A method and device for positioning a target in a mobile scene

Abstract

The present application provides a method and device for target positioning in a mobile scenario, which relates to the field of artificial intelligence and can also be used in the financial field, including: clustering the customer behavior images acquired during the business handling process according to the business environment to obtain a cluster image; matching the cluster image according to a predetermined template image to obtain a customer initial positioning image; removing mismatched points from the customer initial positioning image to obtain a customer target positioning image. The present application can cluster the customer behavior images acquired during the business handling process according to the business environment, and finally obtain a customer target positioning image.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence and can be used in the financial field. Specifically, it is a method and device for target positioning in a mobile scenario. Background Art

[0002] At present, information technology is developing rapidly, and the banking industry has broad prospects in the direction of intelligent development. When customers go to the bank to handle business, how to accurately grasp customer needs, provide customers with convenient and targeted services, and improve customer experience is an urgent need for bank business personnel, and it is also a challenge for the banking industry in the future intelligent development process.

[0003] When customers conduct business at a bank, we can analyze their behavior patterns by locating their mobile devices and collecting continuous images, thereby improving the pertinence of the customer service process. By analyzing customer behavior patterns and conducting scientific customer demand analysis, we can accurately identify customer needs, quantify customer sensitivity, refine customer service strategies in a targeted manner, effectively control the labor costs of offline services, and enhance the public image of the company.

[0004] To achieve the above objectives, existing technologies often use grayscale image matching algorithms and cluster analysis algorithms. However, if obstacles block the movement of the target (customer), the image matched by the grayscale image matching algorithm may be inaccurate, resulting in fuzzy positioning. At the same time, most cluster analysis algorithms will become very vulnerable when there are isolated points in the data. Summary of the invention

[0005] In response to the problems in the prior art, the present application provides a target positioning method and device in a mobile scenario, which can cluster the acquired customer behavior images of the business processing process according to the business environment, and finally obtain a customer target positioning image.

[0006] In order to solve the above technical problems, this application provides the following technical solutions:

[0007] In a first aspect, the present application provides a target positioning method in a mobile scenario, comprising:

[0008] Clustering the acquired customer behavior images of the business handling process according to the business environment to obtain a cluster image;

[0009] Matching the cluster image according to a predetermined template image to obtain an initial positioning image of the customer;

[0010] The mismatched points of the customer's initial positioning image are removed to obtain the customer's target positioning image.

[0011] Furthermore, clustering the acquired customer behavior images of the business handling process according to the business environment to obtain cluster images includes:

[0012] Randomly dividing the customer behavior image into data subsets;

[0013] Run the clustering algorithm on each data subset to obtain the cluster corresponding to each data subset;

[0014] Each data subset is clustered according to the cluster to obtain a cluster image.

[0015] Furthermore, clustering the acquired customer behavior images of the business handling process according to the business environment to obtain cluster images includes:

[0016] Randomly dividing the customer behavior image into data subsets;

[0017] Run the clustering algorithm on each data subset to obtain the cluster corresponding to each data subset;

[0018] Hierarchical clustering is performed on each data subset according to the cluster to obtain hierarchical clustering images.

[0019] Furthermore, performing hierarchical clustering on each data subset according to the cluster to obtain each hierarchical clustering image includes:

[0020] Performing hierarchical clustering on each data subset according to a pre-set hierarchy and the cluster to obtain a new cluster;

[0021] In each new cluster, each representative point moves toward the center of the new cluster to obtain the shape and movement characteristics of the new cluster;

[0022] A hierarchical clustering image is determined based on the shape and motion characteristics.

[0023] Furthermore, matching the cluster image according to a predetermined template image to obtain the customer initial positioning image includes:

[0024] Positioning each pixel point obtained through matrix transformation in the cluster image, and determining a feature point corresponding to each pixel point and a main direction of the feature point;

[0025] Determine a descriptor of the feature point according to the main direction;

[0026] The cluster images are matched according to the descriptors corresponding to the cluster images and a predetermined template image to obtain a customer initial positioning image; the template image is the center image of each cluster.

[0027] Furthermore, the customer's initial positioning image is subjected to mismatch point elimination to obtain the customer's target positioning image, including:

[0028] Calculating the projection error between the customer's initial positioning image and a pre-established mismatch point elimination model;

[0029] The optimal interior point set is updated according to the preset iteration rounds and the projection error to obtain the customer target positioning image.

[0030] Furthermore, the steps of pre-establishing the mismatch point elimination model include:

[0031] Select four data points from the customer's initial positioning image for fitting to obtain a parameter model;

[0032] Calculating the distance between the remaining pixels in the client's initial positioning image and the parameter model;

[0033] Dividing the pixel points into inside points and outside points according to a preset distance threshold and the distance;

[0034] The step of iteratively dividing the in-zone points and out-zone points is performed, and the parameter model with the most in-zone points is used as a mismatch point elimination model.

[0035] In a second aspect, the present application provides a target positioning device in a mobile scenario, comprising:

[0036] A cluster image generating unit is used to cluster the acquired customer behavior images of the business handling process according to the business environment to obtain a cluster image;

[0037] An initial positioning image generating unit, used to match the cluster image according to a predetermined template image to obtain a customer initial positioning image;

[0038] The target positioning image generating unit is used to remove mismatched points from the customer initial positioning image to obtain the customer target positioning image.

[0039] Furthermore, the cluster image generating unit comprises:

[0040] A subset division module, used for randomly dividing the customer behavior image into data subsets;

[0041] A cluster determination module is used to run a clustering algorithm on each data subset to obtain a cluster corresponding to each data subset;

[0042] The cluster image generation module is used to cluster each data subset according to the cluster to obtain a cluster image.

