CN114238772A - Intelligent network map recommendation system with content self-adaptive perception - Google Patents

Abstract

The method integrates clustering and image recommendation based on image content into retrieval and recommendation of a network map service, greatly improves the experience of map service search in the prior art, improves the efficiency of service discovery, proposes a whole set of technical process from data collection processing, thumbnail generation, image clustering and platform construction, realizes automatic sampling of the network map service based on information quantity and approximation degree, drives image clustering map recommendation by color and texture, designs a map service retrieval recommendation platform for realizing a Web end on the basis, the grasping and the identification of the map content are accurate, the intelligent recommendation of the network map can be carried out based on the content self-adaptive perception, a map service retrieval platform based on keyword search and image content recommendation is developed, the matching degree of the retrieved map content and the actual demand is high, the rationality and the effectiveness of the method are verified, and the method has high practical value.

Description

Intelligent network map recommendation system with content self-adaptive perception

Technical Field

The application relates to an intelligent network map recommending system, in particular to an intelligent network map recommending system capable of sensing content in a self-adaptive manner, and belongs to the technical field of intelligent network map content recommending.

Background

The network map service is a novel map organization mode in the mobile internet era, integrates the characteristics of multi-scale coupling, high interoperability and the like on the basis of basic map visualization, enters the big data era, and becomes the optimal knowledge expression and display service by lower resource demand, visual visualization form and seamless cross-platform connection. In recent years, the quantity and quality of map services are continuously improved, the demand of the public on the map services is also gradually improved, and how to effectively manage the map services and quickly match the demand of users becomes a research, development and application hotspot.

At present, a large amount of geographic information data, geographic information networks, geographic information portals, geographic information documents and the like are scattered on heterogeneous network nodes, and constitute massive network geographic information resources. Meanwhile, the units and individuals can perform assistant decision-making, application management, assistant scientific research and the like by acquiring various geographic information resources. In particular, network map services represented by OGC WMSs are widely used in various industries, and how to quickly search and find map services that meet the needs is a problem that needs to be solved at present.

Image content based searches have a better visual experience and more accurate search results than traditional keyword text based searches. Baidu, Google, etc. have all introduced image content-based search functionality. If the two internet ideas are introduced into the map service, the map service recommendation function based on the image content is developed, the keyword search function can be effectively assisted, the map service search experience in the prior art is improved, and the service discovery efficiency is improved.

The content-aware map recommendation system is an important and effective supplement to the text-based service search in the prior art, and the recommendation function based on the image content makes up for a single keyword search-based geographic information service discovery mode in the prior art, so that the system has important effects and great application values on enhancing user experience and improving map service discovery efficiency.

Image retrieval, which is primarily performed by retrieving images through text information such as the date of shooting, the author of the image, the category and size of the image, and the like, is a text-based image retrieval technique. Through the development of more than ten years, image retrieval is developed into content-based retrieval, images are retrieved mainly through characteristics of colors, textures, shapes and the like of the images, and the purpose is to avoid using character description and retrieve similar images through images queried by users or image characteristics specified by the users on the basis of visual similarity.

Currently, text-based image retrieval is well established, but has its drawbacks and deficiencies: firstly, the text and keyword labeling of the image needs to be completed manually; secondly, the manual labeling has the problems of different observation angles, different subjective ideas and the like, and can generate labeling ambiguity, namely, the manual labeling is difficult to completely express an image containing a plurality of targets and even covering some emotional colors. The image retrieval based on the text still has the defects, along with the improvement of market requirements and application, image retrieval applications based on image contents such as Baidu recognization drawings and Google recognization drawings appear in succession, although the accuracy is still to be improved, the image characteristics can be automatically extracted, and the ambiguity is avoided.

In summary, the prior art network map recommendation system has obvious disadvantages, and its main defects and design difficulties include:

firstly, in recent years, the quantity and quality of map services are continuously improved, the demand of the public on the map services is also gradually improved, and the prior art cannot effectively manage the map services and quickly match the demand of users; at present, a large amount of geographic information data, geographic information networks, geographic information portals, geographic information documents and the like are scattered on heterogeneous network nodes, so that massive network geographic information resources are formed, units and individuals can perform assistant decision-making, application management, assistant scientific research and the like by acquiring various geographic information resources, the prior art cannot quickly search and find map services meeting requirements, and a more intelligent map search recommendation method based on contents is lacked;

secondly, compared with the traditional retrieval based on keyword texts, the search based on image content has better visual experience and more accurate search results, the prior art does not introduce the internet idea into map service, and the map service recommendation function based on image content is lacked, so that the map service in the prior art has poor search experience and low service discovery efficiency; because the content-aware map recommendation is an effective supplement to the text-based service search in the prior art, the image content-based recommendation can make up for a single keyword search geographic information service discovery mode in the prior art, has an important role and a great application value in enhancing user experience and improving map service discovery efficiency, but the image content-based network map intelligent recommendation in the prior art is difficult to implement;

third, the prior art text-based map image retrieval is well established, but has its drawbacks and deficiencies: firstly, the text and keyword labeling of the image needs to be completed manually; secondly, the manual annotation has the problems of different observation angles, subjective idea difference and the like, and annotation ambiguity can be generated, namely the manual annotation is difficult to completely express an image containing a plurality of targets and even covering some emotional colors, the image retrieval based on the text still has defects, and image retrieval applications based on image content are sequentially appeared along with the improvement of market requirements and application, but the images cannot be applied to a map, or the problems of low accuracy, incapability of automatically extracting image characteristics, excessive ambiguity and the like exist;

fourthly, network map recommendation based on content perception has a plurality of problems, firstly, a feasible network map service automatic sampling method based on information quantity and approximation degree is lacked, a map service image sampling method under the constraint of information quantity and approximation degree is lacked, a plurality of sampling images cannot be generated for each image layer, and an image library cannot be established by using the sampling images; secondly, a feasible image clustering map recommendation method is lacked, a clustering method based on image groups is designed according to the characteristics of a plurality of sampling maps of each image layer to classify the image layers, and a similar image layer recommendation method considering the characteristics of the image groups of the image layers is lacked; and thirdly, a feasible network map service retrieval recommendation platform design and development method is lacked, and a high-quality and efficient network map intelligent retrieval recommendation platform is lacked.

Disclosure of Invention

Aiming at the defects of the prior art, the method integrates the clustering and image recommendation based on the image content into the retrieval and recommendation of the network map service, greatly improves the experience of the map service search in the prior art, improves the efficiency of service discovery, provides a whole set of technical process from data collection processing, thumbnail generation, image clustering and platform building, realizes automatic sampling of the network map service based on information quantity and approximation degree, drives image clustering map recommendation by color and texture, designs a map service retrieval recommendation platform for realizing a Web end on the basis of the information quantity and the approximation degree, has accurate grasp and identification of the map content, can carry out intelligent network map recommendation based on content self-adaptive perception, develops the map service retrieval platform based on keyword search and image content recommendation, has high matching degree of the retrieved map content with actual requirements, and verifies the reasonability and effectiveness of the method, has high practical value.

In order to achieve the technical effects, the technical scheme adopted by the application is as follows:

the intelligent network map recommending system based on content adaptive perception integrates clustering and image recommendation based on image content into retrieval and recommendation of network map service, a whole set of technical process is built from data collection processing, thumbnail generation, image clustering and platform, and a map service retrieval recommending platform for realizing a Web end is designed on the basis, and the intelligent network map recommending system mainly comprises:

first, a network map service automatic sampling method based on information quantity and approximation degree includes: firstly, calculating and evaluating the information quantity of the map tiles, and secondly, calculating the similarity of the cross-scale map tiles; positioning an area with rich content in the map by adopting an information quantity combined quad-tree structure, wherein the information quantity is approximately expressed by the complexity of an image; the approximation degree is calculated through a color quantization unified histogram and is used for screening the scaling scale with mutation, the two scales realize the ordered traversal of the map service from the plane position dimension and the scaling scale dimension respectively, and effective data are provided for clustering recommendation based on image content;

second, color and texture driven image clustering map recommendation, comprising: the method comprises the steps of firstly, the map feature collaborative representation of comprehensive color and texture features, secondly, the map content image clustering, thirdly, the clustering method based on image groups, and fourthly, the map service recommendation based on the image groups; the image group recommendation method provides a multi-feature representation method of comprehensive colors and textures by giving different weight values to groups with different approximation degrees to reduce the difference of approximation degrees caused by layer positioning, so as to realize map service recommendation based on the image group;

thirdly, a network map intelligent retrieval recommendation platform is set up, comprising: firstly, constructing a map intelligent retrieval recommendation platform, secondly, acquiring and processing network map data, thirdly, searching a map keyword database, fourthly, preprocessing a map image and searching the image, and fifthly, establishing access for the network map intelligent retrieval recommendation platform; a large number of map services are crawled by compiling a web crawler, metadata of the map services are analyzed and stored in a database, a keyword retrieval base is established and keyword retrieval services are issued based on an Apache Solr service, thumbnails are generated and stored for each service, the color and texture characteristics of each thumbnail are calculated and stored in a characteristic base, and a clustering result is stored in a category base. When recommending similar images, firstly finding the category of the images, then finding the most similar images in the category, and recommending the most similar images to the user.

