CN114022772B - Method, system, device and storage medium for predicting space distribution of mobile vendor - Google Patents
Method, system, device and storage medium for predicting space distribution of mobile vendor Download PDFInfo
- Publication number
- CN114022772B CN114022772B CN202111353459.6A CN202111353459A CN114022772B CN 114022772 B CN114022772 B CN 114022772B CN 202111353459 A CN202111353459 A CN 202111353459A CN 114022772 B CN114022772 B CN 114022772B
- Authority
- CN
- China
- Prior art keywords
- vendor
- mobile
- subset
- spatial distribution
- street view
- 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
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000002372 labelling Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 48
- 238000012795 verification Methods 0.000 claims description 20
- 238000011176 pooling Methods 0.000 claims description 10
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 239000000969 carrier Substances 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000007306 turnover Effects 0.000 claims description 2
- 238000012545 processing Methods 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 10
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000001514 detection method Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 5
- 238000011835 investigation Methods 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000007423 decrease Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000010200 validation analysis Methods 0.000 description 3
- 239000004677 Nylon Substances 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 229920001778 nylon Polymers 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a prediction method, a prediction system, a prediction device and a prediction storage medium for spatial distribution of mobile vendors, which can be applied to the technical field of image processing. The method comprises the following steps: acquiring road network data of a preset area and a plurality of street view images corresponding to preset coordinate points in the road network data; respectively storing a plurality of street view images into a first subset and a second subset, and carrying out type labeling of mobile vendors on the street view images in the first subset; inputting the first subset and the second subset into a mobile vendor identification model to obtain the spatial position of the mobile vendor in the preset area; acquiring association factors of mobile vendors; inputting the spatial position and the correlation factor of the flow vendor in the preset area into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flow vendor in the area to be predicted. The invention can reduce labor cost and workload, and can rapidly and accurately obtain the distribution state of the flowing stall in different areas.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method, a system, a device and a storage medium for predicting the spatial distribution of mobile vendors.
Background
Stall economies are an integral part of the urban social ecosystem. In the related art, knowledge of the distribution state of mobile vendors is mainly performed in two ways: first, manual individual case investigation: that is, a typical area where the mobile vendor is active is selected, individual case data collection is performed on vendors, consumers, community residents and the like through investigation modes such as observation, interview, access and the like, and then the distribution state of the mobile vendor in the current area is analyzed through the collected data. Such investigation requires high labor costs and the investigated data can only reflect the status of the investigated area. Second, urban general survey: i.e. analysing vendor size and distribution status by means of census data. The nature of such investigation is also dependent upon the process of manually collecting data, so that the collected data can only reflect the status of the area under investigation.
Disclosure of Invention
The present invention aims to solve at least one of the technical problems existing in the prior art. Therefore, the invention provides a method, a system, a device and a storage medium for predicting the spatial distribution of mobile booths, which can quickly obtain the distribution states of the mobile booths in different areas without manually collecting data.
In one aspect, an embodiment of the present invention provides a method for predicting spatial distribution of mobile vendors, including the steps of:
acquiring road network data of a preset area;
acquiring a plurality of street view images corresponding to preset coordinate points in the road network data;
Respectively storing the street view images into a first subset and a second subset, and labeling the types of the mobile vendors for the street view images in the first subset;
inputting the first subset and the second subset into a mobile vendor identification model to obtain an identification result of the mobile vendor in the preset area;
Acquiring association factors of mobile vendors;
inputting the spatial position of the flowing vendor in the preset area and the correlation factor into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flowing vendor in the area to be predicted.
In some embodiments, the inputting the first subset and the second subset into a mobile vendor identification model to obtain an identification result of a mobile vendor in the preset area includes:
Extracting partial data of the first subset and partial data of the second subset to be stored in a training set, extracting residual data of the first subset to be stored in a verification set;
After the training set is input into a mobile vendor identification model, sequentially extracting features, pooling the features and fusing the features from street view images in the training set to obtain a first prediction category and a first boundary frame of mobile vendors on the street view images in the training set;
after inputting the verification set into a mobile vendor identification model, sequentially extracting features, pooling the features and fusing the features from street view images in the verification set to obtain a second prediction category and a second bounding box of mobile vendors on the street view images in the verification set;
determining a confidence score and an intersection region corresponding to the second bounding box;
When the confidence score is greater than or equal to a confidence threshold and the intersection area is greater than or equal to an intersection area threshold, inputting the street view image into a mobile vendor identification model, and sequentially extracting features, pooling features and fusing the features from the street view image to obtain an identification result of the mobile vendor in the preset area.