[0043] Furthermore, the cluster image generating unit further includes:

[0044] The subset division module is also used to randomly divide the customer behavior image into data subsets;

[0045] The cluster determination module is also used to run a clustering algorithm on each data subset to obtain a cluster corresponding to each data subset;

[0046] The hierarchical clustering image generation module is used to perform hierarchical clustering on each data subset according to the cluster to obtain each hierarchical clustering image.

[0047] Furthermore, the hierarchical clustering image generation module comprises:

[0048] A new cluster determination submodule, used for performing hierarchical clustering on each data subset according to a preset hierarchy and the cluster to obtain a new cluster;

[0049] A shape movement characteristic determination submodule, used for moving each representative point in each new cluster toward the center of the new cluster to obtain the shape and movement characteristic of the new cluster;

[0050] The hierarchical clustering image generation submodule is used to determine the hierarchical clustering image according to the shape and movement characteristics.

[0051] Furthermore, the initial positioning image generating unit comprises:

[0052] A feature point main direction determination module is used to locate each pixel point obtained through matrix transformation in the cluster image, and determine the feature point corresponding to each pixel point and the main direction of the feature point;

[0053] A descriptor determination module, used to determine the descriptor of the feature point according to the main direction;

[0054] The initial positioning image generation module is used to match the cluster image according to the descriptor corresponding to the cluster image and a predetermined template image to obtain the customer's initial positioning image; the template image is the center image of each cluster.

[0055] Furthermore, the target positioning image generating unit comprises:

[0056] A projection error calculation module, used to calculate the projection error between the customer's initial positioning image and a pre-established mismatch point elimination model;

[0057] The customer target positioning image generation module is used to update the optimal interior point set according to a preset iteration round and the projection error to obtain the customer target positioning image.

[0058] Furthermore, the target positioning device in the mobile scene also includes:

[0059] A parameter model determination unit, used for selecting four data points from the customer's initial positioning image for fitting to obtain a parameter model;

[0060] A distance calculation unit, used to calculate the distance between the remaining pixels in the customer's initial positioning image and the parameter model;

[0061] An inside-outside distinction unit, used for dividing the pixel points into inside-outside points and outside-outside points according to a preset distance threshold and the distance;

[0062] The elimination model determination unit is used for iteratively dividing the steps of the internal points and the external points, and taking the parameter model with the most internal points as the mismatch point elimination model.

[0063] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the target positioning method in the mobile scenario when executing the program.

[0064] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the target positioning method in the mobile scenario.

[0065] In response to the problems in the prior art, the target positioning method and device in mobile scenarios provided by the present application can combine clustering algorithms with image matching, collect customer behavior images of customers in the business processing process in multiple scenarios, and perform clustering analysis on them, and perform correlation matching on the clustered images, thereby reducing the false matching rate and improving the accuracy of customer behavior pattern analysis, so as to provide customers with more suitable products and services and improve customer satisfaction on the premise of understanding customer behavior patterns. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0067] Figure 1 This is a flow chart of a method for locating a target in a mobile scenario in an embodiment of the present application;

[0068] Figure 2 One of the flow charts for obtaining a cluster image in an embodiment of the present application;

[0069] Figure 3 This is the second flow chart of obtaining a cluster image in an embodiment of the present application;

[0070] Figure 4A flowchart of obtaining each hierarchical clustering image in an embodiment of the present application;

[0071] Figure 5 To obtain the initial positioning image of the customer in the embodiment of the present application;

[0072] Figure 6 To obtain a customer target positioning image in the embodiment of the present application;

[0073] Figure 7 This is a flow chart of a mismatch point elimination model pre-established in an embodiment of the present application;

[0074] Figure 8 This is one of the structural diagrams of the target positioning device in the mobile scenario in the embodiment of the present application;

[0075] Fig. 9 This is one of the structural diagrams of the cluster image generation unit in the embodiment of the present application;

[0076] Fig.10 This is the second structural diagram of the cluster image generation unit in the embodiment of the present application;

[0077] Fig.11 This is a structural diagram of a hierarchical clustering image generation module in an embodiment of the present application;

[0078] Fig.12 This is a structural diagram of an initial positioning image generation unit in an embodiment of the present application;

[0079] Fig.13 This is a structural diagram of a target positioning image generation unit in an embodiment of the present application;

[0080] Fig.14 This is the second structural diagram of the target positioning device in the mobile scenario in the embodiment of the present application;

[0081] Fig.15 A schematic diagram of the structure of an electronic device in an embodiment of the present application;

[0082] Fig.16 This is an overall architecture diagram of a target positioning model based on cluster analysis and differential image matching in an embodiment of the present application;

[0083] Fig.17 A schematic diagram of scale space construction in an embodiment of the present application;

[0084] Fig.18 A schematic diagram of sample selection and model establishment in an embodiment of the present application. DETAILED DESCRIPTION

[0085] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0086] It should be noted that the target positioning method and device in the mobile scenario provided by the present application can be used in the financial field, and can also be used in any field other than the financial field. The application field of the target positioning method and device in the mobile scenario provided by the present application is not limited.

[0087] See also Figure 1 In order to cluster the acquired customer behavior images of the business handling process according to the business environment and finally obtain the customer target positioning image, the present application provides a target positioning method in a mobile scenario, including:

[0088] S101: clustering the acquired customer behavior images of the business handling process according to the business environment to obtain a cluster image;

[0089] It should be noted that the target positioning method in the mobile scenario provided in this application is based on the Cure (Clustering Using Representative) clustering algorithm, the Surf (Speeded Up Robust Feature) image matching algorithm and the RANSAC (Random Sample Consensus) random sampling consensus algorithm. It is suitable for a variety of target movement scenarios and can locate targets in target movement scenarios.

