Method for generating urban-level large-scale population activity trajectories based on mobile phone signaling

Abstract

The present invention proposes a method for generating large-scale crowd activity trajectories at the city level based on mobile phone signaling, belonging to the field of intelligent transportation, including: S1: Obtain the signaling data of a specific group of people in the selected city for one day, perform preprocessing operations, and divide the map area into grids; S2: Use a masked autoencoder model to process each preprocessed trajectory sequence to obtain the probability that each trajectory point belongs to different grids; S3: Construct a loss function for trajectory accuracy and distribution diversity, and train the parameters by reconstructing the masked trajectory points, and finally obtain the parameters of the model in the pre-training stage; S4: Use the model in the pre-training stage to generate a complete final trajectory; S5: Overlay the corresponding longitude and latitude coordinates of a specific person at a certain moment on the generated trajectory points, and finally form a complete final trajectory of a specific person. The method of the present invention can generate large-scale crowd activity trajectory data that is consistent with the distribution of the original signaling trajectory data without exposing personal privacy.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation, and particularly relates to a method for generating urban-level large-scale population activity trajectories based on mobile phone signaling. Background Art

[0002] In recent years, with the popularization of mobile devices, especially mobile phones, the user trajectory data collected by network operators has become an important information source for studying individual and group movement patterns. Based on the mobile big data extracted from mobile device communications, which contains rich user geographical information and time information, it has greatly promoted the research on the movement trajectories of individual users at the million-data level within several months, providing an important basis for individual behavior semantic analysis. The analysis based on user trajectory data is of great significance for big data research and applications such as smart cities.

[0003] However, since mobile phone signaling data contains a large amount of personal information and sensitive data, the processing and use of this data must strictly comply with the laws and regulations on privacy protection. Therefore, in most cases, such data is strictly restricted to a few internal departments of communication operators. Due to the lack of a wide access channel for signaling data, researchers face many difficulties in conducting research in related fields. This limitation in data acquisition not only affects the depth and breadth of research, but also delays the accumulation of knowledge and the progress of technology in this field. Therefore, how to fully release the huge potential of signaling data while protecting personal privacy and being able to carry out geospatial analysis and population flow statistics without involving privacy has become an urgent problem to be solved. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a method for generating urban-level large-scale population activity trajectories based on mobile phone signaling, including the following steps:

[0005] Step S1: Obtain the signaling data of a specific population in a selected city for one day, perform preprocessing operations on the signaling data, and divide the map area into grids to obtain a preprocessed trajectory sequence of equal length;

[0006] Step S2: Construct a masked autoencoder model, including: a masking module, an encoder, a decoder, and a loss function module; first, use the masking module to sample different masking rates greater than 0.5 for each of the preprocessed trajectory sequences, mask each of the preprocessed trajectory sequences according to the masking rate to obtain a masked trajectory sequence; pass the masked trajectory sequence through the encoder and the decoder to obtain the probability that each trajectory point belongs to a different grid ;

[0007] Step S3: Construct a loss function for trajectory accuracy and distribution diversity, and train the parameters by reconstructing the masked trajectory points, finally obtaining the parameters of the model in the pre-training stage;

[0008] Step S4: In the trajectory generation stage, use the model in the pre-training stage and adopt the method of gradually increasing the generation of trajectory points. In each step, a certain proportion of trajectory points are incrementally generated until the obtained trajectory points form a complete final trajectory;

[0009] Step S5: Given the corresponding longitude and latitude coordinates of a specific person at a certain moment, and input them into the position of the initial trajectory sequence corresponding to this moment. On the basis of Step S4, after each step of generating trajectory points, cover the above-known corresponding longitude and latitude coordinates of the specific person at a certain moment in the generated trajectory points, and repeat multiple steps to generate trajectory points according to Step S4 until finally forming a complete final trajectory of the specific person. Repeat Step S5 until the final trajectories of a predetermined number of specific persons are generated. The set of these trajectories is the large-scale crowd movement trajectories at the city level.

[0010] Beneficial effects:

[0011] 1. The present invention designs a loss function for trajectory accuracy and distribution diversity, considering the authenticity and diversity of the trajectories generated by the network respectively. Existing methods only consider the error of predicting user trajectories when generating trajectories, and tend to generate relatively consistent trajectories when finally generating trajectories, which can only represent the most likely trajectories of individual users and cannot describe the diverse trajectories of user groups. The loss function proposed by the present invention can effectively solve this problem.