In some embodiments, prior to performing the step of inputting the training set into the mobile vendor identification model, the method further comprises the steps of:
Overturning the street view image in the training set to obtain an overturning image;
scaling the turnover image to obtain a scaled image;
And performing color gamut conversion on the scaled image to obtain a target image for training.
In some embodiments, the inputting the spatial location of the flowing vendor in the preset area and the correlation factor into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flowing vendor in the area to be predicted includes:
Inputting the spatial position of the inflow vendor in the preset area into a spatial distribution prediction model to obtain an ROC curve of model performance;
When the ROC curve meets the preset requirement, the correlation factor is input into a spatial distribution prediction model, and the spatial distribution state of the flow vendor in the region to be predicted is predicted.
In some embodiments, the obtaining a plurality of street view images corresponding to a preset coordinate point in the road network data includes:
Setting preset coordinate points at preset intervals in the road network data;
And acquiring a plurality of street view images corresponding to the preset coordinate points.
In some embodiments, the obtaining a plurality of street view images corresponding to the preset coordinate point includes:
And acquiring a plurality of street view images at different shooting angles on the preset coordinate point.
In some embodiments, the labeling the street view images within the first subset for a type of mobile vendor comprises:
Extracting cargo carriers in the street view images in the first subset;
labeling the type of the mobile vendor according to the cargo carrier.
In another aspect, an embodiment of the present invention provides a method for predicting spatial distribution of mobile vendors, including:
The first acquisition module is used for acquiring road network data of a preset area;
the second acquisition module is used for acquiring a plurality of street view images corresponding to preset coordinate points in the road network data;
the decomposition module is used for respectively storing the street view images into a first subset and a second subset, and carrying out type labeling on the street view images in the first subset;
The identification module is used for inputting the first subset and the second subset into a mobile vendor identification model to obtain an identification result of the mobile vendor in the preset area;
the third acquisition module is used for acquiring the association factors of the mobile vendors;
The prediction module is used for inputting the spatial position of the flowing vendor in the preset area and the correlation factor into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flowing vendor in the area to be predicted.
In another aspect, an embodiment of the present invention provides a device for predicting spatial distribution of mobile vendors, including:
at least one memory for storing a program;
at least one processor configured to load the program to perform the method of predicting the spatial distribution of the mobile vendor.
In another aspect, embodiments of the present invention provide a storage medium having stored therein a computer executable program for implementing a method for predicting the spatial distribution of mobile vendors when the computer executable program is executed by a processor.
The prediction method for the spatial distribution of the mobile vendors provided by the embodiment of the invention has the following beneficial effects:
According to the embodiment, a plurality of street view images corresponding to preset coordinate points on road network data of a preset area are firstly obtained, the street view images are respectively stored in a first subset and a second subset, meanwhile, the type of the flowing vendor is marked on the street view images in the first subset, so that training accuracy of an identification model is improved, then the first subset and the second subset are input into the flowing vendor identification model to obtain space positions of the flowing vendor in the preset area, position reference data are provided for the distribution prediction of the flowing vendor in other areas, then the space positions of the flowing vendor in the preset area and the obtained association factors are input into the space distribution prediction model, the space distribution state of the flowing vendor in the area to be predicted is predicted, therefore artificial collection of data is not needed, labor cost and workload are reduced, and meanwhile, the distribution state of the flowing vendor in different areas can be obtained rapidly and accurately.
Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The invention is further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a flow chart of a method for predicting spatial distribution of mobile vendors according to an embodiment of the present invention;
Fig. 2 is a schematic structural diagram of a mobile vendor identification model according to an embodiment of the invention.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative only and are not to be construed as limiting the invention.
In the description of the present invention, the meaning of a number is one or more, the meaning of a number is two or more, and greater than, less than, exceeding, etc. are understood to exclude the present number, and the meaning of a number is understood to include the present number. The description of the first and second is for the purpose of distinguishing between technical features only and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of the technical features indicated.
In the description of the present invention, unless explicitly defined otherwise, terms such as arrangement and the like should be construed broadly, and those skilled in the art can reasonably determine the specific meaning of the terms in the present invention in combination with the specific contents of the technical scheme.