The intelligent network map recommending system based on content adaptive perception further comprises a map service preview adaptive sampling method based on content, wherein the map service preview adaptive sampling method comprises the following steps: the method comprises the following steps of adopting partial network map tiles as preview pictures, under a tile map mode, defining the map service concise expression as screening a plurality of tiles capable of representing the content of the map service from a large number of map tiles, and further decomposing the map service concise expression into two sub-modules: tile selection in the plane position dimension and tile selection in the scale dimension;

(1) tile selection for the planar positional dimension: quantizing the representativeness of the map tiles based on a specific scaling, and approximately representing the information quantity of the map tiles by adopting the complexity factors of the map tiles and using the information quantity as a representative quantization index of the tiles;

(2) tile selection for the scaling dimension: the key point is that a mutated tile is found, the similarity degree between two scale tiles is quantitatively evaluated by adopting the image approximation degree, and the scaling of the significant change is judged according to the similarity degree;

based on two quantization indexes of complexity and approximation degree, the automatic selection of tiles with plane position dimension full coverage and scaling dimension multi-scale is respectively realized by two rounds of screening:

the 1 st round of screening, starting from the smallest scaling, only selecting the most complex tile at each level, and processing the next scaling by taking the geographical range of the tile as a constraint, and filtering most tiles in the first round of screening;

and 2, screening: comparing the similarity of the tiles of the scaling, further removing redundancy, and defining a sampling set as R, wherein the specific steps are as follows:

the method comprises the following steps: from L₀Stage starting, with L₀Adding R into the only map tile in the level, and using the map tile as a seed tile;

step two: requesting 4 map tiles with the same geographical range as the seed tiles at a higher level, and taking the tiles Ringo with the most abundant information amount as new seed tiles;

step three: splicing 4 tiles to obtain Rpda, if the similarity between the Rpda and the seed tile is greater than a critical value, adding the Rpda into the R, otherwise, not adding the Rpda;

repeating the second step and the third step until all scaling ratios are traversed to obtain a final sampling image set;

flexible adjustment of a plurality of variables: the number of tiles of adjacent scaling does not necessarily satisfy 4 times of the relation, the number of the tiles to be spliced is influenced, the first r (r is more than or equal to 1) tiles with the richest information amount are used as seed tiles, and the probability of obtaining the representative tiles is increased.

The intelligent network map recommendation system with content adaptive perception further comprises the following map tile information quantity calculation and evaluation: performing the 1 st round of screening based on the map information quantity, selecting the map tile with the most abundant information quantity as a seed tile under each zoom scale, traversing the next-level zoom scale by taking the geographic space range covered by the seed tile as a constraint heuristic, and dividing L₀And L₁Outside the levels, the seed tiles of other levels are not guaranteed to have the maximum information quantity, but a higher hit rate is obtained at a lower calculation cost, and after rough selection, the number of map tiles participating in comparison at each level is not more than 4;

in the aspect of map information amount calculation and evaluation, the data amount of the compressed map tiles is used as the representation index of the information amount of the map tiles, and the data amount of the map tiles in the PNG format is used as the information amount index.

The intelligent network map recommendation system with the content adaptive perception further calculates the tile approximation degree of the cross-scale map: comparing the similarity of the map tiles between two adjacent scales, and judging mutation or slow change according to the similarity, and only keeping the map tiles which are changed remarkably; analyzing the similarity degree of the map tiles starting from the color features, quantizing the color features of the expression map by adopting a color quantization unified histogram, and measuring the approximation degree of the map tiles based on cosine distance weight on the basis;

the method for generating the square map is improved by combining map expression characteristics:

improvement 1, for color quantization: firstly, quantizing colors of map tiles, splicing two tiles to be compared together, uniformly managing all colors of the two tiles, and quantizing to obtain a remarkable and unique color set;

improvement 2, calculating the frequency of appearance of color: dividing a map into a plurality of continuous color blocks, calculating the occurrence frequency of corresponding colors by using the circumferences of the color blocks, wherein the division standard of the continuous color blocks is that pixels with the same color are adjacent in four adjacent domains, and the circumference calculation adopts a Moore neighborhood boundary tracking method;

in the RGB color space, color quantization adopts a minimum variance method, the number of reserved colors is specified, the occurrence frequency of each color in two tiles is respectively counted based on a quantized color set, a color quantization unified histogram is obtained, and finally the similarity degree between the histograms is calculated by adopting cosine distance; the color weight value with the largest proportion in the histogram is set to be 0, so that the influence of the color weight value on the approximation degree is suppressed.

The intelligent recommendation system for the network map with the content self-adaptive perception is further used for recommending the image clustering map driven by colors and textures: classifying map images according to image characteristics, clustering the image layers before layer recommendation to reduce the similar layer search range, respectively using a color quantization unified histogram and a gray level co-occurrence matrix to represent color and texture characteristics of the images, and recommending the image group clustering of network map service characteristics on the basis of generating 3 sampling maps for each image layer;

and (3) the map feature collaborative representation of the comprehensive color and texture features: for a color map, its color and texture features are respectively marked as G_sAnd G_rFor a given two images D and E, the difference in color characteristics between them is recorded as A_sThe difference of texture features is A_r，k₁And k₂The weighted values of the distances between the two features are respectively, and the comprehensive distance between D and E is as follows:

A(D，E)＝k₁*A_s(GS_D，GS_E)+k2*Ar(Gr_D，Gr_E)

the method is used as a similarity comparison method for map images and retrieval.

The intelligent network map recommendation system with the content self-adaptive perception function further comprises the following steps of map content image clustering: all objects in the map data set are used as clustering centers at the beginning of clustering, each clustering center is updated through a message transmission mode in each iteration, one clustering center is selected as the center of an object failing in competition after each iteration, and a plurality of high-quality clustering results are finally obtained through continuous iteration and selection;

clustering is carried out according to the approximation degree among M map data points, the approximation degree forms an M multiplied by M similarity matrix, all map data points are taken as potential clustering centers Z, the larger the value C (w, w) on the diagonal line of the C matrix is, the higher the probability that the point w becomes the clustering center is, the value is the reference degree q, the transmission mechanism mainly comprises an attraction factor and an attribution factor, the attraction factor and the attribution factor matrix are updated, in order to avoid result oscillation, a new attenuation coefficient k is introduced, when information is updated, each information is set to be k times of the last iteration update plus 1-k times of the information update value, wherein the attenuation coefficient k is a real number between 0 and 1. I.e. the r +1 th iteration of t (j, w) and d (j, w):

d_r+1(j，w)←(1-k)d_r+1(j，w)+kd_r(j，w)

the specific steps of the map content image clustering are as follows:

step 1: initializing parameters and reading map content data;

step 2: calculating an approximation matrix, solving a median value of the approximation matrix and assigning the median value to the parameter illumination q;

and step 3: updating an attraction factor matrix and an attribution factor matrix;

and 4, step 4: judging whether the maximum iteration times is reached or a termination condition is met, if so, skipping to the step 2, otherwise, performing the step 5;

and 5: and obtaining a final clustering center Z, and dividing each data object into corresponding clusters.

The intelligent network map recommendation system based on content adaptive perception further comprises a clustering method based on image groups: generating 3 preview images for each image layer by adopting a content-based map service preview adaptive sampling method, wherein the 3 preview images are expressions of different scaling information changes of the image layer, and the 3 preview images of each image layer need to be synthesized no matter whether clustering or image layer recommendation is carried out;

taking 3 preview pictures of each layer as a picture group, clustering the picture group of each layer, firstly analyzing the characteristics of the 3 preview pictures in the picture group, wherein the 3 pictures come from the same layer at different zoom levels, the layers at different zoom levels express the same content of the layers under different scales, and the color expression of the same object under different scales cannot be changed;

when clustering an image group according to color features, the color features of the image group are integrated, and the unified color quantization histograms of 3 images A, B, C are respectively set as R_A,R_B,R_CThe color features of the image group are R-R due to the similarity of the color features of the images in the image group_A+R_B+R_CAnd calculating and considering the characteristics of the images in the image group and comprehensively considering the color characteristics of the image group.

