing errors according to the output of the map detection model, wherein the map detection model is obtained by training based on the set region proportion parameters of each detection object, and the output of the map detection model comprises the marking of the detection object on the target electronic map data without the layout drawing errors. The embodiment of the invention introduces the deep learning technology into the problem map automatic detection for the first time, solves the problems of low efficiency and high labor cost of the existing map detection method relying on manual detection, realizes the automatic detection of the electronic map and the automatic identification of the problem map, and improves the detection efficiency.
Description
Technical Field
The embodiment of the invention relates to the technical field of surveying and mapping geographic information, in particular to a problem map detection method, a problem map detection device, electronic equipment and a medium.
Background
In ancient times of China, "layout" is a book for registering household and land, "map" is a map, and "layout" represents household and map and gradually evolves into a pronoun of national territory. With the development of times, maps are closely related to the daily life of people, and the development of the economy and the society, even the national security, is directly influenced. However, in recent years, map use safety problems are frequent, and if a 'problem map' comes, the map not only has adverse effects on social public and damages the safety and benefits of national ownership, but also becomes a handle for adversary attacks in the open world, even causes international disputes, and therefore high attention must be paid to the map use safety problems.
As is known, the map market in China is rapidly expanded and prosperous. Since 2012, published maps were about 2000 more than 3 hundred million copies each year. In 2016, the navigation electronic map can complete the service total value of 66 hundred million yuan, and the mapping qualification unit with the Internet map service qualification can complete the service total value of 282 hundred million yuan. However, in the meantime, the map market and the display and use of the map are still prohibited from damaging the 'problem maps' such as the territorial ownership, the safety, the ocean rights and the like in China. According to statistics, law enforcement actions are carried out in the whole country for 11000 times since 2012, about 1000 relevant cases are checked and dealt, more than 20 illegal map products are collected and paid, and the disposal of more than 1000 websites with the problem map is completed. In the 2016 map market inspection, a large number of illegal cases are found and legally examined in various map markets, cultural goods markets, internet map service units, exhibitions (exhibitions), memorial halls, museums and the like. Some cases, such as missing pictures and wrong pictures of important islands and national boundaries in China endanger the national mastership; some of the sensitive and confidential information are uploaded and marked in the internet, and public login, illegal transaction and confidential maps and the like are disclosed, so that the national security is endangered; some related countries and regions are not marked according to the external political exchange claims of China, and the national benefits are damaged by illegal mapping, drawing, map service provision and the like. Especially, some problem maps published by the internet and published by media have wide distribution, fast transmission and great harm.
At present, the method for checking the 'problem map' mainly relies on manual visual interpretation. Compared with a correct map, the problem map has the defects that the difference is very small, a detector needs to be skilled in a drawing method of the correct map, the detection speed is low, and the labor intensity is high.
Disclosure of Invention
The embodiment of the invention provides a problem map detection method, a problem map detection device, electronic equipment and a medium, which are used for improving the detection efficiency of a problem map and ensuring the detection accuracy.
In a first aspect, an embodiment of the present invention provides a problem map detection method, where the method includes:
acquiring target electronic map data, wherein the target electronic map data comprises at least one detection object;
inputting the target electronic map data into a pre-trained map detection model, and determining whether the target electronic map data has layout drawing errors according to the output of the map detection model, wherein the map detection model is obtained by training based on the set region proportion parameters of each detection object, and the output of the map detection model comprises the marking of the detection object on the target electronic map data without the layout drawing errors.
In a second aspect, an embodiment of the present invention further provides a problem map detection apparatus, where the apparatus includes:
the system comprises an acquisition module, a detection module and a processing module, wherein the acquisition module is used for acquiring target electronic map data, and the target electronic map data comprises at least one detection object;
and the detection module is used for inputting the target electronic map data into a pre-trained map detection model and determining whether the target electronic map data has layout drawing errors according to the output of the map detection model, wherein the map detection model is obtained by training based on the set area proportion parameters of each detection object, and the output of the map detection model comprises the marking of the detection object on the target electronic map data without the layout drawing errors.
In a third aspect, an embodiment of the present invention further provides an electronic device, including:
one or more processors;
a storage device for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement a problem map detection method as in any embodiment of the invention.
In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the problem map detection method according to any embodiment of the present invention.