[0090] Specifically, in one embodiment, the method provided in the present application can be used to target (customer) positioning of customers who conduct banking business in bank branches, so that business personnel can subsequently analyze customer behavior patterns based on positioning information.

[0091] Before targeting customers who come to a bank branch to handle business within a certain period of time, it is first necessary to collect customer behavior images corresponding to the customers who handle business at the bank branch within the period of time. The customer behavior image can be each frame in the collected continuous video image. Clustering algorithm and image matching algorithm can be used to process each frame in the continuous video image to obtain the target positioning information of the customer within the period of time. Then, the business personnel can analyze their movement and behavior, and transmit the obtained movement and behavior as customer behavior information to the bank business backend system, so that the business personnel can provide targeted business handling services to the customer after understanding the customer's behavior information, thereby improving the efficiency and satisfaction of customer business handling.

[0092] Specifically, after obtaining the customer behavior image of the business handling process, the Cure clustering algorithm can be used to cluster the customer according to the business environment in which the customer handles the business, and obtain a cluster image. The so-called business environment refers to the surrounding environment of the customer when the customer handles the business in the branch hall. Clustering customer behavior images according to the business environment can make customer behavior images with similar or related business environments clustered into one category, and then complete image matching in this category, so as to achieve the purpose of fast matching and accurate matching.

[0093] S102: Matching the cluster image according to a predetermined template image to obtain an initial positioning image of the customer;

[0094] It is understandable that the embodiment of the present application uses the Surf image matching algorithm to complete cluster image matching. The Surf image matching algorithm is a feature-based image matching algorithm. The algorithm can complete feature point extraction by constructing a scale space, and then use the extracted feature points to perform feature matching on the clustered customer behavior image, thereby completing the initial customer positioning. The so-called initial customer positioning refers to the process of obtaining a customer positioning image without removing the mismatched points.

[0095] It should be noted that the common grayscale-based image matching algorithm is computationally complex and has low real-time performance. The image matching algorithm used in the embodiment of the present application has a good behavior description effect for locating multiple targets in an image, and can achieve matching of each frame image in a continuous image, which is conducive to finally obtaining the required target behavior information from the matching information.

[0096] S103: Eliminate mismatched points from the customer's initial positioning image to obtain the customer's target positioning image.

[0097] It is understandable that the embodiment of the present application uses the RANSAC (Random Sample Consensus) algorithm to eliminate mismatched points. The so-called mismatched points refer to points that obviously should not belong to a certain cluster but are included in a certain cluster.

[0098] Specifically, although the customer's initial positioning image can be obtained through the above image matching steps, due to the noise that may exist in the image acquisition process, errors may occur in the feature point selection process, resulting in inaccurate image matching. Therefore, it is necessary to use the RANSAC algorithm to eliminate the mismatched points, ensure the accuracy of the matching, and finally obtain the customer's target positioning image.

[0099] In summary, see Fig.16For a large number of images collected in a continuous time period, the embodiment of the present application can use the Cure clustering algorithm to cluster the images. The scale space is established and the feature points are extracted using the Surf algorithm for the images on each clustering layer. Image matching is performed in sequence according to the order in which the images are collected for the feature points extracted from each image. In this process, the RANSAC algorithm is applied to the matched feature points, and the mismatched points can be eliminated through the creation of a parameter model to obtain an accurate matching result, and finally the movement and behavior information of the target in the continuous time period are obtained.

[0100] From the above description, it can be seen that the target positioning method in mobile scenarios provided by the present application can combine clustering algorithms with image matching, collect customer behavior images of customers in the business processing process in multiple scenarios, and perform clustering analysis on them, and perform correlation matching on the clustered images, thereby reducing the false matching rate and improving the accuracy of customer behavior pattern analysis, so as to provide customers with more suitable products and services on the premise of understanding customer behavior patterns and improve customer satisfaction.

[0101] See also Figure 2 , cluster the acquired customer behavior images of the business handling process according to the business environment to obtain cluster images, including:

[0102] S201: Randomly divide the customer behavior image into data subsets;

[0103] It is understandable that the embodiment of the present application uses the Cure clustering algorithm to complete the clustering of customer behavior images. The Cure clustering algorithm can select a fixed number of representative points in the data space and cluster these points into one category. Usually, each cluster includes more than one representative point, so the Cure clustering algorithm is better at processing isolated points.

[0104] Considering that the customer behavior images collected in actual business scenarios may be frames in continuous video images, the number is large. In order to facilitate subsequent clustering processing, it is necessary to randomly divide the customer behavior images into data subsets.

[0105] Subset division can be performed in the following way: a random sample S is extracted from the collected customer behavior images at regular intervals as a subset. For example, every 5 minutes can be set as a time interval, and the image collection frequency can also be limited to 3 to 5 times per second.

[0106] It should be noted that most clustering algorithms are often fragile when dealing with clusters of similar size or samples with isolated points. The embodiment of the present application adopts the Cure clustering algorithm to cluster customer behavior images. Compared with the centroid-based strategy and the representative object-based strategy, the Cure clustering algorithm chooses an intermediate strategy, taking into account both the centroid-based strategy and the representative object-based strategy, ensuring the reliability of clustering.

[0107] S202: Run the Cure clustering algorithm on each data subset to obtain a cluster corresponding to each data subset;

[0108] It is understandable that the purpose of dividing subsets is to facilitate the determination of clusters in clustering. When processing large amounts of information, the data partitioning algorithm can be used to divide the sample into several parts, and then the objects in each part are locally clustered to form subsets. This ensures the efficiency and reliability of the clustering process. By running the Cure clustering algorithm on each data subset, the clusters corresponding to each data subset can be obtained.