The intelligent network map recommendation system based on content adaptive perception further recommends a map service based on image groups: recommending 3 preview images of the comprehensive consideration layer based on the similar layer of the image content, and on the basis, providing a similar service recommendation method based on a map service sampling image;

is provided with a layer D and a layer E, each layer is provided with 3 preview pictures respectively D₁、D₂、D₃And E₁、E₂、E₃Calculating the similarity between the sampling image groups of the layer D and the layer E; considering that 3 images of each image layer reduce the low similarity of two images caused by similar services being positioned to different levels or different areas, the following image group similarity calculation method is proposed:

the first step is as follows: separately calculate an image group D₁、D₂、D₃And image group E₁、E₂、E₃The image approximation degree between every two is C by adopting a color and texture feature calculation method_jiWherein j is {1,2,3} is D_jA subscript of (1), (2), (3) is E_iSubscript of (1), assume C_jiMaximum value of (1) is C₁₁In these 9 sets of comparisons, D₁And E₁Most similar, let C₁＝C₁₁Excluding from containing D_iAnd E_iAll groups of (i.e. remaining C)₂₂、C₂₃、C₃₂、C₃₃；

The second step is that: in the same way, C₂₂、C₂₃、C₃₂、C₃₃Finding out the maximum value of the four groups, and setting the maximum value as C₃₂Record C₂＝C₃₂Excluding the image D₃And image E₂At this time, D remains₂And E₃This set of images, denoted C₃＝C₂₃；

The third step: after the first step and the second step, the most similar three groups of image combination modes are found between the two image groups, in order to reduce the influence on the similarity caused by different zoom levels and different positions, different weight values k are set for each group of similarity, and the similarity between the last image groups is as follows:

C＝k₁×C₁+k₂×C₂+k₃×C₃

wherein 1. gtoreq.k₁＞k₂＞k₃≥0。

The intelligent network map recommendation system with the content self-adaptive perception function further comprises a network map intelligent retrieval recommendation platform which is set up: the method comprises the steps of geographic information service data acquisition, data processing, design and construction of a keyword search library and image feature search, construction of a Web-end network map service keyword search and similar service recommendation platform, analysis of metadata of a large number of map services by crawling of a Web crawler and storage of the metadata in a database, construction of the keyword search library, issuing of keyword search services based on Apache Solr, sampling of representative images for each layer, calculation of color and texture features of each image to construct a feature library, storage of clustering results in a category library through clustering, finding out the category of the similar images when the similar images need to be recommended, finding out the layer to which the most similar images belong under the category, and recommending the layers to users.

Map intelligent retrieval recommendation platform architecture

The method comprises the steps of collecting network map services scattered at each node of the Internet by constructing a network crawler, analyzing collected structured network map service data, storing the data into a database as metadata displayed for a user, generating thumbnails for each layer by combining a layer acquisition interface provided by the service and an automatic thumbnail generation method of the network map service, wherein the thumbnails are used as service content representatives and are data bases for subsequently carrying out image feature extraction and image clustering, and a map intelligent retrieval recommendation platform framework is divided into 3 layers:

layer 1: the network map service acquisition and preprocessing layer: the method comprises the steps of capturing scattered map services on the Internet by compiling a network map service theme crawler, downloading and analyzing metadata of the map services according to service versions and specifications of the map services, establishing a database table structure according to parameter information of the map services based on a structured organization form, and storing processed data into a MySQL database;

layer 2: image preprocessing and keyword search library construction: the layer is divided into two parts, and keyword retrieval and similar service recommendation of the platform are respectively aimed at; aiming at keyword retrieval, establishing a direction index library from the keywords to the service; generating a thumbnail for each service for similar service recommendations; the color and texture characteristics of the service which is reserved by the application are calculated in the background in advance and stored in a characteristic library, so that convenience is provided for subsequent clustering;

layer 3: platform query and recommendation logic implementation: the background logic for realizing the query recommendation platform mainly comprises the following steps: and inquiring and displaying according to the map service of the keywords, and recommending according to the similar service of the thumbnail content.

(II) network map data acquisition and processing

The method comprises the following steps of collecting OGC WMS and ArcGIS MapService on a network by using a crawler developing geographic information service topics, wherein the thought and logic of the crawler developing geographic information service topics comprise:

(1) acquiring a seed point URL by using the keyword through a search engine;

(2) reading a seed point URL webpage through a compiling crawler program, and extracting other acquired URL links from the webpage;

(3) analyzing the webpage content, and if the webpage contains WMS, web map service and ARCGIS REST Services keywords, reserving the webpage as a URL to be processed;

(4) if the URL link contains a service & WMS & request & gettapcoordinates keyword, and the page contains WMS-capcoordinates, determining that the URL link is an OGC WMS service; if the link contains/ArcGIS/REST/Services and the page content contains ArcGIS REST Services Directory, the Service is determined to be ArcGIS Service, and the Service link is stored in the database.

And storing the captured linked service into a database to obtain metadata of the service, requesting and analyzing the metadata of the webpage according to the service release specification type and the service version thereof, and only selecting more important information to store in a warehouse.

The content self-adaptive perception network map intelligent recommendation system further comprises a map keyword retrieval library: establishing a reverse index library of keywords for services, establishing mapping of the keywords for service layers, and selecting two fields of description information and a title of each service layer as fields matched with the keywords;

(1) regarding the information of each layer served by WMS and ArcGIS Map Service as a Document, selecting Title and Description in the information of each layer as a field to be segmented, delivering the information to be indexed to a Tokenizer component, and obtaining the Token Token information of each Document through the operations of de-marking point symbols and de-pause times;

(2) then, the word is handed to a language analysis tool, and the obtained word elements are processed, wherein the word elements comprise capitalization, lowercase and word root forms, and the word Term is obtained after the word elements are processed;

(3) transmitting the Term to an Indexer, establishing a word dictionary for the indexed terms, sorting the indexed terms according to the alphabetical order, and combining the same Term to form a document inverted chain List Posting List;

the application objects are two map services, namely OGC WMS and ArcGIS MapService, and the application objects are used for carrying out keyword index on service layer information and finding out proper service description information in metadata of the service layer information;

establishing a retrieval service by establishing an inverted index table based on words, English uses spaces and punctuation marks between words to separate, and retrieval result display information comprises: and connecting the layer title, the layer description information, the layer name and the sampling graph, wherein the corresponding fields are maprle, mapAesc, mapMeme and images respectively.

Keyword retrieval service publishing: the method comprises the steps of releasing keyword retrieval as an independent service to decouple the keyword retrieval service from the whole system, selecting Tomcat as a release server of Solr service, checking information in Solr through a visual web page after the service is successfully released, inquiring a retrieval result through a page configuration inquiry condition, integrating the service into a platform, adopting a SolrJ tool, and calling the Solr search service.

Compared with the prior art, the innovation points and advantages of the application are as follows:

firstly, the method integrates the clustering based on the image content and the image recommendation into the retrieval and recommendation of the network map service, greatly improves the experience of the map service search in the prior art, improves the efficiency of service discovery, provides a whole set of technical processes of data collection processing, thumbnail generation, image clustering and platform building, designs a map service retrieval recommendation platform for realizing a Web end on the basis, has accurate grasp and identification on the map content, can carry out intelligent network map recommendation based on the content self-adaptive perception, a map service retrieval platform based on keyword search and image content recommendation is developed, the recommended map layer has higher similarity with the comparison map layer, the matching degree of the retrieved map content and the actual demand is high, the rationality and the effectiveness of the method are verified, and the method has good practical value;

secondly, the method introduces the search based on the image content into the network map service, develops the map service recommendation function based on the image, assists the search based on the keywords, improves the search experience of the network map service in the prior art, and provides an automatic sampling method of the network map service based on the information quantity and the approximation degree, and the information quantity is approximately expressed by the image complexity by adopting the information quantity combined with a quad-tree structure to position the area with rich content in the map; the similarity is calculated through a color quantization unified histogram and is used for screening the mutated zoom scale, the two scales realize the ordered traversal of the map service from a plane position dimension and a zoom scale dimension respectively, effective data are provided for clustering recommendation based on image content, experiments prove that the image complexity is very effective in positioning the key position of the map, map tiles with the most abundant content under the same zoom scale can be accurately selected, good effect is achieved by using the similarity to screen the key scale, and the map service discovery efficiency is improved;

thirdly, the color feature and the texture feature are provided to comprehensively represent the image feature, the color feature is represented by a color quantization uniform histogram, the texture feature is represented by a gray level co-occurrence matrix, the inclusion relationship and the color similarity of the image group are considered, the image group recommendation method reduces the similarity difference caused by layer positioning by endowing different weight values to different groups with different similarity, the color and texture driven image clustering map recommendation method considers the inclusion relationship and the color similarity of the image group, the image group recommendation method reduces the similarity difference caused by layer positioning by endowing different weight values to different groups with different similarity, and experiments show that the color and texture feature of the image group are comprehensively considered, and the image group feature is considered, so that the image group feature has a better recommendation result on layers with different types of surface features;

fourthly, a network map intelligent retrieval recommendation platform is built, a large number of map services are crawled by compiling a network crawler, metadata of the map services are analyzed and stored in a database, a keyword retrieval base is built and the keyword retrieval services are distributed based on an Apache Solr service, thumbnails are generated and stored for each service, the color and texture characteristics of each thumbnail are calculated and stored in a characteristic base, a clustering result is stored in a category base, when similar images are recommended, the category of the thumbnail is found, and then the most similar image is found in the category and recommended to a user.

Drawings

Fig. 1 is a diagram illustrating the data amount of map symbol output PNG images of different complexities.

Fig. 2 is an exemplary diagram for mapping a complex area based on PNG data volume.