According to the embodiment of the invention, the acquired target electronic map data is input into the pre-trained map detection model, and whether the target electronic map data has a map drawing error is determined according to the output of the map detection model, wherein the map detection model is obtained by training based on the set area proportion parameter of each detection object, the output of the map detection model comprises the marking of the detection object on the target electronic map data without the map drawing error, namely, the depth learning technology is introduced into the problem map automatic detection for the first time, so that the problems of low efficiency and high labor cost of the existing map detection method relying on manual detection are solved, the automatic detection of the electronic map and the automatic identification of the problem map are realized, the detection efficiency is improved, and the detection accuracy is ensured.
Drawings
Fig. 1 is a flowchart of a problem map detection method according to an embodiment of the present invention;
fig. 2 is a flowchart of a problem map detection method according to a second embodiment of the present invention;
fig. 3 is a schematic structural diagram of a problem map detection apparatus according to a third embodiment of the present invention;
fig. 4 is a schematic structural diagram of an electronic device according to a fourth 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 invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
Example one
Fig. 1 is a flowchart of a problem map detection method according to an embodiment of the present invention, where the embodiment is applicable to a situation where a problem map in an electronic map is detected, and the method may be executed by a problem map detection apparatus, where the apparatus may be implemented in a software and/or hardware manner, and may be integrated on an electronic device, where the electronic device includes a server.
As shown in fig. 1, the problem map detection method provided in this embodiment may include:
s110, obtaining target electronic map data, wherein the target electronic map data comprise at least one detection object.
The target electronic map data refers to any electronic map data to be detected, and comprises any types such as a world map, a national map and a city map. The detection object refers to a specific area in the target electronic map data and is determined according to the current detection requirement. For example, if the current detection requirement is to detect whether the region information corresponding to the designated province in the country map is correct, the detection object is a sub-region of the designated province in the country map.
And S120, inputting target electronic map data into a pre-trained map detection model, and determining whether the target electronic map data has layout drawing errors according to the output of the map detection model, wherein the map detection model is obtained by training based on the set region proportion parameter of each detection object, and the output of the map detection model comprises the marking of the detection object on the target electronic map data without the layout drawing errors.
Considering that the electronic map data is different from the general image data, after the geographic area is determined, the outline shape of each sub-area included in the electronic map data is usually determined and does not change with the change of factors such as the format, size and resolution of the electronic map data. Therefore, in the training process of the map detection model, the area proportion parameters corresponding to each detection object one to one are set, and the frame area determined by the area proportion parameters just can enclose the detection object, namely the area of the interference area in the frame area is the minimum except the area occupied by the detection object. In the detection process of the electronic map, the map outline is an important factor for determining whether the map is correct, and therefore the matching degree of the frame area determined by the area proportion parameter and the detection object directly influences the accuracy of map detection. The smaller the interference area included in the frame area is, the clearer the area outline of the detection target is, and the higher the detection accuracy for the detection target is.
The process for determining the set area proportion parameter of each detection object comprises the following steps: determining a target scale of the electronic map data, wherein the target scale can be the scale with the highest use frequency when the electronic map is currently drawn or displayed; determining the proportion between the geometric side lengths corresponding to the detection objects in the electronic map data represented by the target scale; the ratio between the geometric side lengths is determined as a set area ratio parameter of the detection object, for example, the aspect ratio of the detection object is taken as its set area ratio parameter.
In this embodiment, the map detection model is obtained by training using a deep learning method. When the model is trained according to the constructed neural network structure, besides setting common network model hyper-parameters such as learning rate, iteration times, initialization method, model training structure, threshold value of training process, gradient cutting strategy and the like, the set region proportion parameter corresponding to the detection object is set as the training parameter for each detection object, and simultaneously participates in model training. Due to the uniqueness of the electronic map data, the outline shapes of different detection objects are different, and therefore the set area proportion parameters of the detection objects are different. The set area proportion parameter corresponding to each detection object is set as a training parameter, so that the effect of accurately selecting different detection objects can be realized, and the method is an important basis for ensuring the accuracy and pertinence of model training. When the model is trained to be used for detecting the electronic map, the set area proportion parameter can also be used as a screening condition to screen the area on the target electronic map data, so that the detection object can be quickly positioned.
In the model training process, for the national map training data, the detection objects are sub-regions A, B, C and D, respectively, and according to the aspect ratios of the four sub-regions on the map under the target scale, the set region scale parameters of the four sub-regions are set to be 1: 3.61, 3.62:1, 2.80:1, 1:1.28, and 1: 1.53.