[0012] 2. The present invention designs a multi-step autoregressive trajectory generation architecture for gradually increasing the generation of trajectory points, which can balance the quality of the generated trajectories and the required time. Existing methods generate a fixed number of trajectory points in each step when generating the final trajectory. If the number of trajectory points generated in each step is large, the quality of the generated trajectory will not be high. Since the number of finally generated trajectory points does not change, if the number of trajectory points generated in each step is small, the number of steps will be large and the generation time will be long. The multi-step generation method of gradually increasing the generation of trajectory points proposed by the present invention can improve the quality of the generated trajectories in the early stage and reduce the trajectory generation time in the later stage.

[0013] 3. The present invention designs a method for generating trajectories based on specific spatio-temporal information. By inputting the spatio-temporal position of the user at the current moment into the model, specific trajectories can be obtained. Existing methods cannot insert the spatio-temporal information of users into the model and cannot generate user trajectories with specific spatio-temporal characteristics. The trajectory generation method proposed by the present invention can solve this problem. Description of the Drawings

[0014] Figure 1Schematic flow chart of a method for generating large-scale population activity trajectories at the city level based on mobile phone signaling according to the present invention;

[0015] Figure 2 It is a framework diagram of a method for generating large-scale population activity trajectories at the city level based on mobile phone signaling. Specific implementation manners

[0016] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0017] Embodiment 1

[0018] As Figure 1 shown, a method for generating large-scale population activity trajectories at the city level based on mobile phone signaling provided by an embodiment of the present invention includes the following steps:

[0019] Step S1: Obtain the signaling data of a specific population in a selected city for one day, perform preprocessing operations on the signaling data, and divide the map area into grids to obtain a preprocessed trajectory sequence of equal length;

[0020] Step S2: Construct a masked autoencoder model, including: a masking module, an encoder, a decoder, and a loss function module; first, use the masking module to sample different masking rates greater than 0.5 for each preprocessed trajectory sequence, and mask each preprocessed trajectory sequence according to the masking rate to obtain a masked trajectory sequence; pass the masked trajectory sequence through the encoder and decoder to obtain the probability that each trajectory point belongs to different grids ;

[0021] Step S3: Construct a loss function for trajectory accuracy and distribution diversity, and finally obtain the parameters of the model in the pre-training stage by reconstructing and training the parameters of the masked trajectory points;

[0022] Step S4: In the trajectory generation stage, use the model in the pre-training stage, and adopt the method of gradually increasing the generation of trajectory points. A certain proportion of trajectory points are generated incrementally in each step until the generated trajectory points form a complete final trajectory;

[0023] Step S5: Given the corresponding longitude and latitude coordinates of a specific person at a certain moment, and input them into the position of the initial trajectory sequence corresponding to that moment. On the basis of Step S4, after generating a trajectory point in each step, cover the known corresponding longitude and latitude coordinates of the specific person at a certain moment in the generated trajectory points. Repeat generating trajectory points in multiple steps according to Step S4 until finally forming a complete final trajectory of the specific person. Repeat Step S5 until the final trajectories of a predetermined number of specific persons are generated. The set of these trajectories is the urban-level large-scale crowd activity trajectories.

[0024] In one embodiment, the above-mentioned Step S1: Obtain the signaling data of a specific group of people in a selected city in one day, perform preprocessing operations on the signaling data, and divide the map area into grids to obtain a preprocessed trajectory sequence of equal length, specifically including:

[0025] Step S11: Obtain the original signaling data, including: user ID, base station ID, base station longitude, base station latitude, arrival time, and departure time;

[0026] Step S12: Group the user IDs, sort all the original signaling data in ascending order according to the arrival time, and convert the original signaling data of each user into a sequence data of equal length in units of 15 minutes. The length of this sequence data is set to , and the value corresponding to the sequence data of one day is 24;

[0027] Step S13: Eliminate the signaling data of users with ping-pong, drift phenomena or those who have been in the same area without traveling within a given time according to simple rules. Specifically, if the distance difference between the (i + 1)-th trajectory point and the i-th trajectory point is greater than 3 km, and among the subsequent trajectory point changes within 20 minutes or within 5 times, the distance between the j-th trajectory point and the i-th trajectory point is less than 1.5 km, it indicates that the user data is abnormal, and the trajectory points from the (i + 1)-th to the (j - 1)-th can be eliminated;

[0028] Step S14: Divide the map area into grids of 100m×100m, regard the base stations within the same grid as the same base station, set the ID of each grid, and obtain the preprocessed trajectory sequence. Each preprocessed trajectory sequence includes: user ID, grid ID, and current time.