In the description of the present invention, a description of the terms "one embodiment," "some embodiments," "an exemplary embodiment," "an example," "a particular example," or "some examples," etc., means that a particular feature or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Referring to fig. 1, an embodiment of the present invention provides a method for predicting spatial distribution of mobile vendors. The embodiment can be applied to a background processor corresponding to a server or a vendor management platform.
In the implementation process, the embodiment includes the following steps:
S11, road network data of a preset area are obtained.
In this embodiment, the road network data includes urban road network data, and the road network data may be directly called from a preset website, for example, directly called from OpenStreetMap. Wherein OpenStreetMap is abbreviated as OSM, which is a freely editable world map. The map on the OSM is drawn by the user holding the GPS device, aerial photographs, satellite imagery, other free content, even by the user having spatial knowledge due to familiarity with the target area.
S12, acquiring a plurality of street view images corresponding to preset coordinate points in the road network data.
In this embodiment, after the road network data is acquired, preset coordinate points are set at preset intervals on the road network data, for example, 10 meters as an interval, 20 meters as an interval, and the like. The size of the preset interval can be set by a user according to the actual condition of the road network data. After coordinate points on the current road network data are set, street view images corresponding to the coordinate points are collected. Specifically, the image acquired at each coordinate point includes at least three images. The at least three images may be street view images corresponding to the same angle, or street view images corresponding to three different shooting angles. When the street view image is obtained from street view open service of the online map, the network crawler software can be opened through an API of the online map to crawl photographed images of different angles corresponding to preset coordinate points on the road network data. For example, street view images with shooting angles of 90 °, 180 °, and 270 ° are crawled.
And S13, respectively storing a plurality of street view images into the first subset and the second subset, and carrying out type labeling on mobile vendors on the street view images in the first subset.
In this embodiment, in order to improve the training accuracy of the mobile vendor identification model, after a plurality of street view images are obtained, the street view images are respectively stored in two different subsets, and then different processing modes are adopted for the data in the two subsets. For example, the streetscape images in the first subset are subject to type criteria of the mobile vendor, but the streetscape images in the second subset are not processed to provide different types of data for model training. It will be appreciated that since different types of mobile vendors may use different tools, the type of mobile vendor may be determined by identifying the cargo vehicles included on the street view image within the first subset and then determining the type of mobile vendor based on the identified cargo vehicles. For example, the identified cargo carrier is a nylon bag that is placed on the ground and the cargo is placed on the nylon bag, then the mobile vendor may be identified as being of the ground booth type. For example, if the mobile vendor is identified as a table, and the goods are placed on the table, the mobile vendor may be identified as a table booth. For example, if the cargo carrier is identified as a tricycle and the cargo is placed in the tricycle, the type of the mobile vendor can be determined as a tricycle stall. Also, if the cargo carrier is identified as a wagon, and the cargo is placed in the wagon, then it may be determined that the mobile vendor is of the wagon booth type.
S14, inputting the first subset and the second subset into a mobile vendor identification model to obtain an identification result of the mobile vendor in the preset area.
In this embodiment, after all street view images in the first subset are processed, training the mobile vendor identification model through the first subset and the second subset, and predicting to obtain the identification result of the mobile vendor in the preset area, so as to provide data support for the mobile booth prediction process in other areas. It will be appreciated that during the training process, partial data may be extracted from the first subset and partial data may be stored together with the second subset extraction to the training set, while the remainder of the data extracted from the first subset is stored to the validation set. The partial data of the first subset in the training set can be used as positive training samples, and the number of the positive training samples is equal to that of the second subset in the training set. For example, a total of 10 samples were extracted from the first subset, 9 samples were saved into the training set, and the remaining 1 sample was used as the validation set, so the number of samples extracted from the second subset was also 9. After the training set and the verification set are obtained, the training set is input into a mobile vendor identification model, then features are sequentially extracted from street view images in the training set in the mobile vendor identification model, feature pooling and feature fusion are carried out, and then a first prediction category and a first boundary frame of mobile vendors on the street view images in the training set are obtained through prediction to serve as identification results in the current training process. After one round of training is completed, inputting the verification set into a mobile vendor identification model, sequentially extracting features, feature pooling and feature fusion from street view images in the verification set in the mobile vendor identification model, and obtaining a second prediction category and a second bounding box of the mobile vendor on the street view images in the verification set as an identification result in the current verification process. And then determining a confidence score and an intersection area corresponding to the second bounding box, wherein the intersection area refers to the overlapping area of the predicted bounding box and the marked bounding box. When the confidence score is smaller than the confidence threshold or the intersection area is smaller than the intersection area threshold, extracting part of data from the first subset to serve as a training set and a verification set respectively, extracting part of data from the second subset to be stored in the training set, retraining the mobile vendor identification model through the reacquired training set, and verifying the trained mobile vendor identification model through the reacquired verification set. And (3) circulating until the confidence score is greater than or equal to the confidence threshold and the intersection area is greater than or equal to the intersection area threshold, inputting the street view image into the trained mobile vendor identification model, and sequentially extracting the features, pooling the features and fusing the features of the street view image to obtain the identification result of the mobile vendor in the preset area.