The map detection model is used for detecting the target electronic map data, whether the map drawing error exists in the target electronic map data can be determined through the difference of the output results of the model, and the intelligent detection of the electronic map data is realized. For example, if there is no layout drawing error in the target electronic map data, the model output result may be: carrying out visual frame marking on the detected object on the target electronic map data; if the target electronic map data has layout drawing errors, that is, the target electronic map is a problem map, the output result of the model may be: the target electronic map data is directly output in early warning modes such as text prompt or voice prompt, and the data belongs to the problem map, and frame marking is not carried out on the detection object, namely, the problem map is automatically identified through the difference of the output results of the model. When a plurality of detection objects exist on the target electronic map, if one detection object has drawing errors, the target electronic map data belongs to the problem map, and when the plurality of detection objects are drawn correctly, the target electronic map data belongs to the correct map. It should be noted that, the detection objects in the electronic map data are usually a fixed limited number of detection areas, the correct drawing situations of the detection objects are unique, and the incorrect drawing situations of the detection objects have multiple possibilities, so that when the map detection model is trained, only the detection objects drawn correctly are learned and identified, and when the model is applied to problem map detection, only the detection objects in the electronic map data without drawing errors are subjected to frame labeling in the model output result, and based on an exclusion method, the electronic map data without the detection objects subjected to frame labeling output belongs to a problem map. Of course, the position of the error area on the problem map can also be adaptively shown, so as to facilitate the subsequent correction of the problem map. In addition, the detection objects included in the target electronic map data are all detection objects participating in training in the training process of the map detection model.
According to the technical scheme, the acquired target electronic map data is input into a pre-trained map detection model, whether layout drawing errors exist in the target electronic map data is determined according to the output of the map detection model, wherein the map detection model is obtained through training based on the set area proportion parameters of each detection object, the output of the map detection model comprises the marking of the detection object on the target electronic map data without the layout drawing errors, the problems that an existing map detection method relying on manual detection is low in efficiency and high in labor cost are solved, automatic detection of an electronic map and automatic identification of a problem map are achieved, detection efficiency is improved, and detection accuracy is guaranteed.
Example two
Fig. 2 is a flowchart of a problem map detection method according to a second embodiment of the present invention, which is further optimized based on the above-mentioned embodiments. As shown in fig. 2, the method may include:
s210, obtaining at least one piece of sample electronic map data as a training set, wherein each detection object is marked on each piece of sample electronic map data by using a set area proportion parameter corresponding to each detection object.
For example, enough electronic map data can be searched from the internet as a training set, and each piece of electronic map data is subjected to frame marking to determine the detection object of the model training process. In order to ensure that the detection object is accurately positioned in the training set, the frame marking is realized in a manual marking mode. Illustratively, according to a set region proportion parameter predetermined for each detection object by a user, a corresponding rectangular frame is created by taking the region center of the detection object as the center of the frame, so as to adaptively surround the detection object.
Optionally, the obtaining at least one piece of sample electronic map data as a training set includes:
and cleaning at least one piece of sample electronic map data, and taking the cleaned sample electronic map data as a training set. The data cleaning refers to removing drawing error data in the sample electronic map data. And the correctly drawn electronic map data is used as a training set, so that the accuracy of the obtained training model is ensured.
Optionally, before inputting the training set into the preset neural network structure for model training to obtain the map detection model, the training process of the map detection model further includes: and performing real-time enhancement processing on each sample electronic map data in the training set to obtain the training set for model training based on the enhancement processing. Among these, real-time enhancement processing includes but is not limited to: data decentralization, data standardization, data ZCA (Zero-phase Component Analysis) whitening, data graying, data random rotation, data horizontal offset, data vertical offset, data size random scaling, data random channel offset, data random horizontal flip, data random vertical flip, data binarization and data random clipping. By carrying out real-time enhancement processing on the sample electronic map data, the real-time transformation and updating of the sample electronic map data can be realized, and the diversity of sample data is increased; and the transformed sample electronic map data obtained by the enhancement processing is stored in a cache mode, and after the sample data is used for model training, the cache is randomly released, so that the storage space of the electronic equipment is not excessively occupied, and the storage pressure is relieved on the basis of realizing enrichment of the sample data.
S220, inputting the training set into a preset neural network structure to perform model training to obtain a map detection model, wherein the preset neural network structure comprises a regional suggestion network.
The preset neural Network structures in this embodiment may include fast-RCNN and Mask-RCNN Network structures, and these neural Network structures include Region suggestion networks (RPNs), and ROI (Region Of Interest) is generated by using RPN networks. In the model training process, a user presets a set area proportion parameter of each detection object as a training parameter, the area suggestion network determines the corresponding detection object in a frame mode by using the set area proportion parameter, then map recognition is continuously carried out on the basis of a preset neural network structure, and finally a map detection model is obtained through training.
