Scenic spot visitor behavior detection clustering method, system, device and storage medium

Abstract

The application provides a scenic spot tourist behavior detection clustering method, system, device and storage medium, the method comprises the following steps: collecting pictures with geographical information based on a scenic spot; identifying the behavior of tourists in the pictures based on the positional relationship between the skeleton points, obtaining a behavior label, and establishing a mapping relationship between the behavior label and the geographical information of the picture; gridizing an electronic map of the scenic spot, and collecting the number of times of various behavior labels appearing in each grid; when the number of behavior labels in the grid meets the preset threshold value corresponding to the behavior label, the grid obtains the behavior label and generates an interest point on the electronic map; based on the matching of the historical behavior label of the tourist and the interest point in the electronic map, the interest point meeting the matching is provided to the tourist. The application can identify the behavior of tourists in the scenic spot, add interest points in various places of the scenic spot according to the behavior label of the tourist, optimize the scenic spot map, and also enhance the data-guided experience of the scenic spot.

Description

Technical Field

[0001] This invention relates to the field of digital scenic area management, and more specifically, to a method, system, device, and storage medium for detecting and clustering tourist behavior in scenic areas. Background Technology

[0002] With the development of the tourism industry, the number of tourism products has become increasingly diverse. Tourists also have higher expectations for their visitor experience. Scenic spots often provide maps with pre-defined facilities to guide visitors. However, for large national parks and forest parks, the maps may not be detailed enough, or there may not be any maps at all. Furthermore, typical scenic spot maps usually only mark facilities built by the scenic area, such as restrooms and convenience stores, but they lack any information or markings about areas suitable for picnics, fishing, kite flying, and other activities that tourists can flexibly arrange (non-facility activities). Scenic areas also cannot restrict the areas for these activities.

[0003] This results in tourists being unable to immediately know which areas of a large scenic area offer activities that match their interests (such as picnics, kite flying, fishing, etc.), which areas are more suitable for their preferred activities, or where they can meet like-minded tourists. Consequently, they are left to wander aimlessly, wasting their time and reducing their user experience.

[0004] Therefore, the present invention provides a method, system, device and storage medium for detecting and clustering tourist behavior in scenic areas. Summary of the Invention

[0005] To address the problems in existing technologies, the present invention aims to provide a method, system, device, and storage medium for detecting and clustering tourist behavior in scenic areas. This invention overcomes the difficulties of existing technologies, enables the identification of tourist behavior within scenic areas, adds points of interest to various locations within the scenic area based on tourist behavior tags, optimizes the scenic area map, enhances the data-driven guidance experience of the scenic area, and greatly improves the user's tour experience.

[0006] Embodiments of the present invention provide a clustering method for detecting tourist behavior in scenic areas, comprising the following steps:

[0007] Collect images with geographic information based on the scenic area;

[0008] The behavior of tourists in the image is identified based on the positional relationship between skeletal points, behavior labels are obtained, and a mapping relationship between the behavior labels and the geographic information of the image is established.

[0009] The electronic map of the scenic area is gridded, and the frequency of various behavior tags in each grid is collected.

[0010] When the number of occurrences of the behavior label in the grid meets the preset threshold corresponding to the behavior label, the grid obtains the behavior label;

[0011] At least based on the behavior labels of the grid, a point of interest is generated on the electronic map;

[0012] Collect tourists' current location information, match the tourists' historical behavior tags in various scenic spots with points of interest in the electronic map, and provide tourists with points of interest that match the information.

[0013] Preferably, the acquisition of images with geographic information based on the scenic area includes:

[0014] Photos of the scenic area were collected based on geographic information gathered via the network.

[0015] Collect photos taken by the surveillance cameras within the scenic area; and

[0016] Collect photos taken by drones or unmanned vehicles within the scenic area.

[0017] Preferably, the step of identifying tourist behavior in the image based on the positional relationship between skeletal points, obtaining behavior tags, and establishing a mapping relationship between the behavior tags and the geographic information of the image includes:

[0018] The MoveNet pre-trained model was used to extract 17 skeletal key points of the human body from the image.

[0019] The coordinates of the 17 skeletal key points are transformed, that is, their coordinates are transformed into a coordinate system with one of the key points as the origin;

[0020] The coordinate-transformed key points of the 17 human skeletons were input into a multilayer perceptron for training.

[0021] The behavior label with the highest confidence was obtained by converting the multilayer perceptron model to a TFLite model; and

[0022] Establish a mapping relationship between the behavior tags and the geographic information of the images.

[0023] Preferably, the step of gridding the electronic map of the scenic area and collecting the frequency of various behavior tags in each grid includes:

[0024] Based on the preset side length, the electronic map of the scenic area is gridded to obtain the geographical information of each grid.

[0025] Based on the geographic information corresponding to the behavior tags, the behavior tags are classified into corresponding grids; and

[0026] Count the number of times each behavior label appears in each grid.

[0027] Preferably, the step of obtaining the behavior label when the number of occurrences of the behavior label in the grid meets the preset threshold corresponding to the behavior label includes:

[0028] Determine whether the number of occurrences of each behavior label in the grid meets the preset threshold corresponding to the behavior label. If yes, the grid obtains the behavior label; otherwise, the grid does not obtain the behavior label.