FIG. 3 is a multi-level conceptual division diagram of a multi-source vector road network real-time fusion model.

FIG. 4 is a similar color feature display in the image set of the present application.

FIG. 5 is a schematic diagram of a first step filtering of a map service recommendation based on a group of images.

FIG. 6 is a diagram illustrating a second filtering step of a map service recommendation based on a group of images.

Fig. 7 is a schematic diagram of an input image set according to experiment one of the present application.

Fig. 8 is a schematic diagram of a recommended image group in experiment one of the present application.

Fig. 9 is a schematic diagram of an input image set in experiment two of the present application.

FIG. 10 is a schematic diagram of a set of recommended images for experiment two of the present application.

Fig. 11 is a schematic diagram of an input image set of experiment three of the present application.

Fig. 12 is a schematic diagram of a recommended image group in experiment three of the present application.

FIG. 13 is a frame diagram of a network map intelligent recommendation platform with content adaptive awareness.

Fig. 14 is a diagram illustrating specific steps of keyword indexing according to the present application.

FIG. 15 is a diagram illustrating steps of a map image pre-processing and image retrieval process.

Detailed description of the invention

The technical solution of the content adaptive sensing network map intelligent recommendation system provided in the present application is further described below with reference to the accompanying drawings, so that those skilled in the art can better understand the present application and can implement the present application.

The user experience of map service searches and the efficiency of service discovery are of paramount importance, and image content-based searches have a better visual experience and better image search efficiency than keyword text-based searches. The image content-based search function is introduced into the network map service, and if the idea is introduced into the network map service, the image-based map service recommendation function is developed, so that the keyword-based search can be effectively assisted, the search experience of the network map service in the prior art is improved, and the map service discovery efficiency is improved.

The method integrates clustering and image recommendation based on image content into retrieval and recommendation of the network map service, improves the experience of map service search in the prior art, improves the efficiency of service discovery, provides a whole set of technical process from data collection processing, thumbnail generation, image clustering and platform building, designs a map service retrieval recommendation platform for realizing a Web end on the basis, and mainly comprises the following steps:

first, a network map service automatic sampling method based on information quantity and approximation degree includes: calculating and evaluating the information quantity of the map tiles, and calculating the similarity of the cross-scale map tiles;

the method for generating the map service thumbnail based on the information quantity and the similarity adopts the information quantity combined with a quadtree structure to position the rich-content area in the map, wherein the information quantity is approximately expressed by the image complexity; the approximation degree is calculated through a color quantization unified histogram and is used for screening the scaling scale with sudden change, the two scales respectively realize the ordered traversal of the map service from the plane position dimension and the scaling scale dimension, and effective data are provided for clustering recommendation based on image content.

Second, color and texture driven image clustering map recommendation, comprising: the method comprises the steps of map feature collaborative representation of comprehensive color and texture features, map content image clustering, a clustering method based on image groups and map service recommendation based on the image groups;

the image group recommendation method provides a multi-feature representation method of comprehensive colors and textures by giving different weight values to groups with different approximation degrees to reduce the difference of approximation degrees caused by layer positioning, so as to realize map service recommendation based on the image group;

thirdly, a network map intelligent retrieval recommendation platform is set up, comprising: the method comprises the following steps of constructing a map intelligent retrieval recommendation platform, acquiring and processing network map data, searching a map keyword database, preprocessing a map image and retrieving the image, and establishing access for the network map intelligent retrieval recommendation platform;

the map service retrieval platform based on keyword search and image content recommendation comprises the following steps: a large number of map services are crawled by compiling a web crawler, metadata of the map services are analyzed and stored in a database, a keyword retrieval base is established and keyword retrieval services are issued based on an Apache Solr service, thumbnails are generated and stored for each service, the color and texture characteristics of each thumbnail are calculated and stored in a characteristic base, and a clustering result is stored in a category base. When recommending similar images, firstly finding the category of the images, then finding the most similar images in the category, and recommending the most similar images to the user. The platform adopts a B/C structure, and can stably run through a web page after being released.

The map service recommendation method based on the image content is provided, a map service retrieval platform based on keyword search and image content recommendation is developed, the recommended layer and the comparison layer have higher similarity, and the reasonability and the effectiveness of the method are verified.

Content-based map service preview adaptive sampling method

Map services in the prior art mostly exist in the form of map tiles, and the map tiles are small-size map images which are pre-rendered by a server side and are segmented according to a quadtree rule. For a network map service involving n levels of scaling, the minimum scale L₀A level contains only 1 tile covering the full map area, the maximum scale L_n-1The level contains 4n-1 map tiles, the size of the map tiles is close to the requirement of the thumbnail, compression is not needed, flexible splicing can be carried out, therefore, part of network map tiles are used as preview images, in a tile map mode, the map service concise expression is defined as screening a plurality of tiles capable of representing the content of the map service from a large number of map tiles, and the tiles are further decomposed into two sub-modules: tile selection in the planar position dimension is compared to tile selection in the scale dimension.

(1) Tile selection for the planar positional dimension: based on a specific scaling, the representativeness of the map tiles is quantized, and for a static map, people tend to pay more attention to more complicated parts, such as densely-distributed symbols or compactly-placed notes, because such areas can provide richer information quantity, and from the idea, the information quantity of the map tiles is approximately represented by the complexity factors of the map tiles and is used as a representative quantization index of the tiles;

(2) tile selection for the scaling dimension: the key point is that the mutated tiles are found, and map tiles with different scales have common points and also have differences; the common points are more, and the change among the scales is slight; different points are various, and the change among the scales is obvious; subtle and dramatic correspond to gradual and abrupt changes in the map scale space, which are more noticeable to the user and more desirable for incorporation into the map service thumbnail. The method and the device adopt the image approximation degree to quantitatively evaluate the similarity degree between the two scale tiles, and determine the scaling with the obvious change according to the similarity degree.

Based on two quantization indexes of complexity and approximation degree, the automatic selection of tiles with plane position dimension full coverage and scaling dimension multi-scale is respectively realized by two rounds of screening:

the 1 st round of screening, starting from the smallest scaling, only selecting the most complex tile at each level, and processing the next scaling by taking the geographical range of the tile as a constraint, and filtering most tiles in the first round of screening;

and 2, screening: comparing the similarity of the tiles of the scaling, further removing redundancy, and defining a sampling set as R, wherein the specific steps are as follows:

the method comprises the following steps: from L₀Stage starting, with L₀Adding R into the only map tile in the level, and using the map tile as a seed tile;

step two: requesting 4 map tiles with the same geographical range as the seed tiles at a higher level, and taking the tiles Ringo with the most abundant information amount as new seed tiles;

step three: splicing 4 tiles to obtain Rpda, if the similarity between the Rpda and the seed tile is greater than a critical value, adding the Rpda into the R, otherwise, not adding the Rpda;

repeating the second step and the third step until all scaling ratios are traversed to obtain a final sampling image set;

flexible adjustment of a plurality of variables: the number of tiles of adjacent scaling does not necessarily satisfy 4-fold relationship, the number of tiles affecting splicing, and the first r (r is more than or equal to 1) tiles with the richest information content are used as seed tiles, so that the probability of obtaining representative tiles is increased.

Map tile information quantity calculation and evaluation

From rough to detailed, the number of map tiles at each level of scaling increases exponentially, and the amount of calculation is very large if all the tiles are compared. The method comprises the steps of performing 1 st round of screening based on map information quantity, selecting the map tile with the most abundant information quantity as a seed tile under each zoom scale, traversing the next-level zoom scale by taking the geographic space range covered by the seed tile as a constraint heuristic, and dividing L₀And L₁Beyond the level, there is no guarantee that the seed tiles of other levels have the maximum amount of information, but a higher amount can be obtained with less computational costHit rate, after rough selection, each level of map tiles participating in comparison does not exceed 4, and calculated amount is greatly reduced;

in the aspect of map information quantity calculation and evaluation, the prior art is mostly based on the statistical analysis of vector targets and is not suitable for the information quantity calculation of map tiles (raster images). Compared with a natural scene image, the map comprises regular characters and map symbols, the number of colors is small, the region boundary is sharp, the homogeneity of color blocks is high, and a lossless compression method is adopted to obtain a high compression ratio. The data volume of the map tiles after compression is used as the representation index of the information volume. Most map tiles are in a PNG format, are very suitable for processing homogeneous areas in images, have high compression ratio and can better meet the compression requirement of maps. Fig. 1 lists data volumes of map symbols with different complexities for outputting PNG images, and data volumes of square symbols, dot symbols, annular symbols and color symbols are sequentially increased, which shows that the PNG is very sensitive to image changes generated by visual variables such as shapes, textures, colors and the like, and can reflect the complexity of map contents. Fig. 2 shows an example of positioning a map complex area based on PNG data volume, where each image is formed by splicing four tiles, the dashed box indicates the tile with the largest data volume and requests the next scaling in this range, and the data volume of the PNG image can better reflect the complexity factor of the map content from the three times of amplification processes. Therefore, the data volume of the map tiles in the PNG format is used as the information volume index.