For example, when the RPN network is trained based on the fast-RCNN network structure, classification and regression calculations are performed simultaneously using a multi-task loss function (multi-task loss) for each piece of sample electronic map data, and the classification probability and the frame position are jointly trained. The multitask Loss function includes a classification Loss function (Softmax Loss) and a bounding box regression Loss function (Smooth L1Loss), which are defined as follows:
wherein: n is a radical ofclsRepresents the number of input samples in the mini-batch, NregRepresents the number of anchor (anchor) positions, i is the index of anchor in the mini-batch; p is a radical ofiIs the probability that the anchor is predicted to be a target;
is the probability of marking a positive sample in the sample, an
Is a vector representing the location where the RPN predicted the anchor;
is the true position; the positive sample refers to the detection object, and the negative sample refers to the image background.
considering the error condition of the problem map, there are countless possibilities, such as color error, contour line error, area missing, etc., therefore, in the model training process, the area suggestion network is used to mark only positive samples in the original image of the electronic map data, rather than negative samples. For example, the threshold T is set to 0.7, the target with the probability P greater than T is a positive sample, the target with the probability P less than or equal to T is a negative sample, and only the target with P greater than T is marked in the original image of the electronic map data.
The loss function corresponding to the map detection model obtained through final training comprises the loss of generating the ROI and the loss of classifying the ROI into specific categories, namely determining the loss corresponding to the position of the area where the detection object is located, and classifying the detection object into the loss corresponding to the correct map area or the wrong map area:
Loss＝Lrpn+Lspe
wherein:
Lrpn＝Lrpn_class+Lrpn_bbox,Lrpn_classthe classification loss of the area recommendation network is the loss of dividing the proposed detection frame into foreground/background by the network; l isrpn_bboxIs the loss of the border position of the area proposal network; l isspe＝Lspe_class+Lspe_bbox，Lspe_classIs a classification loss that classifies the proposed region of interest into a specific category; l isspe_bboxIs the calculated frame position loss by comparing the detection frame proposed by the area suggestion network with the real frame (i.e. the frame of the detection object marked in the training set).
And S230, acquiring target electronic map data, wherein the target electronic map data comprises at least one detection object.
In this embodiment, the detection objects included in the target electronic map data all belong to the detection objects used for model training in the model training process.
S240, inputting the target electronic map data into a pre-trained map detection model, and determining whether the map drawing error exists in the target electronic map data according to the output of the map detection model.
On the basis of the above technical solution, optionally, inputting the training set into a preset neural network structure for model training to obtain a map detection model, including:
extracting the characteristics of each sample electronic map data by using a characteristic extraction network of a preset neural network structure to obtain a characteristic image of each sample electronic map data;
inputting the characteristic image into a regional suggestion network with a preset neural network structure, and determining a suggestion region of each detection object on each sample electronic map data based on a set regional proportion parameter of each detection object;
inputting each feature image carrying the suggestion area into a preset neural network structure, adjusting training parameters of the preset neural network structure according to the area difference, and obtaining a map detection model through training processing, wherein the area difference refers to the difference between the position of the suggestion area of each detection object and the position of a labeling area of each detection object on each sample electronic map data.
The preset neural network structure is exemplified as follows as the fast-RCNN network structure:
inputting sample electronic map data into a fast-RCNN network structure, firstly extracting the characteristics of the sample electronic map data by using a shared convolution layer in a fast-RCNN characteristic extraction network to obtain a characteristic image of the sample electronic map data; inputting the feature images obtained by sharing the convolutional layers into a region suggestion network of fast-RCNN, and setting 9 rectangular windows (3 length-width ratios and 3 scales) on an original image corresponding to each pixel point of a sliding window, wherein one rectangular window is an anchor point; in the area suggestion network, continuously inputting the feature images into the convolution layer for convolution calculation, then performing regression and classification calculation according to convolution results and anchor points, and outputting suggestion areas of the detection objects; inputting the feature image carrying the suggested region into the ROI pooling layer, generating a feature map with a fixed size again, inputting the feature map with the fixed size into the full-connection layer to determine whether the detection object has layout drawing errors or not, and positioning the wrong detection object. After the area suggestion network outputs the suggestion area of the detection object, the frame position of the detection object, which is manually marked in the sample electronic map data, is compared with the position of the suggestion area, and the training parameters of the preset neural network structure are adjusted according to the difference between the frame position and the position of the suggestion area, for example, the gradient clipping strategy parameters are adjusted according to the difference, so that the accuracy of the map detection model obtained by training is ensured.