[0029] Preferably, generating a point of interest on the electronic map based at least on the behavior labels of the grid includes:

[0030] A tag combination is established based on the behavior tags of the grid;

[0031] Based on tag combinations and a preset mapping table, a point of interest is generated on the electronic map. The mapping table includes the mapping relationship between various tag combinations, various types of points of interest, and schematic icons of points of interest.

[0032] Add the point of interest icon corresponding to the point of interest to the location of the point of interest in the electronic map.

[0033] Preferably, the step of collecting tourists' current location information and matching the tourists' historical behavior tags in various scenic spots with points of interest in the electronic map to provide tourists with matching points of interest includes:

[0034] Collect tourists' current location information and their historical behavior tags in various scenic spots;

[0035] Matching is performed based on the combination of tags corresponding to the nearest points of interest in the electronic map, using the most frequent historical behavior tags of the tourists.

[0036] A navigation path is generated based on the current location information and the point of interest with the most hits in the behavior tag.

[0037] Embodiments of the present invention also provide a clustering system for detecting tourist behavior in scenic areas, used to implement the above-described clustering method for detecting tourist behavior in scenic areas. The clustering system for detecting tourist behavior in scenic areas includes:

[0038] The image acquisition module collects images with geographic information based on the scenic area;

[0039] The mapping module identifies the behavior of tourists in the image based on the positional relationship between skeletal points, obtains behavior tags, and establishes a mapping relationship between the behavior tags and the geographic information of the image.

[0040] The grid statistics module divides the electronic map of the scenic area into grids and collects the frequency of various behavior tags in each grid.

[0041] The behavior labeling module determines that when the number of occurrences of a behavior label in the grid meets a preset threshold corresponding to that behavior label, the grid obtains the behavior label.

[0042] The point of interest generation module generates a point of interest on the electronic map, based at least on the behavior labels of the grid.

[0043] The point of interest matching module collects the tourist's current location information, matches the tourist's historical behavior tags in various scenic spots with the points of interest in the electronic map, and provides the tourist with points of interest that match the information.

[0044] Embodiments of the present invention also provide a clustering device for detecting tourist behavior in scenic areas, comprising:

[0045] processor;

[0046] A memory in which executable instructions of the processor are stored;

[0047] The processor is configured to perform the steps of the above-mentioned method for detecting and clustering tourist behavior in scenic areas by executing the executable instructions.

[0048] Embodiments of the present invention also provide a computer-readable storage medium for storing a program, which, when executed, implements the steps of the above-described method for detecting and clustering tourist behavior in scenic areas.

[0049] The purpose of this invention is to provide a method, system, device, and storage medium for detecting and clustering tourist behavior in scenic areas. It can identify tourist behavior within scenic areas, add points of interest to various parts of the scenic area based on tourist behavior tags, optimize the scenic area map, enhance the data-driven guidance experience of the scenic area, and greatly improve the user's tour experience. Attached Figure Description

[0050] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0051] Figure 1 This is a flowchart of the clustering method for detecting tourist behavior in scenic areas according to the present invention.

[0052] Figures 2 to 5 This is a framework diagram of the product detection method based on the detection and clustering method of tourist behavior in scenic areas according to the present invention.

[0053] Figure 6 This is a schematic diagram of the module of the scenic area visitor behavior detection and clustering system of the present invention.

[0054] Figure 7 This is a schematic diagram of the structure of the scenic area visitor behavior detection and clustering device of the present invention.

[0055] Figure 8 This is a schematic diagram of the structure of a computer-readable storage medium according to an embodiment of the present invention. Detailed Implementation

[0056] The following specific examples illustrate the implementation methods of this application. Those skilled in the art can easily understand the other advantages and effects of this application from the content disclosed herein. This application can also be implemented or applied through other different specific embodiments, and various details in this application can be modified or changed according to different viewpoints and application systems without departing from the spirit of this application. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0057] The embodiments of this application will now be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily implement the application. This application may be embodied in many different forms and is not limited to the embodiments described herein.

[0058] In this application, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics represented in connection with that embodiment or example, which are included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics represented may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate different embodiments or examples represented in this application, as well as features of different embodiments or examples.

[0059] Furthermore, the terms "first" and "second" are used for illustrative purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the representation of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0060] For the purpose of clearly describing this application, devices that are not relevant to the description are omitted, and the same or similar components throughout the specification are given the same reference numerals.

[0061] Throughout this specification, when it is said that a device is "connected" to another device, this includes not only "direct connection" but also "indirect connection" by placing other components in between. Furthermore, when it is said that a device "comprises" a certain constituent element, unless otherwise stated otherwise, this does not exclude other constituent elements, but rather implies that other constituent elements may be included.

[0062] When we say that a device is "above" another device, this can mean that it is directly above the other device, or it can mean that other devices are present in between. Conversely, when we say that a device is "directly" "above" another device, there are no other devices present in between.