For example, a YOLOv algorithm is adopted to construct a mobile vendor identification model, so that the mobile vendor identification model can efficiently detect the target object in image data of different illumination conditions, observation view angles, blocked target objects and complex backgrounds or scenes. The method for constructing the mobile vendor identification model by adopting YOLOv algorithm comprises the following four steps: deep network construction, model training, model inspection and model application.
Specifically, as shown in fig. 2, the mobile vendor identification model constructed based on YOLOv' 4 algorithm includes three parts: backbones for extracting features from training images, CSPDARKNET53 is adopted as Backbones; neck, consisting of SPP and PANet, where SPP is used for feature pooling and PANet is used for feature fusion; heads for predicting and outputting the predicted category and bounding box of the vendor. After model construction is completed, the street view images with the types marked are combined with some street view images without any vendor to form a model training set, and the training set is input into the recognition model. In this embodiment, in order to improve the recognition efficiency, a K-means algorithm may be used to determine a set of standard bounding boxes with widths and heights (17, 24), (29, 34), (38, 51), (50, 76), (65, 54), (69, 108), (91, 81), (107, 34), (165, 181) according to the distribution mean of the labeled bounding boxes, so as to detect vendors in street view images. In addition, in order to eliminate the interference of complex colors and backgrounds in different images, the full-color images in the training set can be converted by multiple combinations of turning, scaling, color gamut change and the like, and then introduced Backbones for recognition model training. Specifically, the K-means algorithm is a cluster analysis algorithm for iterative solution, and the processing procedure of the K-means algorithm is to decompose preset data into K groups, randomly calculate K objects as initial cluster centers, then calculate the distance between each object and each cluster center, and allocate each object to the cluster center closest to the cluster center. The cluster center and the assigned objects represent a cluster. In the training process, the confidence threshold and the aggregate part IoU formed by dividing the overlapping part of the two areas by the two areas are key parameters of the vendor identification model, and the embodiment controls the output result through the confidence threshold and the IoU threshold and affects the performance of the model. The higher the threshold, the lower the error tolerance of the detection result, and in general, the confidence threshold and IoU threshold are not properly set too low, otherwise too many false detection results are generated. To test the sensitivity of these two thresholds and calibrate the parameters, the present example conducted 8 sets of control experiments with different parameters. The confidence threshold is set between 0.4 and 0.8 at intervals of 0.1, and IoU is set between 0.5 and 0.8 at intervals of 0.1. And finally, the parameter value when the performance is optimal is found by comparing the model performance of the result obtained by different value combinations of the two thresholds.
In the process of training the recognition model, when the detection result obtained by the recognition model meets two conditions, the recognition result of the recognition model is correct. Wherein the two conditions include a confidence score equal to or greater than a confidence threshold and an intersection region equal to or greater than a IoU threshold. In the judgment process, the model performance was evaluated using F 1 values (F 1 Score, harmonic mean of accuracy and recall) and average accuracy mAP (MEAN AVERAGE Precision, average of average accuracy). The calculation formula of the F 1 value is shown as formula (1), wherein P is the precision, and the calculation is carried out by detecting the ratio of the correct number of the bootlegger labels (TP) to the total number of the detected frames (TP+FP); r is recall, calculated by detecting the ratio of the correct number of vendor Tags (TP) to the actual number of frames (TP+FN). mAP is the area of the region surrounded by the precision-recall curve and the x-axis, and can be calculated by an accumulated integral formula shown in formula (2):
where p=tp/(tp+fp), r=tp/(tp+fn), and K is the number of tag classes.