[0029] In one embodiment, the above-mentioned Step S2: Construct a masked autoencoder model, including: a masking module, an encoder, a decoder, and a loss function module; first, use the masking module to sample different masking rates greater than 0.5 for each preprocessed trajectory sequence, and mask each preprocessed trajectory sequence according to the masking rate to obtain a masked trajectory sequence; pass the masked trajectory sequence through the encoder and the decoder to obtain the probability that each trajectory point belongs to a different grid , specifically including:

[0030] Step S21: Sample a probability according to the Gaussian distribution , and sample from the preprocessed trajectory sequence masks of different metrics, and perform masking operations on them, where is the length of the preprocessed trajectory sequence, as shown in formulas (1) - (2):

[0031] (1)

[0032] (2)

[0033] Among them, represents the trajectory sequence after masking, represents the trajectory sequence in after masking operation, that is, the corresponding grid ID is specified as ; represents the trajectory sequence in without masking operation;

[0034] Step S22: Define the encoding of each grid as , add the positional encoding , and finally map it to a -dimensional vector, as shown in formula (3). This formula uses the encoding of each grid plus the positional encoding in the sequence to represent the initial sequence:

[0035] (3)

[0036] Among them, the superscript 0 represents that this is the encoding of the 0th layer; represents the positional encoding; represents the one-hot encoding;

[0037] Step S23: Randomly discard a part of the masked data at a ratio of , and its index set is represented as , represents the data not discarded;

[0038] (4)

[0039] At the beginning of , connect an additional class vector , to form the input of the encoder:

[0040] (5)

[0041] Step S24: Input into an encoder, which consists of four sub-layers in sequence: layer normalization, multi-head self-attention layer, layer normalization, and feed-forward neural network; first, perform layer normalization on the input of the th layer and pass it into the multi-head self-attention layer as the input to obtain ; then, adopt residual connection to add the input to the output of the multi-head self-attention layer to obtain ; Go through layer normalization and the feed-forward neural network layer to obtain ; finally, after stacking the encoder layers, perform layer normalization again to achieve trajectory embedding encoding, as shown in Formulas (6) - (9):

[0042] (6)

[0043] (7)

[0044] (8)

[0045] (9)

[0046] wherein, is layer normalization, is the multi-head self-attention layer, is the feed-forward neural network, is the output of the encoder;

[0047] Step S25: Input into the decoder module, use the decoder to fill the previously masked positions, map the space of the encoder to the decoder space, is the bias term of this mapping, represents the trajectory points not discarded, represents the encoding of the masked positions in Prediction Step S21, and then introduce a new position encoding , and pass the current representation to a decoder with the same structure as the encoder but different parameters to obtain the final output of the decoder, as shown in Formulas (10) - (14):

[0048] (10)

[0049] (11)

[0050] (12)

[0051] (13)

[0052] (14)

[0053] Step S26: Through an additional linear layer , a GELU activation function and a layer normalization to convert it into the encoding ,Will and Multiply the map back to the grid to get a weight coefficient for each grid class and add this weight coefficient to an additional bias And process it through the Softmax layer to get each trajectory point belonging to different Probability , as shown in formulas (15) to (17):

[0054] (15)

[0055] (16)

[0056] (17).

[0057] In one embodiment, the above step S3: constructing a loss function of trajectory accuracy and distribution diversity, and finally obtaining the parameters of the pre-training stage model by reconstructing the training parameters of the mask trajectory points, specifically includes:

[0058] Step S31: represents the likelihood probability that the i-th trajectory point belongs to different grids, Represents a grid with masked operation, using trajectory accuracy loss function Predict the true position of the masked trajectory points, As shown in formula (18):

[0059] (18)

[0060] in, Represents the likelihood probability that the mask trajectory points belong to different grids;

[0061] Trajectory Accuracy Loss Function The goal is to maximize the log-likelihood of the actual position of the mask trajectory point, ensuring that the probability of the correct position is maximized;

[0062] Step S32: To obtain more diverse trajectories, design a distribution diversity loss function As shown in formula (19):

[0063] (19)