[0063] Although the terms first, second, etc., are used in some instances herein to refer to various elements, these elements should not be limited by these terms. These terms are used only to distinguish one element from another. For example, first interface and second interface, etc., are used. Furthermore, as used herein, the singular forms “a,” “an,” and “the” are intended to also include the plural forms unless the context indicates otherwise. It should be further understood that the terms “comprising,” “including,” indicate the presence of features, steps, operations, elements, components, items, kinds, and / or groups, but do not exclude the presence, occurrence, or addition of one or more other features, steps, operations, elements, components, items, kinds, and / or groups. The terms “or” and “and / or” as used herein are interpreted as inclusive, or mean any one or any combination thereof. Thus, “A, B, or C” or “A, B, and / or C” means “any one of: A; B; C; A and B; A and C; B and C; A, B, and C.” Exceptions to this definition will only occur if the combination of elements, functions, steps, or operations is inherently mutually exclusive in some way.

[0064] The technical terms used herein are for reference only to specific embodiments and are not intended to limit the scope of this application. The singular form used herein includes the plural form unless the statement explicitly indicates otherwise. The word "comprising" as used in the specification means to specify a particular characteristic, region, integer, step, operation, element, and / or component, and does not exclude the presence or addition of other characteristics, regions, integers, steps, operations, elements, and / or components.

[0065] Although not explicitly defined, all terms, including technical and scientific terms used herein, shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. Terms defined in commonly used dictionaries shall be further interpreted as having a meaning consistent with the relevant technical literature and the content of this present application, and shall not be over-interpreted as having an ideal or overly formulaic meaning unless otherwise defined.

[0066] Figure 1This is a flowchart of the clustering method for detecting tourist behavior in scenic areas according to the present invention. Figure 1 As shown, an embodiment of the present invention provides a method for detecting and clustering tourist behavior in scenic areas, comprising the following steps:

[0067] S110. Collect images with geographic information based on the scenic area.

[0068] S120. Based on the positional relationship between skeletal points, identify the behavior of tourists in the image to obtain behavior labels and establish a mapping relationship between behavior labels and the geographic information of the image.

[0069] S130. Grid the electronic map of the scenic area and collect the frequency of various behavior tags in each grid.

[0070] S140. When the number of times a behavior label is displayed in the grid meets the preset threshold corresponding to that behavior label, the grid obtains a behavior label.

[0071] S150, At least based on grid-based behavior labels, generate a point of interest on the electronic map.

[0072] S160. Collect tourists' current location information, match tourists' historical behavior tags in various scenic spots with points of interest in the electronic map, and provide tourists with points of interest that match the information.

[0073] In a preferred embodiment, step S110 includes:

[0074] S111. Photos located within the scenic area, collected via the network. The photos include geographical information and the time of capture.

[0075] S112. Collect photos taken by surveillance cameras within the scenic area. And...

[0076] S113. Collect photos taken by drones or unmanned vehicles within the scenic area, thereby collecting as many photos as possible within the scenic area, but not limited to this.

[0077] In a preferred embodiment, step S120 includes:

[0078] S121. Use the MoveNet pre-trained model to extract 17 skeletal key points of the human body from the image.

[0079] S122. Perform coordinate transformation on the 17 skeletal key points, that is, transform their coordinates to a coordinate system with one of the key points as the origin.

[0080] S123. Input the 17 human skeleton key points after coordinate transformation into the multilayer perceptron for training.

[0081] S124. Convert the multilayer perceptron model to a TFLite model to obtain the behavior label with the highest confidence.

[0082] S125. Establish a mapping relationship between behavioral tags and geographic information of images, but not limited to this.

[0083] In a preferred embodiment, step S130 includes:

[0084] S131. Based on the preset side length, the electronic map of the scenic area is gridded to obtain the geographical information of each grid.

[0085] S132. Based on the geographic information corresponding to the behavior tags, classify the behavior tags into the corresponding grids.

[0086] S133. Count the number of times each behavior label appears in each grid to obtain the behavior that tourists like to perform in the grid, but this is not the only limit.

[0087] In a preferred embodiment, step S140 includes:

[0088] S141. Determine whether the number of times each type of behavior label in the grid meets the preset threshold corresponding to the behavior label. If yes, proceed to step S142; otherwise, proceed to step S143.

[0089] S142, the grid obtains behavior labels. And...

[0090] S143. The grid did not obtain behavior labels, so the behaviors that tourists frequently perform in a certain grid are labeled to make the grid information easier for subsequent data processing, but this is not the only option.

[0091] In a preferred embodiment, step S150 includes:

[0092] S151. Establish tag combinations based on grid-based behavior tags.

[0093] S152. Based on the tag combination and the preset mapping table, generate a point of interest on the electronic map. The mapping table includes the mapping relationship of various tag combinations, various types of points of interest, and the schematic diagram of the point of interest.

[0094] S153. Add the corresponding point of interest map icon to the location of the point of interest on the electronic map, so that users can more clearly understand the types of recreational experiences in each area of ​​the scenic spot, but this is not a limitation.