(II) calculating the tile approximation degree of the cross-scale map

The screening is done inside each scale according to the amount of information, without considering the relationship between the different scales. After screening according to the information quantity, obtaining a map tile at each zoom scale, and forming a map tile set. The set may contain several very similar tiles, with expression redundancy when used as a preview. The map tile approximation degree of two adjacent scaling ratios is compared, mutation or gradual change is judged according to the map tile approximation degree, only the map tiles with obvious changes are reserved, although maps with different scaling ratios have differences in content detail degree and symbol style, overall ground object layout and distribution modes are similar, the tile approximation degree calculation difficulty is reduced to a certain degree and is used as an important visual variable of map symbols, color features play a decisive role in human map perception and cognition processes, and therefore the map tile similarity degree is analyzed starting from the color features.

The method for generating the square map is improved by combining map expression characteristics:

improvement 1, for color quantization: for the sake of readability and comprehensiveness, the types of symbols included in the map are limited, and the number of corresponding colors is also greatly lower than that of a natural scene image, however, due to application of technologies such as anti-aliasing and the like, color gradient is introduced into edge pixels of the map symbols, so that the total amount of colors is indirectly improved, as shown in fig. 3, the map tile in a totally includes 8238 colors, so that the map tile is firstly subjected to color quantization, B, C and D in fig. 3 are quantization results of a with different color amounts, a visible tile can be accurately expressed by only 16 colors, if two tiles are quantized respectively, the obtained color sets may not be consistent, the similarity of colors itself needs to be additionally considered in subsequent histogram distance calculation, and the calculation complexity is high. Firstly, splicing two tiles to be compared together, uniformly managing all colors of the two tiles, and quantizing to obtain a remarkable and unique color set;

improvement 2, calculating the frequency of appearance of color: the conventional method is to count the number of pixels belonging to a certain color, and the method cannot reasonably reflect the map content composition, because map symbols, especially linear symbols and dot symbols, usually express important information by a small number of pixels, and the histogram based on the number of pixels can greatly weaken the positions of the two types of symbols. Aiming at the problem, the map is firstly divided into a plurality of continuous color blocks, the occurrence frequency of corresponding colors is calculated by using the circumferences of the color blocks, the division standard of the continuous color blocks is that pixels with the same color are adjacent in four adjacent domains, and the circumference calculation adopts a Moore neighborhood boundary tracking method.

In the RGB color space, the color quantification adopts a minimum variance method to specify the reserved color quantity. Based on the quantized color set, the occurrence frequency of each color in the two tiles is respectively counted to obtain a color quantization unified histogram, and finally the similarity degree between the histograms is calculated by adopting cosine distance. From the histograms of the two tiles at a color quantity of 16, where the approximation is 0.172. the color quantity affects the value of the approximation, it is found after many experimental attempts that the approximation gradually becomes stable when the number of colors is greater than the colors actually present in the map. Therefore, subsequent experiments set a relatively high amount of color. In addition, the thematic map often contains large-area blank areas or ground color filling, and the areas with too large weight values weaken the positions of other map symbols, thereby causing interference to the similarity index. The method and the device adjust during the calculation of the approximation degree, and set the color weight value with the largest proportion in the histogram to be 0 to inhibit the influence of the color weight value on the approximation degree.

(III) map service preview adaptive sampling experimental analysis

Three mapping services were selected for verification of the method in the experiment, including a public mapping service and two thematic mapping services (a medical facility map and a river grade map). First acquire three services at minimum scaling (i.e., L)₀Level) map tiles are used as initial seed tiles, and from the automatically acquired 10-level zoom scale representative map tiles, public map services can be finally positioned to the building dense area with rich contents, and the whole amplification process well reflects the content change of the map services. The two thematic map services get a little map tile content at the maximum zoom scale, but from the previous several levels of zoom, the positioning of the complex area is accurate.

Color quantization with 32 colors followed by inter-scale tile proximity calculations, overall, two types of map services exhibit different regularity in the magnification process: the sudden change of the public map service occurs in the later period, and the content of the map is remarkably changed due to the introduction of elements such as blocks, buildings, greenbelts and the like; abrupt changes of thematic map services occur at the early stage because of the mitigation of symbol conflicts, the selection of element targets, the conversion of expression models, the replacement of symbol patterns, and the like. Through repeated research and judgment of multiple experiments, the thumbnail screening is carried out by setting 0.1 as a critical value, and if the similarity is greater than 0.1, the thumbnail is considered to have sudden change and is brought into the thumbnail set; otherwise, they are not thumbnails.

Second, color and texture driven image clustering map recommendation

The map image clustering recommendation method based on the network map service features is provided on the basis of generating 3 sampling maps for each map layer.

Map feature collaborative representation of composite color and texture features

For a color map, its color and texture features are respectively marked as G_sAnd G_rFor a given two images D and E, the difference in color characteristics between them is recorded as A_sThe difference of texture features is A_r，k₁And k₂The weighted values of the distances between the two features are respectively, and the comprehensive distance between D and E is as follows:

A(D，E)＝k₁*A_s(G_SD，G_SE)+k₂*A_r(G_rD，G_rE)

the method is used as a similarity comparison method for map images and retrieval.

(II) map content image clustering

All objects in the map data set are used as clustering centers at the beginning of clustering, each clustering center is updated through a message transmission mode in each iteration, one clustering center is selected as the center for the object failing in competition after each iteration, and a plurality of high-quality clustering results are finally obtained through continuous iteration and selection, so that the influence of the initial random clustering center in the partitioning method on the final result is avoided, and the result is very stable.

Clustering is carried out according to the approximation degree among M map data points, the approximation degree forms an M multiplied by M similarity matrix, all map data points are taken as potential clustering centers Z, the larger the value C (w, w) on the diagonal line of the C matrix is, the higher the probability that the point w becomes the clustering center is, the value is the reference degree q, the transmission mechanism mainly comprises an attraction factor and an attribution factor, the attraction factor and the attribution factor matrix are updated, in order to avoid result oscillation, a new attenuation coefficient k is introduced, when information is updated, each information is set to be k times of the last iteration update plus 1-k times of the information update value, wherein the attenuation coefficient k is a real number between 0 and 1. I.e. the r +1 th iteration of t (j, w) and d (j, w):

d_r+1(j，w)←(1-k)d_r+1(j，w)+kd_r(j，w)

the specific steps of the map content image clustering are as follows:

step 1: initializing parameters and reading map content data;

step 2: calculating an approximation matrix, solving a median value of the approximation matrix and assigning the median value to the parameter illumination q;

and step 3: updating an attraction factor matrix and an attribution factor matrix;

and 4, step 4: judging whether the maximum iteration times is reached or a termination condition is met, if so, skipping to the step 2, otherwise, performing the step 5;

and 5: and obtaining a final clustering center Z, and dividing each data object into corresponding clusters.

(III) clustering method based on image group

According to the method, 2355 WMS layers are selected, a map service preview self-adaptive sampling method based on contents is adopted, 3 preview images are generated for each layer, the 3 preview images are expressions of different scaling information changes of the layer, and no matter clustering or layer recommendation is carried out, the 3 preview images of each layer need to be synthesized.

The 3 preview images of each layer are taken as an image group, the image group of each layer is clustered, the characteristics of the 3 preview images in the image group are firstly analyzed, the 3 images come from the same layer at different zoom levels, the layers at different zoom levels express the same content of the layers under different scales, the color expression of the same object under different scales cannot be changed, and most of the 3 images in the image group have similar color characteristics, as shown in fig. 4.

When clustering an image group according to color features, the color features of the image group are integrated, and the unified color quantization histograms of 3 images A, B, C are respectively set as R_A,R_B,R_CThe color features of the image group are R-R due to the similarity of the color features of the images in the image group_A+R_B+R_CAnd calculating and considering the characteristics of the images in the image group and comprehensively considering the color characteristics of the image group.

(IV) map service recommendation based on image groups

Based on 3 preview images of the similar image layer recommendation comprehensive consideration image layer of the image content, the similar service recommendation method based on the map service sampling image is provided.

Is provided with a layer D and a layer E, each layer is provided with 3 preview pictures respectively D₁、D₂、D₃And E₁、E₂、E₃And calculating the similarity between the sampling image groups of the layer D and the layer E.

Considering that 3 images of each layer come from different zoom levels, even if layer D is similar to layer E (layers of each zoom level are similar), in the image sampling and screening process, the 3 images of each layer are most likely to be selected in different areas or different zoom levels due to a single level of local information difference, which most likely results in lower similarity among the group of images.