[0064] Since only using the trajectory accuracy loss function may lead to insufficient diversity in the model output. To obtain more diverse trajectories in terms of distribution, the present invention designs a distribution diversity loss function , if the probability distribution is concentrated on a certain value, the logarithmic mean value is larger; if the probability distribution is more uniform, the logarithmic mean value is smaller. This loss function regularizes the probability distribution, thereby enhancing the distribution diversity of the model;

[0065] Step S33: Combine the above two different loss functions to obtain a loss function that comprehensively considers the accuracy and distribution diversity of the reconstructed trajectory as follows:

[0066] (20)

[0067] wherein, is a parameter; in the embodiment of the present invention, is set to 0.001;

[0068] Step S34: Use the loss function to perform parameter training and backpropagation of the model to obtain the updated model parameters, and the pre-training process ends, retaining the pre-trained model parameters .

[0069] In one embodiment, in the above step S4: Trajectory generation stage, use the model in the pre-training stage and adopt a method of gradually incrementally generating trajectory points. A certain proportion of trajectory points are incrementally generated in each step until the obtained trajectory points form a complete final trajectory, which specifically includes:

[0070] Step S41: Denote the conditional probability model defined by the pre-trained model parameters as . The generation process starts from . All the data initially input into the pre-trained model are masked and used as the input for the first step to pass through the pre-trained model to obtain the first generated trajectory sequence. A certain proportion of the trajectory points at the beginning of the sampling sequence are used as the trajectory points generated in the first step , and then, taking as the condition, mask the subsequent trajectory points of and use them as the input for the second step to pass through the pre-trained model to obtain the second generated trajectory sequence. A larger proportion of the trajectory point data at the beginning of the sampling sequence are used as the trajectory points generated in the second step. Sample from sample , and so on. The sampling ratio for each step is calculated in step S42;

[0071] Step S42: Sampling ratio for each round As shown in formula (21):

[0072] (21)

[0073] Wherein, represents the current sampling round number, represents the total sampling round number. In the embodiments of the present invention, takes 12; Relative the variable monotonically increases, and the value range of

[0074] is between 0 and 1, and is used to represent the sampling ratio for each step in step S41;

[0075] Since a smaller number of trajectory points are selected in the initial sampling and the number of trajectory points is increased in subsequent samplings, in the early stage, due to fewer trajectory points, only a smaller number of trajectory points are generated to avoid introducing large errors due to insufficient trajectory information and reducing the trajectory generation quality. In the later stage, to reduce the number of steps for trajectory generation, it is necessary to increase the number of trajectory points generated in each step and reduce the running time. Using the sampling ratio designed in the embodiments of the present invention can meet the balance requirements of the generated trajectory quality and the generated trajectory time;

[0075] Step S43: According to the sampling ratio and in accordance with the sampling method in step S41, continuously sample from until the sum of the sampling ratios of the trajectory is 1. The number of generated trajectory points is equal to the number of original trajectory points, forming a complete trajectory sequence. All the trajectories obtained in the last step are taken out as a complete final trajectory.

[0076] In one embodiment, in the above step S5: Given the corresponding longitude and latitude coordinates of a specific person at a certain moment and inputting them into the position of the initial trajectory sequence corresponding to this moment. On the basis of step S4, after generating trajectory points in each step, cover the above-known corresponding longitude and latitude coordinates of the specific person at a certain moment among the generated trajectory points, and repeat step S4 for multiple steps to generate trajectory points until finally forming a complete final trajectory of the specific person. Repeat step S5 until a predetermined number of final trajectories of the specific person are generated. The set of the final trajectories is the urban-level large-scale crowd activity trajectories, specifically including:

[0077] Step S51: Set the residence and work place of specific personnel according to the corresponding tasks, generate large-scale crowd activity trajectories, set the spatial position of part of the evening time period in the input sequence of step S4 to the residence in advance, set the spatial position of part of the daytime time period to the residence, and set other positions as masks for subsequent trajectory generation;

[0078] Step S52: Based on step S4, After the wheel trajectory is generated, to The track of the time period is retained, and the tracks of the evening and daytime preset in step S51 are overlaid on the retained track. The overlaid track is used to continue to generate the i+1th round using step S4 until the sum of the sampling ratios of the multi-step tracks is 1. After the final complete track is obtained, the step of generating the track is terminated;

[0079] Step S53: Repeat steps S51 to S52 until the final trajectories of a predetermined number of specific persons are obtained.