Further, the feature extraction network of the preset neural network structure comprises a feature pyramid network; correspondingly, the method for extracting the characteristics of each sample electronic map data by using the characteristic extraction network of the preset neural network structure to obtain the characteristic image of each sample electronic map data comprises the following steps:
carrying out multi-scale feature processing on each piece of sample electronic map data by using a feature pyramid network;
and performing feature extraction based on the sample electronic map data after the multi-scale feature processing to obtain a feature image. In order to solve the problem of low accuracy of electronic map data identification under multiple scales, a Feature Pyramid Network (FPN) is introduced to enhance the expression capability of a map detection model obtained by training on multiple scales, so that the effects of integrating map features under multiple scales and neglecting the influence of a variable scale effect on electronic map detection are achieved, the electronic map data of multiple scales are accurately detected, and the self-adaptive detection of the map data of multiple scales is realized. For example, the feature pyramid network can be used to automatically down-sample electronic map data with ultrahigh resolution (e.g., 30000 × 30000), thereby solving the problem that the conventional image processing tool cannot directly process electronic map data with ultrahigh resolution.
On the basis of the above technical solution, optionally, the training process of the map detection model further includes:
acquiring an electronic map data verification set;
detecting electronic map data in the electronic map data verification set by using at least one map detection model obtained in the training process, and evaluating a preset neural network structure adopted by the current map detection model according to the current detection result of the electronic map data verification set;
and determining a preset neural network structure with the evaluation value larger than a preset threshold value as a target neural network structure, determining a map detection model obtained based on the training of the target neural network structure as a target map detection model, and detecting target electronic map data by using the target map detection model.
Illustratively, in the data preparation stage, a large amount of electronic map data is collected from the network, and a training set and a verification set are constructed according to a certain proportion. The training set is used for model training, and the verification set is used for verifying the trained model in the model training process.
The base network is a part of a preset neural network structure, different base networks are selected, and the performances of the preset neural network structure are different, so that the performances of the map detection models obtained based on the preset neural network structure training are different. In the model training process, an optimal base network can be determined firstly, and then an optimal preset neural network structure is determined based on the optimal base network.
For example, first, network hyper-parameters are set, and different base networks are screened based on the same preset neural network structure, and the selectable base networks include but are not limited to: ResNet50, ResNet101, VGG19, ResNext101-64x4d, ResNext101-32x8d, ResNext152-32x8d-IN5k, and the like. Specifically, in the model training process, a monitor is used to evaluate the models obtained by training based on different base networks, and the model calculation speed and the detection accuracy are balanced to determine the optimal base network under the same hyper-parameter condition, for example, the optimal base network may be ResNet 101. And then, evaluating different preset neural network structures comprising the optimal base network by using the model detection result of the verification set. The selectable preset neural network structures comprise fast-RCNN, Mask-RCNN and the like. Specifically, in the whole period of the model training, the monitor can be used to perform all-around monitoring and evaluation on the map detection model obtained based on different preset neural network structure training, and the optimal model and the preset neural network structure corresponding to the ideal model in the whole training period are stored, wherein the preset neural network structure is the target neural network structure. In the estimation process of the preset neural network structure or the training model, the monitoring object may be used as an estimation factor, for example, the smaller the value corresponding to the monitoring object is, the higher the estimation result of the training model and the corresponding neural network structure is. Monitoring the subject may include: training loss, area candidate network training loss, verification loss, area candidate network verification loss, and the like. The higher the evaluation value of the neural network structure, the higher the detection accuracy of the map detection model.
In addition, the data preparation phase can also create a test set at the same time for testing the trained map detection model. The training set, the validation set and the test set maintain a certain mathematical distribution, for example, the ratio of the three is 8:1: 1. And adjusting different confidence coefficient thresholds according to the model detection result of the test set, so that the model detection precision meets the requirement.
According to the technical scheme of the embodiment, a training set is obtained firstly, each detection object is marked on each piece of sample electronic map data in the training set by using a set area proportion parameter corresponding to each detection object, then the training set is input into a preset neural network structure for model training to obtain a map detection model, finally the obtained target electronic map data is input into the map detection model, and whether map drawing errors exist in the target electronic map data is determined according to the output of the map detection model, so that the problems of low efficiency and high labor cost of the existing map detection method relying on manual detection are solved, automatic detection of an electronic map and automatic identification of a problem map are realized, the detection efficiency is improved, and the detection accuracy is ensured; moreover, based on the regional suggestion network in the preset neural network structure, the effect of accurately positioning the detection object according to the set regional proportion parameter of the detection object is realized; in addition, by introducing the characteristic pyramid network into the preset neural network structure, the problem of low accuracy of problem map data identification under multiple scales is solved, and self-adaptive detection of the problem map data under multiple scales is realized.