[0095] In a preferred embodiment, step S160 includes:

[0096] S161. Collect tourists' current location information and tourists' historical behavior tags in various scenic spots.

[0097] S162. Match the most frequent historical behavior tags of tourists with the tags corresponding to nearby points of interest in the electronic map. The most popular behaviors of users can be obtained through the statistics of tourists' historical behavior tags, which makes it easier to push relevant points of interest to tourists.

[0098] S163. Generate a navigation path based on the current location information and the point of interest with the most hits in the behavior tag, but not limited to this.

[0099] This invention allows users to clearly identify points of interest in large scenic areas without maps, and promotes relevant tour routes based on user interests, greatly enhancing the data-driven guidance experience in scenic areas.

[0100] Figures 2 to 5 This is a framework diagram of the product detection method based on the detection and clustering of tourist behavior in scenic areas, as described in this invention. Figures 2 to 5 The specific implementation process of the present invention is as follows:

[0101] First, photos of the scenic area (undeveloped areas or large national parks) are collected via the internet. These photos include geographic information and the time the photos were taken. Photos taken by surveillance cameras within the scenic area are also collected. Finally, photos taken by drones or unmanned vehicles within the scenic area are also collected, thus gathering as many photos as possible within the scenic area.

[0102] See Figure 2 This study utilizes the MoveNet pre-trained model to extract 17 skeletal keypoints from images. The coordinates of these 17 keypoints are then transformed to a coordinate system with one of the keypoints as the origin. These transformed keypoints are then input into a multilayer perceptron for training. The multilayer perceptron model is then transformed using the TFLite model to obtain the behavior label with the highest confidence. A mapping relationship is established between the behavior label and the geographic information of the image, but this is not a limitation. The process of identifying behavior labels can use Google's state-of-the-art pose detection model, but is not limited to this. MoveNet is a fast and accurate pose detection model capable of detecting 17 keypoints on the human body, running at 50+ fps on laptops and mobile phones. TFLite is a toolkit for deploying deep learning models on mobile and embedded devices, which can improve computing speed and reduce memory and GPU memory usage by transforming, deploying, and optimizing a trained TF model.

[0103] Human pose recognition, a branch of computer vision, aims to improve human-computer interaction by detecting key points and recognizing human poses in images or videos containing people. Human pose recognition has a wide range of applications, including medical rehabilitation, posture guidance in sports and dance, and intelligent security. To make it suitable for mobile and embedded devices, this invention provides a real-time human pose recognition method. This method mainly includes a human pose detection model and a human pose classification model. The main process involves first using the MoveNet human pose detection model to detect 17 skeletal key points. Then, these 17 key points are processed by transforming their coordinates to a coordinate system with one of the key points as the origin. Finally, the processed coordinates of the 17 key points are input into the human pose classification model for classification. The human pose classification model is mainly composed of a multilayer perceptron. To further improve the model's inference speed and achieve real-time performance, the classification model is converted to a TFLite model.

[0104] This invention relates to human pose detection, human pose classification, and the TFLite field. Human pose detection is mainly divided into traditional pose detection and deep learning-based pose detection. Deep learning-based human pose detection methods are further divided into top-down and bottom-up approaches. Top-down methods first detect the human body and then specifically detect the position of each joint; while bottom-up methods first detect the human body's joints and then connect them to form the human skeleton. The MoveNet model belongs to the bottom-up approach, and its main characteristics are high speed and high accuracy. Human pose classification mainly utilizes a multilayer perceptron model to classify the pose of 17 input skeletal keypoints. To meet the model's real-time requirements, the multilayer perceptron model is converted to a TFLite model, with int8 quantization performed during the conversion process, greatly improving the model's inference speed while maintaining classification performance.

[0105] Human pose recognition has wide applications in medical rehabilitation, sports and dance posture guidance, and other fields. To apply this algorithm to mobile and embedded devices, this invention presents a real-time pose recognition method based on skeletal point detection and a multilayer perceptron.

[0106] Currently, human pose detection mainly includes classic algorithms such as OpenPose and PoseNet. OpenPose primarily utilizes VGG19 for feature extraction, but its inference speed is slow, making real-time performance difficult. The PoseNet model, based on MobileNet V1, offers improved inference speed compared to OpenPose, but its detection performance is poor. MoveNet, on the other hand, is a fast and accurate pose detection model.

[0107] The human pose classification method primarily utilizes a multilayer perceptron to classify 17 skeletal key points. Finally, to achieve real-time performance, the multilayer perceptron model is converted to a TFLite model, with int8 quantization performed during the conversion process. This ensures the model's recognition accuracy while further improving its inference speed.

[0108] The purpose of this invention is to develop a human posture recognition model based on skeletal point detection and multilayer perceptron, which can not only accurately detect key points of the human skeleton and identify posture classification, but also has real-time performance to be applicable to mobile or embedded devices, so as to meet the needs of human posture recognition in medical rehabilitation and exercise posture guidance.