[0109] For example, the customer behavior images acquired within the above time interval of 5 minutes are randomly sampled to obtain a subset, and then the subset is randomly divided into N parts using the data segmentation algorithm. The value of N depends on the size of the subset. For example, when the number of images in the subset is 1000, the value of N is 4.

[0110] S203: Clustering each data subset according to the cluster to obtain a cluster image.

[0111] It is understandable that the Cure clustering algorithm regards each data point as a cluster, and combines the closest clusters until the number of clusters reaches the preset K. Cluster is a mathematical concept, referring to a group of points with similar properties. In the embodiment of the present application, a cluster refers to a collection of pixel points with similar attributes. Clustering each data subset according to the cluster can obtain a cluster image. The obtained cluster image can be understood as multiple groups of images with similar or related environmental backgrounds. The customer surrounding environment displayed by each group of images is similar or related.

[0112] It can be seen from the above description that the target positioning method in a mobile scenario provided by the present application can cluster the acquired customer behavior images of the business handling process according to the business environment to obtain a clustered image.

[0113] See also Figure 3 , cluster the acquired customer behavior images of the business handling process according to the business environment to obtain cluster images, including:

[0114] S301: Randomly divide the customer behavior image into data subsets;

[0115] S302: running a clustering algorithm on each data subset to obtain a cluster corresponding to each data subset;

[0116] S303: Perform hierarchical clustering on each data subset according to the cluster to obtain hierarchical clustering images.

[0117] It is understandable that the Cure clustering algorithm can be used to perform hierarchical clustering on customer behavior images. Hierarchical clustering can be understood as dividing the similarities of all customer behavior images into multiple levels, and clustering customer behavior images with the same similarity. Different from S203, after completing S301 (that is, S201) and S302 (that is, S202), this embodiment needs to obtain hierarchical clustering images of all customer behavior images through continuous sample extraction and cluster analysis.

[0118] Hierarchical clustering is beneficial for matching clustered images according to predetermined template images, improving the accuracy of matching results, and finally obtaining the customer's initial positioning image, which can largely ensure the real-time matching in actual business scenarios.

[0119] In summary, in actual business scenarios, a large number of customer behavior images collected over a period of time can be clustered using the Cure hierarchical clustering algorithm according to different pre-divided levels. Each level is a collection of similar images. Through this processing step, the orderly stratification of customer behavior images can be achieved, which improves the matching efficiency of the subsequent image matching step.

[0120] From the above description, it can be seen that the target positioning method in the mobile scenario provided by the present application clusters the acquired customer behavior images of the business handling process according to the business environment to obtain a hierarchical clustering image.

[0121] See also Figure 4 , hierarchical clustering is performed on each data subset according to the cluster to obtain hierarchical clustering images, including:

[0122] S401: performing hierarchical clustering on each data subset according to pre-set levels and clusters to obtain new clusters;

[0123] S402: moving each representative point of the hierarchical clustering image in the new cluster toward the center of the new cluster to obtain the shape and movement characteristics of the new cluster;

[0124] S403: Determine a hierarchical clustering image according to the shape and movement characteristics.

[0125] It should be noted that the so-called pre-set levels are divided according to the different similarities as described above, and the specific division levels can be set according to the actual business scenarios.

[0126] After completing the hierarchical clustering of the images, outliers in each data subset can also be removed to eliminate the influence of outliers and enhance the accuracy of subsequent image matching.

[0127] The Cure clustering algorithm can be divided into two steps to eliminate the influence of outliers:

[0128] The first step is coarse screening, that is, removing outliers that are relatively far from the normal value in the coordinate system to complete the removal of outliers. If not removed, outliers will cause the number of objects in the cluster to increase slowly or even not increase, which is not conducive to clustering.

[0129] The second step is fine screening. When it comes to this step, since the proportion of outliers is relatively small and the size of outliers is very different from that of normal clusters, they can be easily identified and removed after identification.

[0130] It should also be noted that the "representative points" in this embodiment refer to the pixels in each image that are representative of the local image information. The "shape and movement characteristics" refer to the shape formed by the aggregation of representative pixels in the collected customer behavior image, as well as how much they move in the x and y directions when moving toward the center of the cluster, and the angle of movement, which constitutes its movement characteristics. Through the movement of the points in the cluster, after reaching stability, the level at which these points are located indicates the level of the image at which these points are located.

[0131] According to the distance of each point from the cluster center, that is, the distance in the x-direction and y-direction, the farthest point can be obtained. Compared with other points, the calculated distance of these points is very different, so they should be removed as outliers.

[0132] In summary, the collected customer behavior images are divided into subsets, and each clustering operation will remove outliers, that is, abnormal values, to ensure the accuracy of clustering.

[0133] It can be seen from the above description that the target positioning method in a mobile scenario provided by the present application can hierarchically cluster each data subset according to the cluster to obtain each hierarchical cluster image.

[0134] See also Figure 5 , matching the cluster image according to a predetermined template image to obtain a customer initial positioning image, including:

[0135] S501: accurately positioning each pixel point obtained through matrix transformation in the cluster image, and determining the feature point corresponding to each pixel point and the main direction of the feature point;

[0136] It can be understood that this step is to match the cluster images of each level after cluster analysis, and the matching object is still the customer behavior image collected at the outlet within a period of time. The following steps are all carried out in accordance with the steps specified by the Surf algorithm, and the corresponding actual scenario is to extract and match the feature points of the customer behavior image.