Considering that 3 images of each image layer reduce the low similarity of two images caused by similar services being positioned to different levels or different areas, the following image group similarity calculation method is proposed:

the first step is as follows: separately calculate an image group D₁、D₂、D₃And image group E₁、E₂、E₃The image approximation degree between every two is C by adopting a color and texture feature calculation method_jiWherein j is {1,2,3} is D_jA subscript of (1), (2), (3) is E_iSubscript of (1), assume C_jiMaximum value of (1) is C₁₁In these 9 sets of comparisons, D₁And E₁Most similar, let C₁＝C₁₁At this time, as shown in FIG. 5, D is excluded_iAnd E_iAll groups of (i.e. remaining C)₂₂、C₂₃、C₃₂、C₃₃；

The second step is that: in the same way, C₂₂、C₂₃、C₃₂、C₃₃Finding out the maximum value of the four groups, and setting the maximum value as C₃₂Record C₂＝C₃₂Excluding the image D as shown in FIG. 6₃And image E₂At this time, D remains₂And E₃This set of images, denoted C₃＝C₂₃；

The third step: after the first step and the second step, the most similar three groups of image combination modes are found between the two image groups, in order to reduce the influence on the similarity caused by different zoom levels and different positions, different weight values k are set for each group of similarity, and the similarity between the last image groups is as follows:

C＝k₁×C₁+k₂×C₂+k₃×C₃

wherein 1. gtoreq.k₁＞k₂＞k₃≥0。

(V) image clustering map recommendation experiment analysis

On the basis of the map content image clustering result, a similarity recommendation experiment is carried out by adopting the image group similarity measurement method provided by the application, and the experiment result is as follows:

experiment one: the recommendation result of the image group shown in fig. 7 is shown in fig. 8, and it can be obtained through a first experiment that the input data is map base map data, and the recommendation results are map layers, so that the image group similarity calculation method provided by the present application has a better recommendation result for map planar layers. The recommendation result takes both the color feature and the texture feature of the image group into account.

Experiment two: the recommended results of the image group and the image group are shown in fig. 9, the input image of experiment two is a complex image, the image has complex textures and complex colors, the recommended results are all the images with similar textures, and the images with similar colors are arranged in front. The image group similarity measurement method provided by the application has a good effect on complex images.

Experiment three: image group the recommendation result of fig. 11 is as shown in fig. 12, the three inputs of the experiment are layers of simple linear symbols, the recommendation results are linear layers, linear image groups with similar colors have high similarity, and linear image groups with large color difference are arranged in the following. The image group similarity measurement method provided by the application has better recommendation on the image group with linear texture.

The three groups of experiments show that the image group recommendation method aiming at the network map service features comprehensively considers the color and texture features of the image group, and has better recommendation results for different ground object types.

Thirdly, building an intelligent retrieval recommendation platform of network map

The method comprises the steps of geographic information service data acquisition, data processing, design and construction of a keyword search library and image feature search, construction of a Web-end network map service keyword search and similar service recommendation platform, analysis of metadata of a large number of map services by crawling of a Web crawler and storage of the metadata in a database, construction of the keyword search library, issuing of keyword search services based on Apache Solr, sampling of representative images for each layer, calculation of color and texture features of each image to construct a feature library, storage of clustering results in a category library through clustering, finding out the category of the similar images when the similar images need to be recommended, finding out the layer to which the most similar images belong under the category, and recommending the layers to users.

Map intelligent retrieval recommendation platform architecture

The method comprises the steps of collecting network map services scattered at each node of the Internet by building a network crawler, analyzing collected structured network map service data, storing the data into a database as metadata displayed for a user, generating thumbnails for each layer by combining a layer acquisition interface provided by the service and an automatic generation method of network map service thumbnails, wherein the thumbnails are used as service content representatives and are a data basis for subsequently carrying out image feature extraction and image clustering, and a retrieval recommendation platform is built, so that the quick retrieval of the network map service and the acquisition of related map services are facilitated, and the information acquisition capability is improved. Based on this, the application proposes a map intelligent retrieval recommendation platform framework as shown in fig. 13.

The map intelligent retrieval recommendation platform framework is divided into 3 layers:

layer 1: the network map service acquisition and preprocessing layer: the method comprises the steps of capturing scattered map services on the Internet by compiling a network map service theme crawler, downloading and analyzing metadata of the map services according to service versions and specifications of the map services, establishing a database table structure according to parameter information of the map services based on a structured organization form, and storing processed data into a MySQL database;

layer 2: image preprocessing and keyword search library construction: the hierarchical pair is divided into two parts, and keyword retrieval and similar service recommendation of the platform are respectively aimed at; aiming at keyword retrieval, establishing a direction index library from the keywords to the service; generating a thumbnail for each service for similar service recommendations; the application has preset color and texture characteristics of the service, and because the characteristic extraction calculation is complex, the characteristics of the service are calculated in advance in the background and stored in a characteristic library, so that convenience is provided for subsequent clustering;

layer 3: platform query and recommendation logic implementation: the background logic for realizing the query recommendation platform mainly comprises the following steps: and inquiring and displaying according to the map service of the keywords, and recommending according to the similar service of the thumbnail content.

(II) network map data acquisition and processing

By adopting the method for developing the geographic information service topic crawler to collect the OGC WMS and the ArcGIS MapService on the network, the thought and logic of the geographic information service topic crawler are as follows:

(1) acquiring a seed point URL by using the keyword through a search engine;

(2) reading a seed point URL webpage through a compiling crawler program, and extracting other acquired URL links from the webpage;

(3) analyzing the webpage content, and if the webpage contains WMS, web map service and ARCGIS REST Services keywords, reserving the webpage as a URL to be processed;

(4) if the URL link contains a service & WMS & request & gettapcoordinates keyword, and the page contains WMS-capcoordinates, determining that the URL link is an OGC WMS service; if the link contains/ArcGIS/REST/Services and the page content contains ArcGIS REST Services Directory, the Service is determined to be ArcGIS Service, and the Service link is stored in the database.

The captured services are stored in a database in a linked mode, in order to obtain metadata of the services, requests and metadata analysis are carried out on web pages according to the service release specification type and the service version of the services, and only important information is selected to be stored in a warehouse due to the fact that the service information to be displayed is limited.

(III) map keyword search library

If the query of the service based on the keywords is simply matched from the database by using SQL statements, the speed is very low, the jamming is easy to occur under the condition of high concurrency, and the efficiency is low. Therefore, in order to solve the problems and improve the efficiency, a reverse index library of keywords for services is established, mapping of the keywords for service layers is established, and two fields of description information and a title of each service layer are selected as fields matched with the keywords. The specific steps for building the key index are shown in fig. 14.

(1) Regarding the information of each layer served by WMS and ArcGIS Map Service as a Document, selecting Title and Description in the information of each layer as a field to be segmented, delivering the information to be indexed to a Tokenizer component, and obtaining the Token Token information of each Document through the operations of de-marking point symbols and de-pause times;

(2) then, the word is handed to a language analysis tool, and the obtained word elements are processed, wherein the word elements comprise capitalization, lowercase and word root forms, and the word Term is obtained after the word elements are processed;

(3) transmitting the Term to an Indexer, establishing a word dictionary for the indexed terms, sorting the indexed terms according to the alphabetical order, and combining the same Term to form a document inverted chain List Posting List;

the application targets two map services, namely OGC WMS and ArcGIS MapService, and in order to perform keyword indexing on service layer information, proper service description information is found in metadata of the service layer information.

Establishing a retrieval service by establishing an inverted index table based on words, English uses spaces and punctuation marks between words to separate, and retrieval result display information comprises: and connecting the layer title, the layer description information, the layer name and the sampling graph, wherein the corresponding fields are maprle, mapAesc, mapMeme and images respectively.

Keyword retrieval service publishing: in order to relieve the server pressure under high concurrency and improve the system stability, key retrieval is issued as an independent service, so that the key retrieval service is decoupled from the whole system, the system stability can be improved, later-stage system maintenance is facilitated, Tomcat is selected as an issuing server of Solr service, information in Solr is checked through a visual web page after the service is successfully issued, a retrieval result is inquired by configuring inquiry conditions on the page, the service is integrated into a platform, a SolrJ tool is adopted, and the Solr search service is called.

(IV) map image preprocessing and image retrieval

The method comprises the steps of generating sampling graphs for OGC WMS, referring to respective image acquisition interfaces, then previewing a self-adaptive sampling method based on the map service of contents, generating thumbnails for layers of each service, and for a recommendation system, evaluating an important index of performance of the recommendation system, namely retrieval speed, large image resource data quantity, more images in an image library, and low sequential retrieval efficiency.

After extracting the image characteristic database, carrying out clustering operation on the color and texture characteristics of the images, establishing an image classification table, wherein the images in each class are similar image sets, in the process of carrying out keyword query, the thumbnail of the layer selected by a user is automatically used as an input image, the class of the input image is directly searched in a clustering result base, then, retrieval is carried out in the class, and the image which is most similar to the input image or meets the specified condition is found and recommended to the user.

The map image preprocessing and image retrieval process includes the flow of fig. 15.

(V) establishing access of intelligent network map retrieval recommendation platform

The platform is based on a B/S framework, a Java technology stack is used as a B/S framework implementation technology, network map service management, keyword-based service retrieval and image content-based service recommendation are integrated, a technical scheme for integrating all services is provided, a platform front end uses a Bootstrap technology to build a display page, a background uses a JavaEE SpringMVC + Spring + MyBatis integration framework to develop background services, Tomcat is a background server, Solr and Tomcat servers are used for issuing retrieval services, a keyword query interface is provided for the background to call, a MySQL database is used for storing map service metadata information, and Redis is used as a cache database for caching the map service metadata information.