[0109] This invention employs a real-time pose recognition method that can accurately detect skeletal key points and classify poses, and also possesses real-time performance suitable for mobile or embedded devices. This technical solution is implemented as follows:

[0110] (1) Use MoveNet pre-trained model to extract 17 key points of human skeleton;

[0111] (2) Transform the coordinates of the 17 skeletal key points, that is, transform their coordinates to be based on a certain key point;

[0112] In the coordinate system of the origin;

[0113] (3) Input the 17 human skeleton key points after coordinate transformation into the multilayer perceptron for training;

[0114] (4) The multilayer perceptron model is converted to the TFLite model and Int8 quantization is performed during the conversion process to further improve the inference speed while ensuring recognition performance.

[0115] This invention discloses a real-time human pose recognition method based on human skeletal point detection and multilayer perceptron (MLP) to accurately identify human posture. After obtaining the coordinates of 17 skeletal key points using a human pose detection algorithm, this invention does not directly input these coordinates into the human pose classification model. Instead, it transforms the coordinates of the skeletal key points so that their coordinates lie in a coordinate system with one of the skeletal key points as the origin. After training the human pose classification model, its inference speed is relatively slow. Therefore, a TFLite transformation is performed to further improve the model's inference speed.

[0116] This invention presents a real-time human pose recognition method based on human skeletal point detection and a multilayer perceptron. MoveNet is a fast and accurate human pose detection algorithm that can accurately identify 17 skeletal key points of the human body. Transforming the coordinates of these 17 key points to the same coordinate system with a key point as the origin before inputting them into the human pose classification network further improves the classification performance. Finally, by performing a TFLite model transformation on the human pose classification model, the inference speed of the model is further improved while maintaining recognition performance. Therefore, this invention proposes a highly accurate and real-time human pose recognition method applicable to mobile or embedded devices.

[0117] This invention presents a high-accuracy and fast-inference-speed human pose recognition algorithm based on human skeletal point detection and a multilayer perceptron, suitable for mobile or embedded devices. The method considers both human pose detection and human pose classification, requiring both models to achieve high inference speed and accuracy. Since MoveNet is a fast and accurate model, it is used to detect 17 skeletal key points for human pose detection. For human pose classification, the coordinates of the 17 key points output by the human pose detection model are first transformed, and then input into the human pose classification model (multilayer perceptron) to further improve the model's accuracy. Finally, TFLite transformation is applied to the human pose classification model to further improve its inference speed.

[0118] The real-time human pose recognition method used in the embodiments of the present invention mainly comprises the following steps:

[0119] 1. First, the MoveNet.TFLite model is used for human pose detection. Its main structure consists of...

[0120] The three parts are: Backbone, Header, and Loss Function.

[0121] The backbone part mainly utilizes MobileNet v2+FPN to extract features;

[0122] The feature maps extracted from the backbone are input into the header, and after passing through their respective convolutions, each outputs its own dimension.

[0123] The characteristics of the degree. There are four headers: Center, KeypointRegression, KeypointHeatmap, and Local Offset. Its overall structure is as follows: Figure 3 As shown:

[0124] Where Center: [N, H, W, 1], where 1 is the central keypoint of the human body. There are two main ways to extract the central keypoint: the first is the average of all human skeletal keypoints; the second is the center point of the largest bounding rectangle of all keypoints.

[0125] KeypointHeatmap: [N, H, W, K], where K is the number of keypoints in the human skeleton.

[0126] KeypointRegression: [N, H, W, 2K], where K is the number of keypoints in the human skeleton, represented by coordinates (x, y), so there are 2K data points. Here, (x, y) represents the offset of each keypoint from its center point.

[0127] LocalOffsets: [N, H, W, 2K], which corresponds to the offset coordinates of K human body key points, because the feature map will have quantization error during the downsampling process of the model.

[0128] Loss function: KeypointHeadmap and Center use MSE, KeypointRegression

[0129] Both LocalOffset and LocalOffset use L1 Loss.

[0130] 2. Transform the coordinates of the 17 key points of the human skeleton so that the 17 key points are in a coordinate system with the 12th key point (left hip) as the origin;

[0131] 3. A multilayer perceptron was used to train 17 skeletal key points after coordinate transformation. The architecture for classifying human skeletal points using a multilayer perceptron is as follows: Figure 4 As shown.

[0132] Depend on Figure 4 As we know, since the input is the coordinates of key points on the human skeleton, i.e., 17 (x, y) coordinates, the multilayer perceptron has 34 inputs and the output is the accuracy of pose classification. There are 9 hidden layers, with the following numbers of neurons: 512, 256, 128, 128, 64, 48, 36, 12, and 4, respectively. Each input neuron and the neuron connected to the next layer have a weight value w and a bias value b. The relationship between the input and output domains is as follows:

[0133]

[0134] in, The output is the j-th neuron in the (h+1)-th layer; This represents the value of the i-th neuron in the h-th layer; The weights connecting x and y; This is the corresponding bias.

[0135] Between adjacent layers, a ReLU function performs a non-linear transformation on the result of the previous layer. During training, the multilayer perceptron network uses backpropagation to refine the parameters of the hidden layers. To accelerate model training and improve accuracy during backpropagation, the Adam method is used.

[0136] 4. Finally, the trained classification model is converted to TFLite to further improve the inference speed of the model.