In the training effect verification process of the identification model, the embodiment adopts a ten-rule crossing mode for verification. For example, the sample data is randomly divided into 10 parts, 9 of which are alternately used for training, and the remaining 1 part is used for model verification. Model test the average of 10 experiments was used and the model finally used for the test was the one that performed optimally among the 10 experiments.
In this embodiment, the model performance of the results obtained by comparing the confidence threshold with the IoU threshold is shown in table 1.
TABLE 1
As can be seen from table 1, when the confidence threshold monotonically decreases, the recall monotonically increases; the overall trend of the precision is a decline, and the local trend is an ascending or descending. And when IoU threshold monotonically increases, recall and accuracy monotonically decreases. The first three of each index are marked and a suitable value for both thresholds can be found to be equal to 0.5, so the set of threshold parameters is used to calibrate the detection model. The detection model obtained by training the group of parameters has an average F 1 value of 0.77 and an mAP value of 66.75%, which shows that the performance of the model meets the preset requirement, and the method can be applied to the detection of the whole street view image dataset.
After training and verification of the model are completed, the street view image is input into the trained recognition model, the recognition model can output prediction frames for the street view image containing the mobile vendors, the number of the prediction frames appearing in the point position record corresponding to the street view image reflects the number of the mobile vendors detected in the point position. For example, when the street view image of the city a is identified, the number of street view images obtained is 76108, and 3907 mobile booths are identified from the 76108 street view images; for example, when the street view image of the city B is identified, the number of street view images obtained is 3339062, and 26119 mobile booths are identified from the 3339062 street view images. The mobile booths are on street view images and can be marked out through prediction frames so as to be convenient for analyzing and obtaining the space positions of the mobile booths in the current area.
S15, obtaining association factors of mobile vendors.
In this embodiment, since the mobile vendors often prefer to gather and concentrate areas with dense traffic, such as intersections, markets, and vicinity of public transportation sites, in order to obtain more business benefits when participating in business activities. But is also constrained by the proximity effect, so that the vendor is actively away from some areas. Based on this, as shown in table 2, an index system covering 11 factors was established for characterizing the spatial distribution of vendors:
TABLE 2
Wherein "+" in table 2 indicates a correlation factor having a positive effect on the mobile vendor's swing location and "-" indicates a correlation factor having a negative effect on the mobile vendor's swing location.
Specifically, the population distribution in the index 1 is mainly used for representing refined population space aggregation information in the city, information such as mobile phone signaling, smart phone users and the like can be used for inversion to form raster data of the population space distribution of the city, existing population space distribution data products can be used, the pixel size is 100×100 meters, the population quantity in the area is reflected by each piece of pixel information, and the projection coordinate is WGS84.
The intersection of index 2 can be extracted from the acquired urban road network data. Specifically, a network data set constructing tool in GIS software is used first to construct a road network from vector data of a city, and then intersection points in the network data set are extracted, namely intersection point data. And finally, calculating the linear distance between each position in the city range and the intersections by using Euclidean distance in GIS software, and outputting and forming corresponding raster data.
Metro ports, pedestrian overpasses, markets, schools, cultural leisure facilities, large traffic hubs, and CBDs (Central Business Districts, central business areas) from index 3 to index 9 can all be calculated based on POIs (Points of Interest ). Specifically, city infrastructure points including subway stations, pedestrian overpasses, markets, schools, cultural leisure facilities, large transportation hubs (e.g., airports, train stations, public transportation hubs), CBDs are first obtained from API (application programinterface ) services of a high-altitude map or a hundred-degree map. Then, the Euclidean distance of each position from the POIs in the city range is calculated by using the linear distance in GIS software, and the corresponding raster data is output and formed.
The road congestion degree in the index 10 comprises dynamic road traffic data, and is mainly used for measuring traffic speed conditions of urban roads in different time periods, so that the congestion degree is estimated, and further, the quantitative relation between vendor management and road congestion is analyzed. Dynamic passing speed data can be obtained by using an API interface of a Goldmap or a hundred-degree map, and the data are respectively input into road network sections of a research area through design of crawler software, so that the passing time of each road section can be calculated, the passing speed is obtained, and the congestion degree classification is carried out by combining the road speed limit grade.
The house rentals in index 11 are used to analyze the spatial correlation of vendor location preference with high, medium, and low-grade residential areas in a city and different consumption level populations. Lease data may be obtained from the relevant departments, or from a larger property portal or industry research facility. In the embodiment, a web crawler tool is adopted to acquire house renting and listing data from a plurality of large house renting information portal websites, each piece of acquired data records the house area, the renting price and the geographic coordinates, repeated release or error data are further cleaned, the data are converted into vector points containing area unit renting information, and a space interpolation tool is used to acquire a house renting space distribution map of a research area.