(1) Database and Redis cache

Each map service comprises a plurality of map layer information, and the number of the map layer information is from several to thousands. The layer information is fixed. If the layer under the same service is retrieved from MySQL for display every time the service information is displayed, the efficiency is inevitably low. Generating the layer information cache of each service is the solution of the present application.

The specific cache generation time and method are as follows: when the service is accessed for the first time, the description information of the service and the layer information contained in the description information are searched from MySQL and are stored into Redis in a json format; when the access is carried out after the second time, the service information corresponding to the service is directly obtained from the Redis, and the response speed of the page is improved.

(2) Map service publishing and access

The website background server adopts Tomcat, and the main functions are demonstrated as follows: and the website entry page is used for searching by inputting keywords.

Claims (10)

1. The intelligent network map recommending system based on content adaptive perception is characterized in that clustering and image recommendation based on image content are integrated into retrieval and recommendation of a network map service, a whole set of technical processes are built from data collection processing, thumbnail generation, image clustering and a platform, and then a map service retrieval and recommendation platform for realizing a Web end is designed on the basis, and the intelligent network map recommending system mainly comprises the following steps:

first, a network map service automatic sampling method based on information quantity and approximation degree includes: firstly, calculating and evaluating the information quantity of the map tiles, and secondly, calculating the similarity of the cross-scale map tiles; positioning an area with rich content in the map by adopting an information quantity combined quad-tree structure, wherein the information quantity is approximately expressed by the complexity of an image; the approximation degree is calculated through a color quantization unified histogram and is used for screening the scaling scale with mutation, the two scales realize the ordered traversal of the map service from the plane position dimension and the scaling scale dimension respectively, and effective data are provided for clustering recommendation based on image content;

second, color and texture driven image clustering map recommendation, comprising: the method comprises the steps of firstly, the map feature collaborative representation of comprehensive color and texture features, secondly, the map content image clustering, thirdly, the clustering method based on image groups, and fourthly, the map service recommendation based on the image groups; the image group recommendation method provides a multi-feature representation method of comprehensive colors and textures by giving different weight values to groups with different approximation degrees to reduce the difference of approximation degrees caused by layer positioning, so as to realize map service recommendation based on the image group;

thirdly, a network map intelligent retrieval recommendation platform is set up, comprising: firstly, constructing a map intelligent retrieval recommendation platform, secondly, acquiring and processing network map data, thirdly, searching a map keyword database, fourthly, preprocessing a map image and searching the image, and fifthly, establishing access for the network map intelligent retrieval recommendation platform; a large number of map services are crawled by compiling a web crawler, metadata of the map services are analyzed and stored in a database, a keyword retrieval base is established and the keyword retrieval services are issued based on an Apache Solr service, thumbnails are generated and stored for each service, the color and texture characteristics of each thumbnail are calculated and stored in a characteristic base, a clustering result is stored in a category base, when similar images are recommended, the category of the similar images is found, and then the most similar images are found in the category and recommended to a user.

2. The system for intelligent recommendation of network map with adaptive content awareness according to claim 1, wherein the adaptive sampling method for previewing map service based on content comprises: the method comprises the following steps of adopting partial network map tiles as preview pictures, under a tile map mode, defining the map service concise expression as screening a plurality of tiles capable of representing the content of the map service from a large number of map tiles, and further decomposing the map service concise expression into two sub-modules: tile selection in the plane position dimension and tile selection in the scale dimension;

(1) tile selection for the planar positional dimension: quantizing the representativeness of the map tiles based on a specific scaling, and approximately representing the information quantity of the map tiles by adopting the complexity factors of the map tiles and using the information quantity as a representative quantization index of the tiles;

(2) tile selection for the scaling dimension: the key point is that a mutated tile is found, the similarity degree between two scale tiles is quantitatively evaluated by adopting the image approximation degree, and the scaling of the significant change is judged according to the similarity degree;

based on two quantization indexes of complexity and approximation degree, the automatic selection of tiles with plane position dimension full coverage and scaling dimension multi-scale is respectively realized by two rounds of screening:

the 1 st round of screening, starting from the smallest scaling, only selecting the most complex tile at each level, and processing the next scaling by taking the geographical range of the tile as a constraint, and filtering most tiles in the first round of screening;

and 2, screening: comparing the similarity of the tiles of the scaling, further removing redundancy, and defining a sampling set as R, wherein the specific steps are as follows:

the method comprises the following steps: from L₀Stage starting, with L₀Adding R into the only map tile in the level, and using the map tile as a seed tile;

step two: requesting 4 map tiles with the same geographical range as the seed tiles at a higher level, and taking the tiles Ringo with the most abundant information amount as new seed tiles;

step three: splicing 4 tiles to obtain Rpda, if the similarity between the Rpda and the seed tile is greater than a critical value, adding the Rpda into the R, otherwise, not adding the Rpda;

repeating the second step and the third step until all scaling ratios are traversed to obtain a final sampling image set;

flexible adjustment of a plurality of variables: the number of tiles of adjacent scaling does not necessarily satisfy 4 times of the relation, the number of the tiles to be spliced is influenced, the first r (r is more than or equal to 1) tiles with the richest information amount are used as seed tiles, and the probability of obtaining the representative tiles is increased.

3. The intelligent content-adaptive-aware network map recommendation system according to claim 2, wherein the map tile information amount calculation evaluation: performing the 1 st round of screening based on the map information quantity, selecting the map tile with the most abundant information quantity as a seed tile under each zoom scale, traversing the next-level zoom scale by taking the geographic space range covered by the seed tile as a constraint heuristic, and dividing L₀And L₁Outside the levels, the seed tiles of other levels are not guaranteed to have the maximum information quantity, but a higher hit rate is obtained at a lower calculation cost, and after rough selection, the number of map tiles participating in comparison at each level is not more than 4;

in the aspect of map information amount calculation and evaluation, the data amount of the compressed map tiles is used as the representation index of the information amount of the map tiles, and the data amount of the map tiles in the PNG format is used as the information amount index.

4. The intelligent content-adaptive-aware network map recommendation system according to claim 1, wherein the cross-scale map tile approximation is calculated as: comparing the similarity of the map tiles between two adjacent scales, and judging mutation or slow change according to the similarity, and only keeping the map tiles which are changed remarkably; analyzing the similarity degree of the map tiles starting from the color features, quantizing the color features of the expression map by adopting a color quantization unified histogram, and measuring the approximation degree of the map tiles based on cosine distance weight on the basis;

the method for generating the square map is improved by combining map expression characteristics:

improvement 1, for color quantization: firstly, quantizing colors of map tiles, splicing two tiles to be compared together, uniformly managing all colors of the two tiles, and quantizing to obtain a remarkable and unique color set;

improvement 2, calculating the frequency of appearance of color: dividing a map into a plurality of continuous color blocks, calculating the occurrence frequency of corresponding colors by using the circumferences of the color blocks, wherein the division standard of the continuous color blocks is that pixels with the same color are adjacent in four adjacent domains, and the circumference calculation adopts a Moore neighborhood boundary tracking method;

in the RGB color space, color quantization adopts a minimum variance method, the number of reserved colors is specified, the occurrence frequency of each color in two tiles is respectively counted based on a quantized color set, a color quantization unified histogram is obtained, and finally the similarity degree between the histograms is calculated by adopting cosine distance; the color weight value with the largest proportion in the histogram is set to be 0, so that the influence of the color weight value on the approximation degree is suppressed.

5. The system for intelligently recommending a content-adaptive aware network map according to claim 1, wherein the color-and-texture-driven image clustering map recommendation comprises: classifying map images according to image characteristics, clustering the image layers before layer recommendation to reduce the similar layer search range, respectively using a color quantization unified histogram and a gray level co-occurrence matrix to represent color and texture characteristics of the images, and recommending the image group clustering of network map service characteristics on the basis of generating 3 sampling maps for each image layer;

and (3) the map feature collaborative representation of the comprehensive color and texture features: for a color map, its color and texture features are respectively marked as G_sAnd G_rFor a given two images D and E, note themThe difference in color characteristics between is A_sThe difference of texture features is A_r，k₁And k₂The weighted values of the distances between the two features are respectively, and the comprehensive distance between D and E is as follows:

A(D，E)＝k₁*A_s(G_SD，G_SE)+k₂*A_r(G_rD，G_rE)

the method is used as a similarity comparison method for map images and retrieval.