[0137] This invention can be used to conduct experiments on human pose recognition tasks. First, the MoveNet.TFLite model was tested using a self-built dataset. This dataset mainly consists of four actions: standing, squatting, lunging, and raising an arm. Each action dataset contains approximately 2000 images, and images of two or three people were collected for each action. The test results are shown in the following figure:

[0138] For the human pose classification network, a self-built dataset was used for training and testing, with a training set to test set ratio of 8:2. First, images were input into the MoveNet human pose detection model, resulting in the output of (x, y) coordinates for 17 skeletal keypoints. The coordinates were preprocessed by converting them to a coordinate system with the twelfth skeletal keypoint (left hip) as the origin. The transformed coordinate data was then used as input to a multilayer perceptron. The batch size was set to 8, and the learning rate for the human pose classification network was set to 2 × 10⁻⁶. -4 With the number of iterations set to 60, the classification accuracy was 99.87%.

[0139] The inference times for the human pose detection model and the human pose classification model are shown below:

[0140] Model Human posture detection Human posture classification Time (ms) 22.23 0.1ms

[0141] Table 1 Inference time for human posture detection and classification models

[0142] The results above show that the human posture recognition model proposed in this invention achieves high recognition accuracy and fast inference speed, and can be applied to mobile or embedded devices.

[0143] After obtaining the behavior tags (e.g., fishing, kite flying, picnicking) from the photos, the electronic map of the scenic area is gridded according to a preset side length to obtain the geographic information of each grid. Based on the geographic information corresponding to the behavior tags, the behavior tags are categorized into their corresponding grids. The frequency of each behavior tag appearing in each grid is counted to obtain the behaviors that tourists most like to perform in that grid, but this is not a limitation.

[0144] The system determines whether the frequency of each type of behavior label in the grid meets the preset threshold corresponding to the behavior label. If yes, the grid obtains a behavior label; otherwise, the grid does not obtain a behavior label. This allows for the labeling of behaviors that tourists frequently perform in a certain grid, making the grid information easier for subsequent data processing. However, this is not a limitation.

[0145] Then, see Figure 5 A tag combination is established based on grid-based behavior tags. Based on the tag combination and a preset mapping table, a point of interest (POI) is generated on the electronic map. The mapping table includes the mapping relationships between various tag combinations, different types of POIs, and POI schematic icons. In this embodiment, three POIs (1, 2, and 3) are added to the electronic map. POI 1 has a kite-flying behavior tag; the behavior tags in related images indicate that many tourists fly kites here, clearly indicating that POI 1 is a suitable place for kite flying. POI 2 has a fishing behavior tag; the behavior tags in related images indicate that many tourists fish here, clearly indicating that POI 2 is a suitable place for fishing. POI 3 has a picnic behavior tag; the behavior tags in related images indicate that many tourists picnic here, clearly indicating that POI 3 is a suitable place for picnics. The corresponding POI schematic icons are added to the POI's location on the electronic map, allowing users to more clearly understand the types of recreational experiences in different areas of the scenic area.

[0146] Finally, the current location information of the tourists is collected. Figure 5 The system uses the user's current location (4) and the tourist's historical behavior tags in various scenic spots. It matches the most frequently occurring historical behavior tags with the tags corresponding to nearby points of interest (POIs) on the electronic map. By statistically analyzing the tourist's historical behavior tags, it can identify the user's most popular behaviors, making it easier to push relevant POIs to the tourist. A navigation path L is generated based on the current location information and the POI with the most frequently matched behavior tags, facilitating the tourist's journey to POI 3 for fishing and allowing them to meet more like-minded tourists, enhancing the overall experience. This allows users to easily find the areas where their interests lie, even without a scenic map.

[0147] Figure 6 This is a schematic diagram of the modules of the scenic area visitor behavior detection and clustering system of the present invention. Figure 6 As shown, embodiments of the present invention also provide a clustering system for detecting tourist behavior in scenic areas, used to implement the above-described clustering method for detecting tourist behavior in scenic areas. The clustering system for detecting tourist behavior in scenic areas includes:

[0148] Image acquisition module 51 acquires images with geographic information based on the scenic area.

[0149] The mapping module 52 identifies the behavior of tourists in the image based on the positional relationship between skeletal points, obtains behavior labels, and establishes a mapping relationship between the behavior labels and the geographic information of the image.

[0150] The grid statistics module 53 grids the electronic map of the scenic area and collects the frequency of various behavior tags in each grid.

[0151] The behavior label module 54 obtains a behavior label when the number of behavior labels in the grid meets the preset threshold corresponding to the behavior label.

[0152] The point of interest generation module 55 generates a point of interest on the electronic map based on at least grid-based behavior labels.

[0153] The point of interest matching module 56 collects tourists' current location information, matches tourists' historical behavior tags in various scenic spots with points of interest in the electronic map, and provides tourists with points of interest that match the information.

[0154] In a preferred embodiment, the image acquisition module 51 is configured to acquire photos located within the scenic area based on network-based geographic information. This includes acquiring photos taken by surveillance cameras within the scenic area, and photos taken by drones or unmanned vehicles within the scenic area.