[0137] First, see Fig.17 , an inverted pyramid scale space is established through the Hessian Matrix. The construction characteristics of this scale space are: the size of the box filter is gradually increased, the filter size used for the same group of images is the same, and the scale space factor A is continuously increased. In this process, the image size remains unchanged.

[0138]

[0139] det(H approx )=L xx L yy -(0.9L xy ) 2 (3-2)

[0140] According to formula (3-2), the transformation graph of the image is approximated by the determinant of the Hessian matrix, that is, a matrix transformation is performed on the clustered image.

[0141] Since the Surf algorithm uses a box filter, in order to balance the error, the empirical value can be set to 0.9, as shown in formula (3-2). Where L is the pixel size, L xx is the product of pixels in the x direction, i.e. L xx =L x ×L x , L xy , L yy Definition and L xx Similarly, det(H) refers to the determinant of the matrix H.

[0142] After the previous step, each pixel point is numerically compared with the pixel points in the two adjacent areas above and below it to determine its maximum value in direction and accurately locate the feature point. That is, each pixel point is compared with the 26 points in its adjacent scale space, and the points with shorter direction vectors in their respective scale spaces are filtered. Here, 26 points are the fixed number of points taken by the Surf algorithm, 9 points in the upper neighborhood, 9 points in the lower neighborhood, and 8 points around the scale level where it is located, for a total of 9+9+8=26 points.

[0143] Taking the feature point obtained in the previous step as the center, calculate the sum of the responses of all pixels within the 60° sector in the x and y directions within the range of the feature point to form a new vector. Perform the above operation on each pixel within the circle within the range of the feature point to obtain multiple vector values, and use the maximum value of these vectors as the main direction of the feature point.

[0144] S502: Determine the descriptor of the feature point according to the main direction;

[0145] It can be understood that a square frame is selected around the feature point, and the frame is divided into 16 sub-regions. In each sub-region, the Haar wavelet of 25 pixels in the horizontal and vertical directions is counted. The sum of the horizontal vector, the sum of the vertical vector, the sum of the horizontal absolute value, and the sum of the vertical absolute value are calculated. These four direction values ​​are used as the description information of the feature point, that is, the descriptor.

[0146] S503: According to the descriptors corresponding to the cluster images, the central image of each cluster is used as a template image to match the cluster images to obtain the initial positioning image of the customer.

[0147] It is understandable that as a basic operation of image matching, the establishment of scale space will affect the generation of feature descriptors. Since the matching between images is the matching of feature descriptors between different images, the establishment of scale space will greatly affect the accuracy of matching.

[0148] After obtaining the feature descriptor of each image, the image to be matched can be matched with the template image for feature points. The so-called template image is the central image of each cluster as the template image in the embodiment of the present application.

[0149] After feature point matching, the image to be matched in each clustering layer establishes a corresponding relationship with the template image, and the corresponding relationship between the pixels of each two images is marked at the time of image acquisition. Relative to the movement of the target at the time of template image acquisition, the motion information of the target in continuous time can be obtained by outputting the matching information of each feature point at each level, thereby extracting the behavior of the target point.

[0150] The so-called image acquisition time means, for example, image A is acquired at time t, and image B is acquired at time t+t1. The correspondence between these two images indicates the movement of the target point relative to time t when the image is acquired at time t+t1, that is, the movement of the target point in the time period t1.

[0151] After image matching, the matching point information of the image at each level can be obtained. These matching point information indicates the changes in the behavior trajectory and behavior mode of the customers in the branch during the period of image acquisition when they are handling business and waiting, and represents the behavior pattern of each customer during this period of time.

[0152] It can be seen from the above description that the target positioning method in a mobile scenario provided by the present application can match the cluster image according to a predetermined template image to obtain the customer's initial positioning image.

[0153] See also Figure 6 , remove the mismatched points from the customer's initial positioning image to obtain the customer's target positioning image, including:

[0154] S601: Calculating the projection error between the customer's initial positioning image and the pre-established mismatch point elimination model;

[0155] S602: Update the optimal internal point set according to the preset iteration rounds and projection error to obtain the customer target positioning image.

[0156] It is understandable that due to the influence of the business environment presented by the customer behavior image, the initial matching results will inevitably have errors. Therefore, eliminating the wrong matching points to ensure the accuracy of the matching is crucial to ultimately obtain a more accurate matching result.

[0157] See also Fig.18 , the embodiment of the present application adopts the RANSAC (Random Sample Consensus) random sampling consensus algorithm to eliminate mismatched points. The RANSAC algorithm assumes that the distribution of data can be explained by some model parameters, which are called "in-line points". If a set of "in-line points" is given, a parameter model can be obtained, and then the "in-line points" in the observed data that satisfy the parameter model can be found. Based on this idea, it can be determined whether there are matching errors in the matched points, thereby eliminating mismatched points and improving the matching accuracy.

[0158] Specifically, 4 sample data are randomly selected from the matching point data set obtained in the previous step, and the transformation matrix is ​​calculated, which is recorded as model M; then, the projection error between all data in the data set and model M is calculated; finally, the optimal internal point set is updated according to the projection error value, and the updating process is iterated for each data set until the preset iteration round is reached.

[0159] The RANSAC algorithm can be used to identify mismatched points and improve the accuracy of the matching process, thereby obtaining more accurate information on changes in customer behavior during that period of time.

[0160] It can be seen from the above description that the target positioning method in a mobile scenario provided by the present application can remove mismatched points from the customer's initial positioning image to obtain the customer's target positioning image.

[0161] See also Figure 7 , the steps of pre-establishing the mismatch point elimination model include:

[0162] S701: Select four data points from the customer's initial positioning image for fitting to obtain a parameter model;

[0163] S702: Calculate the distance between the remaining pixels in the client's initial positioning image and the parameter model;

[0164] S703: Classifying pixel points into internal points and external points according to a preset distance threshold;

[0165] S704: Iteratively execute the parameter model with the most points in the local area as the mismatch point elimination model.