S16, inputting the spatial position and the correlation factor of the flow vendor in the preset area into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flow vendor in the area to be predicted.
In this embodiment, when the spatial distribution of the flowing vendor in the region to be predicted is predicted, the spatial position of the flowing vendor in the preset region may be input to the spatial distribution prediction model to obtain a ROC curve with model performance, and then when the ROC curve meets the preset requirement, the correlation factor is input to the spatial distribution prediction model to predict to obtain the spatial distribution state of the flowing vendor in the region to be predicted. The ROC curve is a coordinate type analysis tool, and is used for the following two aspects: firstly, selecting an optimal signal detection model and discarding a suboptimal model; second, the optimal threshold is set in the same model. It can be appreciated that the present embodiment may employ a MaxEnt niche model as a spatial distribution prediction model to predict the probability of urban area flow vendor activity to be predicted. Specifically, the identified mobile vendor was taken as the presence sample, 75% of which were randomly selected as training samples for the MaxEnt niche model, and the remaining 25% were taken as validation samples. The grid images of the correlation factors listed in table 2 are converted into ASCII format and input as an environmental layer into the MaxEnt ecomodel compiled from the R language. The MaxEnt ecological niche model is based on a training sample and a verification sample calibration model, then predicts a spatial distribution map of the occurrence probability of a vendor, and performs superposition analysis on the spatial distribution map and the existing streets and open spaces of the area to be predicted, so that the spatial distribution state of the mobile vendor in the area to be predicted can be obtained.
In summary, according to the embodiment, the recognition result of the mobile vendor in the area where the street image is located is obtained by the recognition model according to the known road network data and the street image recognition, and then the spatial distribution state of the mobile vendor in the area to be predicted is predicted and obtained by the spatial distribution prediction model in combination with the association factor and the vendor recognition result of the known area, so that the spatial distribution state of the mobile vendor in the designated area can be obtained quickly without manually collecting a large amount of data, and meanwhile, the error influence caused by manually collecting the data can be avoided.
The embodiment of the invention provides a prediction method for spatial distribution of mobile vendors, which comprises the following steps:
The first acquisition module is used for acquiring road network data of a preset area;
the second acquisition module is used for acquiring a plurality of street view images corresponding to preset coordinate points in the road network data;
the decomposition module is used for respectively storing the street view images into a first subset and a second subset, and carrying out type labeling on the street view images in the first subset;
The identification module is used for inputting the first subset and the second subset into a mobile vendor identification model to obtain an identification result of the mobile vendor in the preset area;
the third acquisition module is used for acquiring the association factors of the mobile vendors;
The prediction module is used for inputting the spatial position of the flowing vendor in the preset area and the correlation factor into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flowing vendor in the area to be predicted.
The content of the method embodiment of the invention is suitable for the system embodiment, the specific function of the system embodiment is the same as that of the method embodiment, and the achieved beneficial effects are the same as those of the method.
The embodiment of the invention provides a prediction device for spatial distribution of mobile vendors, which comprises:
at least one memory for storing a program;
At least one processor configured to load the program to perform the method of predicting the spatial distribution of the mobile vendor of fig. 1.
The content of the method embodiment of the invention is suitable for the device embodiment, the specific function of the device embodiment is the same as that of the method embodiment, and the achieved beneficial effects are the same as those of the method.
An embodiment of the present invention provides a storage medium in which a computer-executable program is stored, where the computer-executable program is used to implement the method for predicting the spatial distribution of mobile vendors shown in fig. 1 when the computer-executable program is executed by a processor.
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of one of ordinary skill in the art without departing from the spirit of the present invention. Furthermore, embodiments of the invention and features of the embodiments may be combined with each other without conflict.