EXAMPLE III
Fig. 3 is a schematic structural diagram of a problem map detection apparatus according to a third embodiment of the present invention, which is applicable to a situation of detecting a problem map in an electronic map. The apparatus can be implemented in software and/or hardware, and can be integrated on an electronic device, which includes a server.
As shown in fig. 3, the problem map detection apparatus provided in this embodiment may include an obtaining module 310 and a detecting module 320, where:
the acquiring module 310 is configured to acquire target electronic map data, where the target electronic map data includes at least one detection object;
the detection module 320 is configured to input target electronic map data into a pre-trained map detection model, and determine whether the target electronic map data has a layout drawing error according to an output of the map detection model, where the map detection model is obtained by training based on a set area proportion parameter of each detection object, and the output of the map detection model includes a label of the detection object on the target electronic map data without the layout drawing error.
Optionally, the apparatus further includes a map detection model training module, where the map detection model training module includes:
the training set acquisition unit is used for acquiring at least one piece of sample electronic map data as a training set, wherein each detection object is marked by using a set area proportion parameter corresponding to each detection object on each piece of sample electronic map data;
and the model training unit is used for inputting the training set into a preset neural network structure to carry out model training to obtain a map detection model, wherein the preset neural network structure comprises a regional suggestion network.
Optionally, the model training unit includes:
the characteristic image determining subunit is used for extracting the characteristics of each piece of sample electronic map data by utilizing a characteristic extraction network of a preset neural network structure to obtain a characteristic image of each piece of sample electronic map data;
the suggested region determining subunit is used for inputting the characteristic image into a region suggested network with a preset neural network structure, and determining the suggested region of each detection object on each piece of sample electronic map data based on the set region proportion parameter of each detection object;
and the model training subunit is used for inputting each feature image carrying the suggestion area into a preset neural network structure, adjusting training parameters of the preset neural network structure according to the area difference, and obtaining the map detection model through training processing, wherein the area difference refers to the difference between the position of the suggestion area of each detection object and the position of the labeled area on each sample electronic map data.
Optionally, the feature extraction network includes a feature pyramid network;
accordingly, the feature image determination subunit is configured to:
carrying out multi-scale feature processing on each piece of sample electronic map data by using a feature pyramid network;
and performing feature extraction based on the sample electronic map data after the multi-scale feature processing to obtain a feature image.
Optionally, the map detection model training module further includes:
and the data real-time enhancement processing unit is used for carrying out real-time enhancement processing on each sample electronic map data in the training set before the model training unit carries out the operation of inputting the training set into the preset neural network structure for model training to obtain the map detection model.
Optionally, the training set obtaining unit is configured to:
and cleaning at least one piece of sample electronic map data, and taking the cleaned sample electronic map data as a training set.
Optionally, the map detection model training module further includes:
the verification set acquisition unit is used for acquiring a verification set of the electronic map data;
the neural network structure evaluation unit is used for detecting the electronic map data in the electronic map data verification set by using at least one map detection model obtained in the training process and evaluating a preset neural network structure adopted by the current map detection model according to the current detection result of the electronic map data verification set;
and the target neural network structure determining unit is used for determining a preset neural network structure with the evaluation value larger than a preset threshold value as a target neural network structure, determining a map detection model obtained based on the training of the target neural network structure as a target map detection model, and detecting target electronic map data by using the target map detection model.
The problem map detection device provided by the embodiment of the invention can execute the problem map detection method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method. Reference may be made to the description in the method embodiments of the invention for details not explicitly described in this embodiment.
Example four
Fig. 4 is a schematic structural diagram of an electronic device according to a fourth embodiment of the present invention. FIG. 4 illustrates a block diagram of an exemplary electronic device 412 suitable for use in implementing embodiments of the present invention. The electronic device 412 shown in fig. 4 is only an example and should not bring any limitations to the functionality and scope of use of the embodiments of the present invention.
As shown in fig. 4, the electronic device 412 is in the form of a general purpose electronic device. The components of the electronic device 412 may include, but are not limited to: one or more processors 416, a storage device 428, and a bus 418 that couples the various system components including the storage device 428 and the processors 416.
A program/utility 440 having a set (at least one) of program modules 442 may be stored, for instance, in storage 428, such program modules 442 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. The program modules 442 generally perform the functions and/or methodologies of the described embodiments of the invention.