[0166] It can be understood that the embodiment of the present application randomly selects at least four non-collinear data points from the matching point data set, calculates the homography matrix, and obtains the model M. The parameter model M can be understood as a threshold that defines "inside points" and "outside points", and the "outside points" will eventually be eliminated; then, the projection distance of the remaining data points in the matching point data set from the parameter M is calculated. Iterate the preset rounds K times, take the model with the most "inside points" as the fitting result, delete the "outside points", and determine the parameter model M.

[0167] It should be noted that the homography matrix corresponding to the four points can be obtained by using the selected four points. This matrix is ​​the mathematical form of the parameter model M. The homography matrix refers to the projection matrix from one plane to another plane, and is a commonly used matrix for image correction in computer vision. From the above description, it can be seen that the target positioning method in a mobile scene provided by the present application can pre-establish a mismatch point removal model.

[0168] By analyzing customer behavior patterns, we can scientifically conduct specific customer demand analysis, thereby accurately identifying customer needs, quantifying customer sensitivity, supporting targeted and refined customer service strategies, effectively controlling offline service labor costs, and enhancing the bank's public image.

[0169] Based on the same inventive concept, the embodiments of the present application also provide a target positioning device in a mobile scenario, which can be used to implement the method described in the above embodiments, as described in the following embodiments. Since the principle of solving the problem by the target positioning device in the mobile scenario is similar to the target positioning method in the mobile scenario, the implementation of the target positioning device in the mobile scenario can refer to the implementation of the method based on the software performance benchmark determination, and the repetitions will not be repeated. As used below, the term "unit" or "module" can implement a combination of software and / or hardware for a predetermined function. Although the system described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0170] See also Figure 8 In order to cluster the acquired customer behavior images of the business handling process according to the business environment and finally obtain the customer target positioning image, the present application provides a target positioning device in a mobile scenario, including:

[0171] The cluster image generation unit 801 is used to cluster the acquired customer behavior images of the business handling process according to the business environment to obtain a cluster image;

[0172] The initial positioning image generating unit 802 is used to match the cluster image according to a predetermined template image to obtain a customer initial positioning image;

[0173] The target positioning image generating unit 803 is used to remove mismatched points from the customer initial positioning image to obtain a customer target positioning image.

[0174] See also Fig. 9 , the cluster image generating unit 801 comprises:

[0175] A subset division module 901 is used to randomly divide the customer behavior image into data subsets;

[0176] A cluster determination module 902 is used to run a clustering algorithm on each data subset to obtain a cluster corresponding to each data subset;

[0177] The cluster image generation module 903 is used to cluster each data subset according to the clusters to obtain a cluster image.

[0178] See also Fig.10 , the cluster image generating unit 801 further includes:

[0179] The subset division module 1001 is also used to randomly divide the customer behavior image into data subsets;

[0180] The cluster determination module 1002 is further used to run a clustering algorithm on each data subset to obtain a cluster corresponding to each data subset;

[0181] The hierarchical clustering image generation module 1003 is used to perform hierarchical clustering on each data subset according to the clusters to obtain hierarchical clustering images.

[0182] See also Fig.11 , the hierarchical clustering image generation module 1003 comprises:

[0183] A new cluster determination submodule 1101 is used to perform hierarchical clustering on each data subset according to a preset hierarchy and the cluster to obtain a new cluster;

[0184] A shape movement characteristic determination submodule 1102 is used to move each representative point of the hierarchical clustering image in the new cluster toward the center of the new cluster to obtain the shape and movement characteristic of the new cluster;

[0185] The hierarchical clustering image generation submodule 1103 is used to determine the hierarchical clustering image according to the shape and movement characteristics.

[0186] See also Fig.12 , the initial positioning image generating unit 802 comprises:

[0187] The feature point main direction determination module 1201 is used to accurately locate each pixel point obtained through matrix transformation in the cluster image, and determine the feature point corresponding to each pixel point and the main direction of the feature point;

[0188] A descriptor determination module 1202, configured to determine a descriptor of the feature point according to the main direction;

[0189] The initial positioning image generation module 1203 is used to match the cluster image according to the descriptor corresponding to the cluster image and a predetermined template image to obtain the customer's initial positioning image; the template image is the center image of each cluster.

[0190] See also Fig.13 , the target positioning image generating unit 803 comprises:

[0191] The projection error calculation module 1301 is used to calculate the projection error between the customer's initial positioning image and the pre-established mismatch point elimination model;

[0192] The customer target positioning image generation module 1302 is used to update the optimal interior point set according to a preset iteration round and the projection error to obtain the customer target positioning image.

[0193] See also Fig.14 , the target positioning device in the mobile scene also includes:

[0194] A parameter model determination unit 1401 is used to select four data points from the customer's initial positioning image for fitting to obtain a parameter model;

[0195] The distance calculation unit 1402 is used to calculate the distance between the remaining pixels in the client's initial positioning image and the parameter model;

[0196] An inside-outside distinction unit 1403, configured to divide the pixel points into inside-outside points and outside-outside points according to a preset distance threshold;

[0197] The elimination model determination unit 1404 is used to iteratively divide the steps of the intra-office points and the extra-office points, and use the parameter model with the most intra-office points as the mismatch point elimination model.