Claims (9)
1. A method for predicting the spatial distribution of mobile vendors, comprising the steps of:
acquiring road network data of a preset area;
acquiring a plurality of street view images corresponding to preset coordinate points in the road network data;
Respectively storing the street view images into a first subset and a second subset, and labeling the types of the mobile vendors for the street view images in the first subset;
inputting the first subset and the second subset into a mobile vendor identification model to obtain an identification result of the mobile vendor in the preset area;
Acquiring association factors of mobile vendors;
Inputting the spatial position of the flowing vendor in the preset area and the correlation factor into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flowing vendor in the area to be predicted;
Inputting the spatial position of the flowing vendor in the preset area and the correlation factor into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flowing vendor in the area to be predicted, wherein the method comprises the following steps:
Inputting the spatial position of the inflow vendor in the preset area into a spatial distribution prediction model to obtain an ROC curve of model performance;
When the ROC curve meets the preset requirement, the correlation factor is input into a spatial distribution prediction model, and the spatial distribution state of the flow vendor in the region to be predicted is predicted.
2. The method for predicting the spatial distribution of mobile vendors according to claim 1, wherein the inputting the first subset and the second subset into the mobile vendor identification model to obtain the identification result of the mobile vendor in the preset area comprises:
Extracting partial data of the first subset and partial data of the second subset to be stored in a training set, extracting residual data of the first subset to be stored in a verification set;
After the training set is input into a mobile vendor identification model, sequentially extracting features, pooling the features and fusing the features from street view images in the training set to obtain a first prediction category and a first boundary frame of mobile vendors on the street view images in the training set;
after inputting the verification set into a mobile vendor identification model, sequentially extracting features, pooling the features and fusing the features from street view images in the verification set to obtain a second prediction category and a second bounding box of mobile vendors on the street view images in the verification set;
determining a confidence score and an intersection region corresponding to the second bounding box;
When the confidence score is greater than or equal to a confidence threshold and the intersection area is greater than or equal to an intersection area threshold, inputting the street view image into a mobile vendor identification model, and sequentially extracting features, pooling features and fusing the features from the street view image to obtain an identification result of the mobile vendor in the preset area.
3. The method of claim 2, wherein prior to performing the step of inputting the training set into the mobile vendor identification model, the method further comprises the steps of:
Overturning the street view image in the training set to obtain an overturning image;
scaling the turnover image to obtain a scaled image;
And performing color gamut conversion on the scaled image to obtain a target image for training.
4. The method for predicting the spatial distribution of mobile vendors according to claim 1, wherein the obtaining a plurality of street view images corresponding to preset coordinate points in the road network data comprises:
Setting preset coordinate points at preset intervals in the road network data;
And acquiring a plurality of street view images corresponding to the preset coordinate points.
5. The method for predicting the spatial distribution of mobile vendors according to claim 4, wherein the obtaining a plurality of street view images corresponding to the preset coordinate point comprises:
And acquiring a plurality of street view images at different shooting angles on the preset coordinate point.
6. The method of claim 1, wherein the labeling the type of the street view image in the first subset comprises:
Extracting cargo carriers in the street view images in the first subset;
labeling the type of the mobile vendor according to the cargo carrier.
7. A method for predicting the spatial distribution of mobile vendors, comprising:
The first acquisition module is used for acquiring road network data of a preset area;
the second acquisition module is used for acquiring a plurality of street view images corresponding to preset coordinate points in the road network data;
the decomposition module is used for respectively storing the street view images into a first subset and a second subset, and carrying out type labeling on the street view images in the first subset;
The identification module is used for inputting the first subset and the second subset into a mobile vendor identification model to obtain an identification result of the mobile vendor in the preset area;
the third acquisition module is used for acquiring the association factors of the mobile vendors;
the prediction module is used for inputting the spatial position of the flowing vendor in the preset area and the correlation factor into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flowing vendor in the area to be predicted;
Inputting the spatial position of the flowing vendor in the preset area and the correlation factor into a spatial distribution prediction model, and predicting to obtain the spatial distribution state of the flowing vendor in the area to be predicted, wherein the method comprises the following steps:
Inputting the spatial position of the inflow vendor in the preset area into a spatial distribution prediction model to obtain an ROC curve of model performance;
When the ROC curve meets the preset requirement, the correlation factor is input into a spatial distribution prediction model, and the spatial distribution state of the flow vendor in the region to be predicted is predicted.
8. A device for predicting the spatial distribution of mobile vendors, comprising:
at least one memory for storing a program;
At least one processor for loading the program to perform the method of predicting the spatial distribution of flowing vendors of any one of claims 1-6.