Claims

1. A method for generating urban-level large-scale population activity trajectories based on mobile phone signaling, characterized in that, include: Step S1: obtaining the signaling data of a specific group of people in a selected city for one day, performing a preprocessing operation on the signaling data, and dividing the map area into grids to obtain a preprocessed trajectory sequence of equal length; Step S2: Construct a masked autoencoder model, including: a masking module, an encoder, a decoder, and a loss function module; First, use the masking module to set different masking rates greater than 0.5 for each of the preprocessed trajectory sequences, and mask each of the preprocessed trajectory sequences according to the masking rate to obtain masked trajectory sequences; Pass the masked trajectory sequences through the encoder and the decoder to obtain the probabilities of each trajectory point belonging to different grids ; Step S3: construct the loss function of trajectory accuracy and distribution diversity, and finally obtain the parameters of the pre-training stage model by reconstructing the training parameters of the mask trajectory points; Step S4: in the trajectory generation stage, the pre-training stage model is used to generate trajectory points in a step-by-step manner, with a certain number of trajectory points generated in each step until the obtained trajectory points form a complete final trajectory; Step S5: Given the corresponding longitude and latitude coordinates of a specific person at a certain moment, and inputting them into the initial trajectory sequence position corresponding to the moment, on the basis of step S4, after each step of generating trajectory points, the corresponding longitude and latitude coordinates of the above-mentioned specific person at a certain moment are covered in the generated trajectory points, and multiple steps of generating trajectory points are repeated according to step S4 until a complete final trajectory of the specific person is finally formed, and step S5 is repeated until the final trajectories of a predetermined number of specific persons are generated, and the collection of the final trajectories is the city-level large-scale crowd activity trajectory.

2. The method for generating urban-level large-scale population activity trajectories based on mobile phone signaling according to claim 1, wherein The said step S2: constructing a masked autoencoder model, including: a masking module, an encoder, a decoder and a loss function module; first, use the masking module to sample different masking rates greater than 0.5 for each of the preprocessed trajectory sequences, and mask each of the preprocessed trajectory sequences according to the masking rate to obtain masked trajectory sequences; pass the masked trajectory sequences through the encoder and the decoder to obtain the probabilities of each trajectory point belonging to different grids , specifically including: Step S21: Obtain a probability according to the Gaussian distribution p rob, sample from the preprocessed trajectory sequence and p perform a masking operation on rob× L different trajectory points, where is the length of the preprocessed trajectory sequence, as shown in Formulas (1) to (2): (1) (2) Among them, represents the trajectory sequence after masking processing, denotes the trajectory sequence in subject to the masking operation, and is the number of grids. Step S22: Define the encoding of each grid as , add the positional encoding , and finally map it to d -dimensional vector, as shown in formula (3), which uses the encoding of each grid plus the positional encoding in the sequence to represent the initial sequence: (3) Among them, The superscript 0 indicates that this is the encoding of the 0th layer; Indicates the positional encoding; Indicates the one-hot encoding; Step S23: Randomly discard some masked data at a ratio of , and the index set thereof is represented as , represents the data that is not discarded; (4) At concatenate an additional class vector at the beginning , which forms the input to the encoder: (5) Step S24: Feed into an encoder, which consists of four sub-layers in sequence: layer normalization, multi-head self-attention layer, layer normalization, and feed-forward neural network; first, perform layer normalization on the input of the layer and feed it into the multi-head self-attention layer as input to obtain ; then, use residual connection to add the input to the output of the multi-head self-attention layer to obtain ; go through layer normalization and the feed-forward neural network layer to obtain ; finally, after stacking the encoder layers, perform layer normalization again to achieve trajectory embedding encoding, as shown in Formulas (6) - (9): (6) (7) (8) (9) Among them, is layer normalization, is the multi-head self-attention layer, is the feed-forward neural network, is the output of the encoder; Step S25: Take input to the decoder module, and use the decoder to fill in the previously masked positions, map the space of the encoder to the space of the decoder, is the bias term of this mapping, represent the trajectory points not discarded, represent the encoding of the masked positions in prediction step S21, and then introduce a new positional encoding , and pass the current representation to a decoder with the same structure as the encoder but different parameters to obtain the output of the final decoder , as shown in formulas (10) - (14): (10) (11) (12) (13) (14) Step S26: Take and transform it through an additional linear layer , a GELU activation function, and a layer normalization to obtain the encoding . Multiply by to map it back to the grid, obtain the weight coefficients for each grid category, and add the weight coefficients to an additional bias . Then process it through a Softmax layer to obtain the probabilities that each trajectory point belongs to different , as shown in Formulas (15) to (17): (15) (16) (17)。 3. The method for generating urban-level large-scale population activity trajectories based on mobile phone signaling according to claim 2, wherein, The step S3: constructing the loss function of trajectory accuracy and distribution diversity, and finally obtaining the parameters of the pre-training stage model by reconstructing the training parameters of the mask trajectory points, specifically includes: Step S31: represents the likelihood probability that the i-th trajectory point belongs to different grids, represents the masked trajectory sequence, and uses the trajectory accuracy loss function to predict the true position of the masked trajectory point, as shown in formula (18): (18) Among them, represents the likelihood probability that the masked trajectory point belongs to the grid corresponding to the true position; Step S32: To obtain trajectories with more diverse distributions, a distribution diversity loss function is designed As shown in formula (19): (19) Step S33: Combine the above two different loss functions to obtain a loss function that comprehensively considers the accuracy and distribution diversity of the reconstructed trajectory as follows: (20) Among them, is a parameter; Step S34: Use the loss function Perform parameter training and backpropagation of the model to obtain the parameters of the updated model. The pre-training process ends, and the model parameters after pre-training are retained .