[0198] From the hardware level, in order to cluster the acquired customer behavior images of the business handling process according to the business environment and finally obtain the customer target positioning image, the present application provides an embodiment of an electronic device for implementing all or part of the content of the target positioning method in the mobile scenario, and the electronic device specifically includes the following content:

[0199] Processor, memory, communication interface and bus; wherein the processor, memory and communication interface communicate with each other through the bus; the communication interface is used to realize information transmission between the target positioning device in the mobile scenario and related devices such as the core business system, user terminal and related database; the logic controller can be a desktop computer, a tablet computer and a mobile terminal, etc., but the present embodiment is not limited thereto. In the present embodiment, the logic controller can be implemented with reference to the embodiment of the target positioning method in the mobile scenario in the embodiment, and the embodiment of the target positioning device in the mobile scenario, and the contents thereof are incorporated herein, and the repeated parts are not repeated.

[0200] It is understandable that the user terminal may include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, a smart wearable device, etc. Among them, the smart wearable device may include smart glasses, a smart watch, a smart bracelet, etc.

[0201] In practical applications, part of the target positioning method in a mobile scenario can be executed on the electronic device side as described above, or all operations can be completed in the client device. The specific selection can be based on the processing capability of the client device and the limitations of the user's usage scenario. This application does not limit this. If all operations are completed in the client device, the client device may also include a processor.

[0202] The client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side, and other implementation scenarios may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, or a server cluster consisting of multiple servers, or a server structure of a distributed device.

[0203] Fig.15 FIG. 9 is a schematic block diagram of the system structure of the electronic device 9600 according to an embodiment of the present application. Fig.15 As shown, the electronic device 9600 may include a central processor 9100 and a memory 9140; the memory 9140 is coupled to the central processor 9100. It is worth noting that Fig.15 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0204] In one embodiment, the target positioning method function in a mobile scenario may be integrated into the central processing unit 9100. The central processing unit 9100 may be configured to perform the following control:

[0205] S101: clustering the acquired customer behavior images of the business handling process according to the business environment to obtain a cluster image;

[0206] S102: Matching the cluster image according to a predetermined template image to obtain an initial positioning image of the customer;

[0207] S103: Eliminate mismatched points from the customer's initial positioning image to obtain the customer's target positioning image.

[0208] From the above description, it can be seen that the target positioning method in mobile scenarios provided by the present application can combine clustering algorithms with image matching, collect customer behavior images of customers in the business processing process in multiple scenarios, and perform clustering analysis on them, and perform correlation matching on the clustered images, thereby reducing the false matching rate and improving the accuracy of customer behavior pattern analysis, so as to provide customers with more suitable products and services on the premise of understanding customer behavior patterns and improve customer satisfaction.

[0209] In another embodiment, the target positioning device in the mobile scenario can be configured separately from the central processing unit 9100. For example, the target positioning device in the mobile scenario of the data composite transmission device can be configured as a chip connected to the central processing unit 9100, and the function of the target positioning method in the mobile scenario can be realized through the control of the central processing unit.

[0210] like Fig.15 As shown, the electronic device 9600 may also include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily have to include Fig.15 In addition, the electronic device 9600 may also include Fig.15 For components not shown, reference may be made to the prior art.

[0211] like Fig.15 As shown, the central processing unit 9100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor device and / or logic device. The central processing unit 9100 receives input and controls the operation of various components of the electronic device 9600.

[0212] The memory 9140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The above-mentioned information related to the failure may be stored, and a program for executing the relevant information may also be stored. The CPU 9100 may execute the program stored in the memory 9140 to implement information storage or processing, etc.

[0213] The input unit 9120 provides input to the central processing unit 9100. The input unit 9120 is, for example, a key or a touch input device. The power supply 9170 is used to provide power to the electronic device 9600. The display 9160 is used to display display objects such as images and texts. The display may be, for example, an LCD display, but is not limited thereto.

[0214] The memory 9140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 9140 may also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142, which is used to store application programs and function programs or processes for executing the operation of the electronic device 9600 through the central processor 9100.

[0215] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0216] The communication module 9110 is a transmitter / receiver 9110 that sends and receives signals via an antenna 9111. The communication module (transmitter / receiver) 9110 is coupled to the central processor 9100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.

[0217] Based on different communication technologies, multiple communication modules 9110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless LAN module. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide an audio output via the speaker 9131 and receive an audio input from the microphone 9132, thereby realizing a common telecommunication function. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is also coupled to the central processor 9100, so that recording can be performed on the local machine through the microphone 9132, and the sound stored on the local machine can be played through the speaker 9131.

[0218] The embodiments of the present application also provide a computer-readable storage medium capable of implementing all steps of the target positioning method in a mobile scenario in which the execution subject is a server or a client in the above embodiments. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, all steps of the target positioning method in a mobile scenario in which the execution subject is a server or a client in the above embodiments are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0219] S101: clustering the acquired customer behavior images of the business handling process according to the business environment to obtain a cluster image;

[0220] S102: Matching the cluster image according to a predetermined template image to obtain an initial positioning image of the customer;

[0221] S103: Eliminate mismatched points from the customer's initial positioning image to obtain the customer's target positioning image.

[0222] From the above description, it can be seen that the target positioning method in mobile scenarios provided by the present application can combine clustering algorithms with image matching, collect customer behavior images of customers in the business processing process in multiple scenarios, and perform clustering analysis on them, and perform correlation matching on the clustered images, thereby reducing the false matching rate and improving the accuracy of customer behavior pattern analysis, so as to provide customers with more suitable products and services on the premise of understanding customer behavior patterns and improve customer satisfaction.