The electronic device 412 may also communicate with one or more external devices 414 (e.g., keyboard, pointing terminal, display 424, etc.), with one or more terminals that enable a user to interact with the electronic device 412, and/or with any terminals (e.g., network card, modem, etc.) that enable the electronic device 412 to communicate with one or more other computing terminals. Such communication may occur via input/output (I/O) interfaces 422. Also, the electronic device 412 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public Network, such as the internet) via the Network adapter 420. As shown in FIG. 4, network adapter 420 communicates with the other modules of electronic device 412 over bus 418. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the electronic device 412, including but not limited to: microcode, end drives, Redundant processors, external disk drive Arrays, RAID (Redundant Arrays of Independent Disks) systems, tape drives, and data backup storage systems, among others.
The processor 416 executes various functional applications and data processing by executing programs stored in the storage device 428, for example, implementing a problem map detection method provided by any embodiment of the present invention, which may include:
acquiring target electronic map data, wherein the target electronic map data comprises at least one detection object;
inputting the target electronic map data into a pre-trained map detection model, and determining whether the target electronic map data has layout drawing errors according to the output of the map detection model, wherein the map detection model is obtained by training based on the set region proportion parameters of each detection object, and the output of the map detection model comprises the marking of the detection object on the target electronic map data without the layout drawing errors.
EXAMPLE five
An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements a problem map detection method according to any embodiment of the present invention, where the method may include:
acquiring target electronic map data, wherein the target electronic map data comprises at least one detection object;
inputting the target electronic map data into a pre-trained map detection model, and determining whether the target electronic map data has layout drawing errors according to the output of the map detection model, wherein the map detection model is obtained by training based on the set region proportion parameters of each detection object, and the output of the map detection model comprises the marking of the detection object on the target electronic map data without the layout drawing errors.
Computer storage media for embodiments of the invention may employ any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or terminal. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (10)
1. A problem map detection method is characterized by comprising the following steps:
acquiring target electronic map data, wherein the target electronic map data comprises at least one detection object;
inputting the target electronic map data into a pre-trained map detection model, and determining whether the target electronic map data has layout drawing errors according to the output of the map detection model, wherein the map detection model is obtained by training based on the set region proportion parameter of each detection object, and the output of the map detection model comprises the marking of the detection object on the target electronic map data without the layout drawing errors;
the determination process of the set region proportion parameter comprises the following steps: determining a target scale of electronic map data, wherein the target scale is the scale with the highest use frequency when the electronic map is currently drawn and/or displayed; determining the proportion between the geometric side lengths corresponding to the detection objects in the electronic map data represented by the target scale; and determining the ratio between the geometric side lengths as a set region ratio parameter of the detection object.
2. The method of claim 1, wherein the training process of the map detection model comprises:
obtaining at least one piece of sample electronic map data as a training set, wherein each detection object is marked by using a set area proportion parameter corresponding to each detection object on each piece of sample electronic map data;
and inputting the training set into a preset neural network structure for model training to obtain the map detection model, wherein the preset neural network structure comprises a regional suggestion network.
3. The method of claim 2, wherein inputting the training set into a preset neural network structure for model training to obtain the map detection model comprises:
extracting the characteristics of each sample electronic map data by using the characteristic extraction network of the preset neural network structure to obtain a characteristic image of each sample electronic map data;
inputting the characteristic image into a regional suggestion network of the preset neural network structure, and determining a suggestion region of each detection object on each piece of sample electronic map data based on a set regional proportion parameter of each detection object;
inputting each feature image carrying a suggestion area into the preset neural network structure, adjusting training parameters of the preset neural network structure according to area difference, and obtaining the map detection model through training processing, wherein the area difference refers to the difference between the position of the suggestion area of each detection object and the position of a labeling area of each detection object on each sample electronic map data.
4. The method of claim 3, wherein the feature extraction network comprises a feature pyramid network;
correspondingly, the method for extracting the characteristics of each sample electronic map data by using the characteristic extraction network of the preset neural network structure to obtain the characteristic image of each sample electronic map data comprises the following steps:
carrying out multi-scale feature processing on each piece of sample electronic map data by using the feature pyramid network;
and performing feature extraction based on the sample electronic map data after the multi-scale feature processing to obtain the feature image.
5. The method of claim 2, wherein before inputting the training set into a preset neural network structure for model training to obtain the map detection model, the training process of the map detection model further comprises:
and performing real-time enhancement processing on each piece of sample electronic map data in the training set.
6. The method of claim 2, wherein the obtaining at least one sample electronic map data as a training set comprises:
and cleaning the at least one piece of sample electronic map data, and taking the cleaned sample electronic map data as the training set.