[0155] In a preferred embodiment, the mapping module 52 is configured to extract 17 skeletal keypoints of the human body from an image using a MoveNet pre-trained model. The coordinates of these 17 keypoints are then transformed to a coordinate system with one of the keypoints as the origin. The transformed 17 skeletal keypoints are then input into a multilayer perceptron for training. The multilayer perceptron model is then transformed using the TFLite model to obtain the behavior label with the highest confidence. A mapping relationship is established between the behavior label and the geographic information of the image.

[0156] In a preferred embodiment, the grid statistics module 53 is configured to grid the electronic map of the scenic area according to a preset side length, obtaining the geographic information of each grid. Based on the geographic information corresponding to the behavior tags, the behavior tags are categorized into their corresponding grids. The frequency of each behavior tag appearing in each grid is counted.

[0157] In a preferred embodiment, the behavior label module 54 is configured to determine whether the number of each type of behavior label in the grid meets the preset threshold corresponding to the behavior label. If yes, the grid obtains a behavior label; otherwise, the grid does not obtain a behavior label.

[0158] In a preferred embodiment, the point of interest (POI) generation module 55 is configured to establish tag combinations based on grid-based behavior tags. Based on the tag combinations and a preset mapping table, a POI is generated on the electronic map. The mapping table includes mapping relationships between various tag combinations, different types of POIs, and POI schematic icons. The corresponding POI schematic icon is then added to the POI's location on the electronic map.

[0159] In a preferred embodiment, the point of interest (POI) matching module 56 is configured to collect the tourist's current location information and the tourist's historical behavior tags in various scenic spots. It matches the most frequently occurring historical behavior tags with the corresponding tags of nearby POIs in the electronic map. A navigation path is generated based on the current location information and the POIs with the most frequent matching behavior tags.

[0160] The purpose of this invention is to provide a clustering system for detecting tourist behavior in scenic areas. This system can identify the behavior of tourists within the scenic area, add points of interest to various parts of the scenic area based on the tourists' behavior tags, optimize the scenic area map, and enhance the data-driven guidance experience of the scenic area, thus greatly improving the user's tour experience.

[0161] This invention also provides a device for detecting and clustering tourist behavior in scenic areas, including a processor and a memory storing executable instructions for the processor. The processor is configured to execute steps of a method for detecting and clustering tourist behavior in scenic areas by executing the executable instructions.

[0162] As shown above, the scenic area visitor behavior detection and clustering system of this invention can identify the behavior of visitors in the scenic area, add points of interest to various parts of the scenic area based on the behavior tags of visitors, optimize the scenic area map, and enhance the data-driven guidance experience of the scenic area, thus greatly improving the user's tour experience.

[0163] Those skilled in the art will understand that various aspects of the present invention can be implemented as systems, methods, or program products. Therefore, various aspects of the present invention can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, collectively referred to herein as a "circuit," "module," or "platform."

[0164] Figure 7 This is a schematic diagram of the structure of the scenic area visitor behavior detection and clustering device of the present invention. See below for reference. Figure 7 To describe an electronic device 600 according to this embodiment of the present invention. Figure 7 The electronic device 600 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0165] like Figure 7As shown, the electronic device 600 is presented in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including storage unit 620 and processing unit 610), a display unit 640, etc.

[0166] The storage unit stores program code, which can be executed by the processing unit 610 to perform the steps described in the above-described section on the electronic prescription transfer processing method according to various exemplary embodiments of the present invention. For example, the processing unit 610 can perform actions such as... Figure 1 The steps are shown in the figure.

[0167] Storage unit 620 may include a readable medium in the form of a volatile storage unit, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include a read-only memory (ROM) 6203.

[0168] Storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0169] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the multiple bus structures.

[0170] Electronic device 600 can also communicate with one or more external devices 700 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 660. Network adapter 660 can communicate with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0171] This invention also provides a computer-readable storage medium for storing a program, which, when executed, implements the steps of a method for detecting and clustering tourist behavior in scenic areas. In some possible implementations, various aspects of this invention can also be implemented as a program product comprising program code, which, when run on a terminal device, causes the terminal device to perform the steps described in the above-described electronic prescription processing method section of this specification according to various exemplary embodiments of the invention.

[0172] As shown above, the scenic area visitor behavior detection and clustering system of this invention can identify the behavior of visitors in the scenic area, add points of interest to various parts of the scenic area based on the behavior tags of visitors, optimize the scenic area map, and enhance the data-driven guidance experience of the scenic area, thus greatly improving the user's tour experience.

[0173] Figure 8 This is a schematic diagram of the structure of the computer-readable storage medium of the present invention. (Reference) Figure 8 As shown, a program product 800 for implementing the above-described method according to an embodiment of the present invention is described. This product may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0174] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0175] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0176] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0177] In summary, the purpose of this invention is to provide a method, system, device, and storage medium for detecting and clustering tourist behavior in scenic areas. This invention can identify tourist behavior within a scenic area, add points of interest to various parts of the scenic area based on tourist behavior tags, optimize the scenic area map, enhance the data-driven guidance experience of the scenic area, and greatly improve the user's tour experience.