6. The system for intelligently recommending a content-adaptive aware network map according to claim 1, wherein the map content image clusters are: all objects in the map data set are used as clustering centers at the beginning of clustering, each clustering center is updated through a message transmission mode in each iteration, one clustering center is selected as the center of an object failing in competition after each iteration, and a plurality of high-quality clustering results are finally obtained through continuous iteration and selection;

clustering is carried out according to the approximation degree among M map data points, the approximation degree forms an M multiplied by M similarity matrix, all map data points are taken as potential clustering centers Z, the larger the value C (w, w) on the diagonal line of the C matrix is, the larger the probability of the point w becoming the clustering center is, the value is a reference degree q, the transmission mechanism mainly comprises an attraction factor and an attribution factor, the attraction factor and the attribution factor matrix are updated, in order to avoid oscillation of the result, a damping coefficient k is introduced, when information is updated, each information is set to be k times of the last iteration update plus 1-k times of the information update, wherein the damping coefficient k is a real number between 0 and 1, namely the iteration value of the r +1 th time t (j, w) and d (j, w):

d_r+1(j，w)←(1-k)d_r+1(j，w)+kd_r(j，w)

the specific steps of the map content image clustering are as follows:

step 1: initializing parameters and reading map content data;

step 2: calculating an approximation matrix, solving a median value of the approximation matrix and assigning the median value to the parameter illumination q;

and step 3: updating an attraction factor matrix and an attribution factor matrix;

and 4, step 4: judging whether the maximum iteration times is reached or a termination condition is met, if so, skipping to the step 2, otherwise, performing the step 5;

and 5: and obtaining a final clustering center Z, and dividing each data object into corresponding clusters.

7. The system for intelligently recommending a content-adaptive aware network map according to claim 1, wherein the clustering method based on image groups comprises: generating 3 preview images for each image layer by adopting a content-based map service preview adaptive sampling method, wherein the 3 preview images are expressions of different scaling information changes of the image layer, and the 3 preview images of each image layer need to be synthesized no matter whether clustering or image layer recommendation is carried out;

taking 3 preview pictures of each layer as a picture group, clustering the picture group of each layer, firstly analyzing the characteristics of the 3 preview pictures in the picture group, wherein the 3 pictures come from the same layer at different zoom levels, the layers at different zoom levels express the same content of the layers under different scales, and the color expression of the same object under different scales cannot be changed;

when clustering an image group according to color features, the color features of the image group are integrated, and the unified color quantization histograms of 3 images A, B, C are respectively set as R_A,R_B,R_CThe color features of the image group are R-R due to the similarity of the color features of the images in the image group_A+R_B+R_CAnd calculating and considering the characteristics of the images in the image group and comprehensively considering the color characteristics of the image group.

8. The intelligent content-adaptive-aware network map recommendation system according to claim 1, wherein the map service recommendation based on image groups is: recommending 3 preview images of the comprehensive consideration layer based on the similar layer of the image content, and on the basis, providing a similar service recommendation method based on a map service sampling image;

is provided with a picture layerD and layers E, each layer respectively has 3 preview pictures, D₁、D₂、D₃And E₁、E₂、E₃Calculating the similarity between the sampling image groups of the layer D and the layer E; considering that 3 images of each image layer reduce the low similarity of two images caused by similar services being positioned to different levels or different areas, the following image group similarity calculation method is proposed:

the first step is as follows: separately calculate an image group D₁、D₂、D₃And image group E₁、E₂、E₃The image approximation degree between every two is C by adopting a color and texture feature calculation method_jiWherein j is {1,2,3} is D_jA subscript of (1), (2), (3) is E_iSubscript of (1), assume C_jiMaximum value of (1) is C₁₁In these 9 sets of comparisons, D₁And E₁Most similar, let C₁＝C₁₁Excluding from containing D_iAnd E_iAll groups of (i.e. remaining C)₂₂、C₂₃、C₃₂、C₃₃；

The second step is that: in the same way, C₂₂、C₂₃、C₃₂、C₃₃Finding out the maximum value of the four groups, and setting the maximum value as C₃₂Record C₂＝C₃₂Excluding the image D₃And image E₂At this time, D remains₂And E₃This set of images, denoted C₃＝C₂₃；

The third step: after the first step and the second step, the most similar three groups of image combination modes are found between the two image groups, in order to reduce the influence on the similarity caused by different zoom levels and different positions, different weight values k are set for each group of similarity, and the similarity between the last image groups is as follows:

C＝k₁×C₁+k₂×C₂+k₃×C₃

wherein 1. gtoreq.k₁＞k₂＞k₃≥0。

9. The intelligent content-adaptive-perception network map recommendation system according to claim 1, wherein a network map intelligent retrieval recommendation platform is constructed: the method comprises the steps of geographic information service data acquisition, data processing, design and construction of a keyword search library and image feature search, construction of a Web-end network map service keyword search and similar service recommendation platform, analysis of metadata of a large number of map services by crawling of a Web crawler and storage of the metadata in a database, construction of the keyword search library, issuing of keyword search services based on Apache Solr, sampling of representative images for each layer, calculation of color and texture features of each image to construct a feature library, storage of clustering results in a category library through clustering, finding the category of the similar image when the similar image needs to be recommended, finding out the layer to which the most similar image belongs under the category, and recommending the layer to a user;

map intelligent retrieval recommendation platform architecture

The method comprises the steps of collecting network map services scattered at each node of the Internet by constructing a network crawler, analyzing collected structured network map service data, storing the data into a database as metadata displayed for a user, generating thumbnails for each layer by combining a layer acquisition interface provided by the service and an automatic thumbnail generation method of the network map service, wherein the thumbnails are used as service content representatives and are data bases for subsequently carrying out image feature extraction and image clustering, and a map intelligent retrieval recommendation platform framework is divided into 3 layers:

layer 1: the network map service acquisition and preprocessing layer: the method comprises the steps of capturing scattered map services on the Internet by compiling a network map service theme crawler, downloading and analyzing metadata of the map services according to service versions and specifications of the map services, establishing a database table structure according to parameter information of the map services based on a structured organization form, and storing processed data into a MySQL database;

layer 2: image preprocessing and keyword search library construction: the layer is divided into two parts, and keyword retrieval and similar service recommendation of the platform are respectively aimed at; aiming at keyword retrieval, establishing a direction index library from the keywords to the service; generating a thumbnail for each service for similar service recommendations; the color and texture characteristics of the service which is reserved by the application are calculated in the background in advance and stored in a characteristic library, so that convenience is provided for subsequent clustering;

layer 3: platform query and recommendation logic implementation: the background logic for realizing the query recommendation platform mainly comprises the following steps: inquiring and displaying the map service according to the keywords, and recommending the similar service according to the thumbnail content;

(II) network map data acquisition and processing

The method comprises the following steps of collecting OGC WMS and ArcGIS MapService on a network by using a crawler developing geographic information service topics, wherein the thought and logic of the crawler developing geographic information service topics comprise:

(1) acquiring a seed point URL by using the keyword through a search engine;

(2) reading a seed point URL webpage through a compiling crawler program, and extracting other acquired URL links from the webpage;

(3) analyzing the webpage content, and if the webpage contains WMS, web map service and ARCGIS REST Services keywords, reserving the webpage as a URL to be processed;

(4) if the URL link contains a service & WMS & request & gettapcoordinates keyword, and the page contains WMS-capcoordinates, determining that the URL link is an OGC WMS service; if the link contains/ArcGIS/REST/Services and the page content contains ArcGIS REST Services Directory, determining the link as ArcGIS Service and storing the Service link into a database;

and storing the captured linked service into a database to obtain metadata of the service, requesting and analyzing the metadata of the webpage according to the service release specification type and the service version thereof, and only selecting more important information to store in a warehouse.

10. The intelligent content-adaptive-aware network map recommendation system according to claim 9, wherein the (three) map keyword search library: establishing a reverse index library of keywords for services, establishing mapping of the keywords for service layers, and selecting two fields of description information and a title of each service layer as fields matched with the keywords;

(1) regarding the information of each layer served by WMS and ArcGIS Map Service as a Document, selecting Title and Description in the information of each layer as a field to be segmented, delivering the information to be indexed to a Tokenizer component, and obtaining the Token Token information of each Document through the operations of de-marking point symbols and de-pause times;

(2) then, the word is handed to a language analysis tool, and the obtained word elements are processed, wherein the word elements comprise capitalization, lowercase and word root forms, and the word Term is obtained after the word elements are processed;

(3) transmitting the Term to an Indexer, establishing a word dictionary for the indexed terms, sorting the indexed terms according to the alphabetical order, and combining the same Term to form a document inverted chain List Posting List;

the application objects are two map services, namely OGC WMS and ArcGIS MapService, and the application objects are used for carrying out keyword index on service layer information and finding out proper service description information in metadata of the service layer information;

establishing a retrieval service by establishing an inverted index table based on words, English uses spaces and punctuation marks between words to separate, and retrieval result display information comprises: the image layer title, the image layer description information, the image layer name and the sampling image are connected, and the corresponding fields are maprle, mapAesc, mapMeme and images respectively;

keyword retrieval service publishing: the method comprises the steps of releasing keyword retrieval as an independent service to decouple the keyword retrieval service from the whole system, selecting Tomcat as a release server of Solr service, checking information in Solr through a visual web page after the service is successfully released, inquiring a retrieval result through a page configuration inquiry condition, integrating the service into a platform, adopting a SolrJ tool, and calling the Solr search service.

Images