7. The method according to any one of claims 2-6, wherein the training process of the map detection model further comprises:
acquiring an electronic map data verification set;
detecting electronic map data in the electronic map data verification set by using at least one map detection model obtained in a training process, and evaluating a preset neural network structure adopted by a current map detection model according to a current detection result of the electronic map data verification set;
and determining a preset neural network structure with the evaluation value larger than a preset threshold value as a target neural network structure, determining a map detection model obtained based on the training of the target neural network structure as a target map detection model, and detecting the target electronic map data by using the target map detection model.
8. A problem map detection apparatus, comprising:
the system comprises an acquisition module, a detection module and a processing module, wherein the acquisition module is used for acquiring target electronic map data, and the target electronic map data comprises at least one detection object;
the detection module is used for inputting the target electronic map data into a pre-trained map detection model and determining whether the target electronic map data has layout drawing errors according to the output of the map detection model, wherein the map detection model is obtained by training based on the set area proportion parameter of each detection object, and the output of the map detection model comprises the mark of the detection object on the target electronic map data without the layout drawing errors;
the determination process of the set region proportion parameter comprises the following steps: determining a target scale of electronic map data, wherein the target scale is the scale with the highest use frequency when the electronic map is currently drawn and/or displayed; determining the proportion between the geometric side lengths corresponding to the detection objects in the electronic map data represented by the target scale; and determining the ratio between the geometric side lengths as a set region ratio parameter of the detection object.
9. An electronic device, comprising:
one or more processors;
a storage device for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the problem map detection method of any of claims 1-7.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the problem map detection method according to any one of claims 1 to 7.
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 CN109977191A (en) | 2019-07-05 |
| CN109977191B true 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) |
Families Citing this family (9)
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| 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 |
Family Cites Families (4)
| 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 |
| CN105513071B (en) * | 2015-12-08 | 2017-12-19 | 江苏省基础地理信息中心 | A kind of topographic map symbols quality evaluating method |
| US11216494B2 (en) * | 2017-09-22 | 2022-01-04 | Accenture Global Solutions Limited | Virtual artificial intelligence based consultant |
| CN109145171B (en) * | 2018-07-23 | 2020-09-08 | 广州市城市规划勘测设计研究院 | Multi-scale map data updating method |
-
2019
- 2019-04-01 CN CN201910257029.0A patent/CN109977191B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| 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 (2)
| Title |
|---|
| "问题地图"现状分析及对策;周佳薇等;《测绘技术装备》;20181231;第20卷(第4期);41-43 * |
| 基于改进贝叶斯判别分析的地图修改检测;包贺先等;《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》;20141103;411-415 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109977191A (en) | 2019-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109977191B (en) | Problem map detection method, device, electronic equipment and medium | |
| CN107067003B (en) | Region-of-interest boundary extraction method, device, equipment and computer storage medium | |
| US9251420B2 (en) | System for mapping and identification of plants using digital image processing and route generation | |
| CN104424466B (en) | Method for checking object, body detection device and image pick up equipment | |
| US20190266434A1 (en) | Method and device for extracting information from pie chart | |
| CN105574550A (en) | Vehicle identification method and device | |
| CN105513038A (en) | Image matching method and mobile phone application test platform | |
| CN110910343A (en) | Method and device for detecting pavement cracks and computer equipment | |
| CN111325717B (en) | Mobile phone defect position identification method and equipment | |
| CN104463240B (en) | A kind of instrument localization method and device | |
| CN108628993B (en) | Electronic map self-adaptive classification method, device, equipment and storage medium | |
| CN112215217B (en) | Digital image recognition method and device for simulating doctor to read film | |
| CN110288612B (en) | Nameplate positioning and correcting method and device | |
| CN110866931B (en) | Image segmentation model training method and classification-based enhanced image segmentation method | |
| CN114462469B (en) | Training method of target detection model, target detection method and related device | |
| Varjo et al. | Image based visibility estimation during day and night | |
| CN110705619A (en) | Fog concentration grade judging method and device | |
| KR20230042706A (en) | Neural network analysis of LFA test strips | |
| CN116721104B (en) | Live three-dimensional model defect detection method and device, electronic equipment and storage medium | |
| CN116415020A (en) | Image retrieval method, device, electronic equipment and storage medium | |
| CN113706636A (en) | Method and device for identifying tampered image | |
| CN111488846A (en) | Method and equipment for identifying water level | |
| CN117456291B (en) | Defect classification method and device, electronic equipment and storage medium | |
| CN111753625B (en) | Pedestrian detection method, device, equipment and medium | |
| CN116468974B (en) | Smoke detection method, device and storage medium based on image generation |
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 |