[0223] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0224] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0225] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0226] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0227] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A target positioning method in a mobile scene, characterized in that: include: Clustering the acquired customer behavior images of the business handling process according to the business environment to obtain a cluster image; Matching the cluster image according to a predetermined template image to obtain an initial positioning image of the customer; Eliminate mismatched points from the customer's initial positioning image to obtain a customer's target positioning image; The step of clustering the acquired customer behavior images of the business handling process according to the business environment to obtain cluster images includes: Randomly dividing the customer behavior image into data subsets; Run the clustering algorithm on each data subset to obtain the cluster corresponding to each data subset; Clustering each data subset according to the cluster to obtain a cluster image; The step of matching the cluster image with a predetermined template image to obtain an initial positioning image of the customer includes: Positioning each pixel point obtained through matrix transformation in the cluster image, and determining a feature point corresponding to each pixel point and a main direction of the feature point; Determine a descriptor of the feature point according to the main direction; Matching the cluster images according to the descriptors corresponding to the cluster images and a predetermined template image to obtain an initial positioning image for the customer; the template image is the central image of each cluster; The step of removing mismatched points from the customer's initial positioning image to obtain the customer's target positioning image includes: Calculating the projection error between the customer's initial positioning image and a pre-established mismatch point elimination model; The optimal interior point set is updated according to the preset iteration rounds and the projection error to obtain the customer target positioning image.

2. The target positioning method in a mobile scenario according to claim 1, characterized in that: The step of clustering the acquired customer behavior images of the business handling process according to the business environment to obtain cluster images includes: Randomly dividing the customer behavior image into data subsets; Run the clustering algorithm on each data subset to obtain the cluster corresponding to each data subset; Hierarchical clustering is performed on each data subset according to the cluster to obtain hierarchical clustering images.

3. The target positioning method in a mobile scenario according to claim 2, characterized in that: The step of performing hierarchical clustering on each data subset according to the cluster to obtain each hierarchical clustering image comprises: Performing hierarchical clustering on each data subset according to a pre-set hierarchy and the cluster to obtain a new cluster; Moving each representative point of the hierarchical clustering image in the new cluster toward the center of the new cluster to obtain the shape and movement characteristics of the new cluster; A hierarchical clustering image is determined based on the shape and motion characteristics.

4. The target positioning method in a mobile scenario according to claim 1, characterized in that: The steps of pre-establishing a mismatch point elimination model include: Select four data points from the customer's initial positioning image for fitting to obtain a parameter model; Calculating the distance between the remaining pixels in the client's initial positioning image and the parameter model; Dividing the pixel points into inside points and outside points according to a preset distance threshold and the distance; The step of iteratively dividing the in-zone points and out-zone points is performed, and the parameter model with the most in-zone points is used as a mismatch point elimination model.

5. A target positioning device in a mobile scene, characterized in that: include: A cluster image generating unit is used to cluster the acquired customer behavior images of the business handling process according to the business environment to obtain a cluster image; An initial positioning image generating unit, used to match the cluster image according to a predetermined template image to obtain a customer initial positioning image; A target positioning image generating unit, used to remove mismatched points from the customer's initial positioning image to obtain a customer target positioning image; Wherein, the cluster image generating unit comprises: A subset division module, used for randomly dividing the customer behavior image into data subsets; A cluster determination module is used to run a clustering algorithm on each data subset to obtain a cluster corresponding to each data subset; A cluster image generation module, used for clustering each data subset according to the clusters to obtain a cluster image; Wherein, the initial positioning image generating unit comprises: A feature point main direction determination module is used to locate each pixel point obtained through matrix transformation in the cluster image, and determine the feature point corresponding to each pixel point and the main direction of the feature point; A descriptor determination module, used to determine the descriptor of the feature point according to the main direction; An initial positioning image generation module is used to match the cluster image according to the descriptor corresponding to the cluster image and a predetermined template image to obtain a customer initial positioning image; the template image is the center image of each cluster; Wherein, the target positioning image generating unit comprises: A projection error calculation module, used to calculate the projection error between the customer's initial positioning image and a pre-established mismatch point elimination model; The customer target positioning image generation module is used to update the optimal interior point set according to a preset iteration round and the projection error to obtain the customer target positioning image.

6. The target positioning device in a mobile scene according to claim 5, characterized in that: The cluster image generating unit further includes: The subset division module is also used to randomly divide the customer behavior image into data subsets; The cluster determination module is also used to run a clustering algorithm on each data subset to obtain a cluster corresponding to each data subset; The hierarchical clustering image generation module is used to perform hierarchical clustering on each data subset according to the cluster to obtain each hierarchical clustering image.

7. The target positioning device in a mobile scene according to claim 6, characterized in that: The hierarchical clustering image generation module comprises: A new cluster determination submodule, used for performing hierarchical clustering on each data subset according to a preset hierarchy and the cluster to obtain a new cluster; a shape movement characteristic determination submodule, used for moving each representative point of the hierarchical clustering image in the new cluster toward the center of the new cluster to obtain the shape and movement characteristic of the new cluster; The hierarchical clustering image generation submodule is used to determine the hierarchical clustering image according to the shape and movement characteristics.

8. The target positioning device in a mobile scene according to claim 5, characterized in that: Also includes: A parameter model determination unit, used for selecting four data points from the customer's initial positioning image for fitting to obtain a parameter model; A distance calculation unit, used to calculate the distance between the remaining pixels in the customer's initial positioning image and the parameter model; An inside-outside distinction unit, used for dividing the pixel points into inside-outside points and outside-outside points according to a preset distance threshold and the distance; The elimination model determination unit is used for iteratively dividing the steps of the internal points and the external points, and taking the parameter model with the most internal points as the mismatch point elimination model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the target positioning method in a mobile scenario described in any one of claims 1 to 4 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the target positioning method in a mobile scenario described in any one of claims 1 to 4 are implemented.

Images