[0178] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A clustering method for detecting tourist behavior in scenic areas, characterized in that, Includes the following steps: Collect images with geographic information based on the scenic area; The MoveNet pre-trained model was used to extract 17 skeletal key points of the human body from the image. The coordinates of 17 skeletal key points are transformed into a coordinate system with one of the key points as the origin. The 17 coordinate-transformed human skeletal key points are then input into a multilayer perceptron for training. The multilayer perceptron model is then transformed into a TFLite model to obtain the behavior label with the highest confidence. A mapping relationship between the behavior label and the geographic information of the image is then established. The electronic map of the scenic area is gridded, and the frequency of various behavior tags in each grid is collected. When the number of occurrences of the behavior label in the grid meets the preset threshold corresponding to the behavior label, the grid obtains the behavior label; A tag combination is established based on the behavior tags of the grid; a point of interest is generated on the electronic map based on the tag combination and a preset mapping table, wherein the mapping table includes the mapping relationship between various tag combinations, various types of points of interest, and point of interest schematic icons; the point of interest schematic icon corresponding to the point of interest is added to the position of the point of interest on the electronic map. Collect tourists' current location information, match the tourists' historical behavior tags in various scenic spots with points of interest in the electronic map, and provide tourists with points of interest that match the information.

2. The method for detecting and clustering tourist behavior in scenic areas as described in claim 1, characterized in that, The collection of images with geographic information based on the scenic area includes: Photos of the scenic area were collected based on geographic information gathered via the network. Collect photos taken by the surveillance cameras within the scenic area; and Collect photos taken by drones or unmanned vehicles within the scenic area.

3. The method for detecting and clustering tourist behavior in scenic areas as described in claim 1, characterized in that, The electronic map of the scenic area is gridded, and the frequency of various behavior tags appearing in each grid is collected, including: Based on the preset side length, the electronic map of the scenic area is gridded to obtain the geographical information of each grid. Based on the geographic information corresponding to the behavior tags, the behavior tags are classified into corresponding grids; and Count the number of times each behavior label appears in each grid.

4. The method for detecting and clustering tourist behavior in scenic areas as described in claim 1, characterized in that, The step of obtaining the behavior label when the number of occurrences of the behavior label in the grid meets the preset threshold corresponding to the behavior label includes: Determine whether the number of occurrences of each behavior label in the grid meets the preset threshold corresponding to the behavior label. If yes, the grid obtains the behavior label; otherwise, the grid does not obtain the behavior label.

5. The method for detecting and clustering tourist behavior in scenic areas as described in claim 1, characterized in that, The process of collecting tourists' current location information, matching tourists' historical behavior tags in various scenic spots with points of interest in the electronic map, and providing tourists with matching points of interest includes: Collect tourists' current location information and their historical behavior tags in various scenic spots; Matching is performed based on the combination of tags corresponding to the nearest points of interest in the electronic map, using the most frequent historical behavior tags of the tourists. A navigation path is generated based on the current location information and the point of interest with the most hits in the behavior tag.

6. A clustering system for detecting tourist behavior in scenic areas, used to implement the clustering method for detecting tourist behavior in scenic areas as described in claim 1, characterized in that, include: The image acquisition module collects images with geographic information based on the scenic area; The mapping module uses the MoveNet pre-trained model to extract 17 skeletal key points of the human body based on the image; it then performs coordinate transformation on the 17 skeletal key points, that is, transforms their coordinates to a coordinate system with a certain key point as the origin; the 17 skeletal key points of the human body after coordinate transformation are input into a multilayer perceptron for training; the multilayer perceptron model is transformed into a TFLite model to obtain the behavior label with the highest confidence. Establish a mapping relationship between the behavior tags and the geographic information of the images; The grid statistics module divides the electronic map of the scenic area into grids and collects the frequency of various behavior tags in each grid. The behavior labeling module determines that when the number of occurrences of a behavior label in the grid meets a preset threshold corresponding to that behavior label, the grid obtains the behavior label. The point of interest (POI) generation module establishes tag combinations based on the behavior tags of the grid; based on the tag combinations and a preset mapping table, it generates a POI on the electronic map, wherein the mapping table includes the mapping relationship between various tag combinations, various types of POIs, and POI schematic icons; and it adds the POI schematic icon corresponding to the POI to the location of the POI on the electronic map. The point of interest matching module collects the tourist's current location information, matches the tourist's historical behavior tags in various scenic spots with the points of interest in the electronic map, and provides the tourist with points of interest that match the information.

7. A clustering device for detecting tourist behavior in scenic areas, characterized in that, include: processor; A memory in which executable instructions of the processor are stored; The processor is configured to perform the steps of the method for detecting and clustering tourist behavior in scenic areas according to any one of claims 1 to 5 by executing the executable instructions.

8. A computer-readable storage medium for storing a program, characterized in that, When the program is executed by the processor, it implements the steps of the method for detecting and clustering tourist behavior in scenic areas as described in any one of claims 1 to 5.

Images