9. A storage medium having stored therein a computer executable program for implementing the method of predicting the spatial distribution of a flowing vendor of any one of claims 1-6 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111353459.6A CN114022772B (en) | 2021-11-16 | 2021-11-16 | Method, system, device and storage medium for predicting space distribution of mobile vendor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111353459.6A CN114022772B (en) | 2021-11-16 | 2021-11-16 | Method, system, device and storage medium for predicting space distribution of mobile vendor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114022772A CN114022772A (en) | 2022-02-08 |
| CN114022772B true CN114022772B (en) | 2024-05-03 |
Family
ID=80064601
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111353459.6A Active CN114022772B (en) | 2021-11-16 | 2021-11-16 | Method, system, device and storage medium for predicting space distribution of mobile vendor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114022772B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458333A (en) * | 2019-07-18 | 2019-11-15 | 华南农业大学 | A kind of population spatial distribution prediction technique and system based on POIs data |
| CN110992645A (en) * | 2019-12-06 | 2020-04-10 | 江西洪都航空工业集团有限责任公司 | Mobile vendor detection and alarm system in dynamic scene |
| CN111444818A (en) * | 2020-03-24 | 2020-07-24 | 哈尔滨工程大学 | CNN-based market stall violation stall detection method |
| CN112651293A (en) * | 2020-10-30 | 2021-04-13 | 华设设计集团股份有限公司 | Video detection method for road illegal stall setting event |
| CN112990517A (en) * | 2019-12-12 | 2021-06-18 | 中移雄安信息通信科技有限公司 | Crowd distribution prediction method and system |
-
2021
- 2021-11-16 CN CN202111353459.6A patent/CN114022772B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458333A (en) * | 2019-07-18 | 2019-11-15 | 华南农业大学 | A kind of population spatial distribution prediction technique and system based on POIs data |
| CN110992645A (en) * | 2019-12-06 | 2020-04-10 | 江西洪都航空工业集团有限责任公司 | Mobile vendor detection and alarm system in dynamic scene |
| CN112990517A (en) * | 2019-12-12 | 2021-06-18 | 中移雄安信息通信科技有限公司 | Crowd distribution prediction method and system |
| CN111444818A (en) * | 2020-03-24 | 2020-07-24 | 哈尔滨工程大学 | CNN-based market stall violation stall detection method |
| CN112651293A (en) * | 2020-10-30 | 2021-04-13 | 华设设计集团股份有限公司 | Video detection method for road illegal stall setting event |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114022772A (en) | 2022-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9547866B2 (en) | Methods and apparatus to estimate demography based on aerial images | |
| CN109493119B (en) | POI data-based urban business center identification method and system | |
| Laumer et al. | Geocoding of trees from street addresses and street-level images | |
| US20130226667A1 (en) | Methods and apparatus to analyze markets based on aerial images | |
| CN107430815A (en) | Method and system for automatic identification parking area | |
| KR102011773B1 (en) | Method for evaluating usability of empty house, server and system using the same | |
| Ding et al. | Towards generating network of bikeways from Mapillary data | |
| US20220357176A1 (en) | Methods and data processing systems for predicting road attributes | |
| CN102663164A (en) | Pass control algorithm test device and method based on multiobjective test case generation | |
| CN111626277A (en) | Vehicle tracking method and device based on over-station inter-modulation index analysis | |
| CN110472999A (en) | Passenger flow pattern analysis method and device based on subway and shared bicycle data | |
| Ghosh et al. | Automated detection and classification of pavement distresses using 3D pavement surface images and deep learning | |
| CN114708521A (en) | City functional area identification method and system based on street view image target relation perception network | |
| CN114548811A (en) | Airport accessibility detection method and device, electronic equipment and storage medium | |
| Yang et al. | Toward country scale building detection with convolutional neural network using aerial images | |
| Smith et al. | Classification of sidewalks in street view images | |
| Ye et al. | Land use classification from social media data and satellite imagery | |
| CN110555432A (en) | Method, device, equipment and medium for processing interest points | |
| CN114022772B (en) | Method, system, device and storage medium for predicting space distribution of mobile vendor | |
| Li | Mapping urban land use by combining multi-source social sensing data and remote sensing images | |
| Panizzi et al. | Private or Public Parking Type Classifier on the Driver’s Smartphone | |
| CN116628531A (en) | Crowd-sourced map road object element clustering method, system and storage medium | |
| Hammoudi et al. | Towards a model of car parking assistance system using camera networks: Slot analysis and communication management | |
| Meedeniya et al. | Land‐Use Classification with Integrated Data | |
| Sun et al. | Automatic building age prediction from street view images |
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 |