4. The method for generating urban-level large-scale population activity trajectories based on mobile phone signaling according to claim 3, wherein, The step S4: trajectory generation stage, using the pre-training stage model, adopts a method of gradually increasing the generation of trajectory points, and each step generates a certain number of trajectory points incrementally until the obtained trajectory points form a complete final trajectory, specifically including: Step S41: The pre-trained model parameters The defined conditional probability model, denoted as , and the generation process starts from . All the data initially input into the pre-trained model are masked and passed through the pre-trained model as the input of the first step to obtain the first generated trajectory sequence. A certain proportion of the trajectory points at the beginning of the sampling sequence are used as the trajectory points generated in the first step . Then, with as the condition, the subsequent trajectory points are masked and passed through the pre-trained model as the input of the second step to obtain the second generated trajectory sequence. More proportion of the trajectory point data at the beginning of the sampling sequence are used as the trajectory points generated in the second step. Sampling is performed from sampling , and so on. The sampling ratio for each step is calculated in step S42; Step S42: Sampling ratio per round As shown in formula (21): (21) Among them, represents the current sampling round number, represents the total sampling round number; relatively monotonically increases with the variable, whose value range is between 0 and 1 and is used to represent the sampling ratio of each step in step S41; Step S43: According to the sampling ratio , according to the sampling method of step S41, from continuously sample until the sum of the sampling ratios of the trajectory is 1. The number of generated trajectory points is equal to the number of original trajectory points, forming a complete trajectory sequence. Take out all the trajectories obtained in the last step as a complete final trajectory.

5. The method for generating urban-level large-scale population activity trajectories based on mobile phone signaling according to claim 4, wherein The step S5: given the corresponding longitude and latitude coordinates of a specific person at a certain moment, and inputting them into the initial trajectory sequence position corresponding to the moment, on the basis of step S4, after each step of generating trajectory points, the corresponding longitude and latitude coordinates of the specific person at a certain moment are covered in the generated trajectory points, and the trajectory points are generated repeatedly in multiple steps according to step S4 until a complete final trajectory of the specific person is finally formed, and step S5 is repeated until a predetermined number of final trajectories of the specific person are generated, and the set of the final trajectories is the city-level large-scale crowd activity trajectory, which specifically includes: Step S51: Set the residence and work place of specific personnel according to the corresponding tasks, generate large-scale crowd activity trajectories, set the spatial position of part of the evening time period in the input sequence of step S4 to the residence in advance, set the spatial position of part of the daytime time period to the residence, and set other positions as masks for subsequent trajectory generation; Step S52: On the basis of Step S4, after the -th round of trajectory generation, to the trajectories in the time period are retained, and the trajectories in the evening and daytime preset in Step S51 are overlaid on the retained trajectories. The overlaid trajectories are then used to continue the (i + 1)-th round of generation using Step S4 until the sum of the sampling ratios of the multi-step trajectories is 1. After obtaining the final complete trajectory, the steps for generating this trajectory are ended; Step S53: Repeat steps S51 to S52 until the final trajectories of a predetermined number of specific persons are obtained.

Images