How to Choose the Best Cloud Hosting Provider

Trajectory data is closely related to our life: as small as the indoor cleaning line of the sweeping robot, as large as the travel between provinces and continents, as short as a shared bicycle ride, as long as ten years of communication base station records. For urban governance, mining the movement patterns of people and vehicles can help decision makers better plan urban traffic, ensure public safety, respond to emergencies, and make cities smarter and more efficient.

Limitations of Classical Trajectory Representation

Trajectory data describes the time-dependent changes in the spatial position and properties of moving objects in the space-time dimension. During the movement of these moving objects (such as people, vehicles, etc.), their positions and other attributes will be recorded by the carried devices (such as mobile phones, sensors, etc.) at regular intervals, forming a sampling point data. For a moving object, the data of a plurality of sampling points are arranged in time sequence to constitute the trajectory data of the object.

However, there are many sources of trajectory data (including but not limited to mobile phone signaling data, vehicle GPS data, Wifi-sniffing data, check-in data, etc.) and new data is generated all the time, resulting in a huge volume of trajectory data, causing serious problems. Visual confusion and rendering pressure. Therefore, by transforming the trajectory data into flow field data with the help of visual analysis technology, we not only retain the main characteristics of the trajectory data, but also greatly reduce the data volume, so as to better gain insight into the movement patterns of people in the city.

When using visual analysis technology to analyze and mine trajectory data, a very important task is to visualize the trajectory on the interactive interface to provide users with space for observation and exploration. For the expression method of trajectory data, there has been some research in this field. The traditional methods include the path connection method and the flying line method. Among them, the path connection method connects the sampling points in the trajectory data of each object in chronological order, and then uses other visual channels, such as color, width and line shape, to encode other attributes of the object. This method is the most intuitive form of trajectory display. It can clearly show the spatial position of the path of moving objects. DataV uses this visualization in the "Linear Thermal Layer" subcomponent in the "Basic Plane Map" and "3D Map" components. form.

The flying line method is similar to the path connection method. The difference is that it simulates the movement of moving objects in the form of animation. In general, a line segment with an arrow is used to encode the moving object, and the line segments will be in chronological order between sampling points. Move, this method can more clearly reflect the moving direction of the moving object, and has a cool visual effect, the "flying line layer" in each map component of DataV and the "arc layer" in the "3D map" component, The "track layer" and "road network track layer" are the realizations of this visualization method. The lightning special effects in the "Lightning Map", which was stunningly unveiled on Double Eleven last year, were also realized by the flying line method.

Application of path connection method: Visualize the national logistics backbone network using the "Line Thermal Layer" component in DataV

Application of Flying Line Method: Using DataV to Visualize the Taxi Trajectory of Hangzhou in One Day

Application of flying line method: DataV double 11 national express "lightning" big screen, each lightning simulates the process of sending baby from seller to buyer along the real road network

The above two visualization forms are both classical trajectory data visualization methods, but they also have limitations. When these methods are applied to massive trajectory data, due to the amazing data volume, a large number of occlusions and overlaps appear in the original intuitive and clear visual expression. , if not handled properly, it will greatly affect the user's observation and exploration. In addition, the increase in data volume has brought huge drawing pressure, and users need to continuously improve the performance of hardware devices to cope, which invisibly raises the threshold for trajectory analysis. Therefore, for massive trajectory data, we need a more effective visualization method to gain insight into the movement laws of moving objects in cities.

Flow Field Generation Algorithm

After research, we propose a flow field generation algorithm for massive trajectory data, which can convert trajectory data in a specific time segment into flow field data, so as to express and describe "people flow" and "vehicle flow". The characteristic of this method is that it does not directly visualize the massive trajectory data, but aggregates the trajectory data to a certain extent, extracts the main features of the trajectory data, converts the trajectory data into flow field data, and then selects an appropriate visualization method. Display the flow field data. Since the flow field data retains the main features of the trajectory data and greatly reduces the data volume, it can clearly and intuitively reflect the movement rules of moving objects in the city while eliminating visual occlusion and reducing the pressure of drawing.

The main flow of the algorithm is shown in the following figure:

1) Statistical trajectory point vector

The trajectory data is composed of several sampling point data, and the sampling point data includes the position, time and other attribute information of the trajectory point. We first calculate the position of the trajectory points according to the data of all sampling points, and then calculate the direction and size of the in-out vector of each trajectory point according to the inflow and outflow between the two sampling points in the trajectory data, where the size also includes the number of trajectories and the movement. the speed of the object;

2) Filter the trajectory point vector

For all the in-out vectors of the trajectory points obtained in the previous step, the method will filter them according to the user-defined trajectory quantity threshold to obtain the main in-out vectors of each trajectory point;

3) Generate the main vector of the trajectory point

In this step, according to the user-defined flow field direction, the method classifies and aggregates all incoming and outgoing principal vectors of each trajectory point obtained in the previous step by direction, and generates at most one principal outgoing vector and at most one principal outgoing vector in each direction. input vector. While generating the main vectors in each direction, it is necessary to count the average speed, average moving distance and average difference angle of the vectors in each direction;

4) Diffusion trajectory point principal vector

Next, the n*m ​​grid is tiled into the area specified by the user, and the main vector of each trajectory point in each direction is spread into the n*m ​​grid according to certain conditions and rules. Among them, during diffusion, the vector, direction and velocity of the diffusion in the grid remain unchanged, and the number of trajectories decreases, and only if the following conditions are met, a grid will be affected by the radiation of a principal vector:

● The distance between the center of the diffused grid and the track point is not greater than the average moving distance of the track point;
● When the diffusion vector is the incoming vector, the angle between the vector formed by the center of the diffused grid and the trajectory point and the diffusion principal vector should be between [180-average difference angle, 180+average difference angle]; when the diffusion vector is When extracting the vector, the angle between the vector formed by the center of the diffused grid and the trajectory point and the diffuse principal vector should be between [-average difference angle, + average difference angle].
5) Calculate the grid principal vector

In the previous step, the same grid may be affected by the radiation of multiple vectors, resulting in multiple diffusion vectors. Therefore, in this step, it is necessary to calculate the aggregated vectors (including the flow field) in all directions in each grid. direction, moving speed and number of trajectories) to get the final flow field data.

In the method, the threshold of the number of trajectories, the direction of the flow field and the number of grids need to be defined. The trajectory number threshold is mainly used to filter the trajectory point vector. The filtering is to retain the main trajectory and prevent the "noise" from interfering with the accuracy of the result; while the flow field is divided into directions to avoid the offset of the movement in the opposite direction, so It can retain more detailed information, making the final result more accurate; and when defining the number of grids, it is necessary to balance the computational pressure brought by a large number of meshes and the rough effect brought by a small number of meshes. In order to make the method have a certain degree of adaptability, we did not use a fixed angle and distance when spreading the vector, we used the average moving distance and average difference angle in each direction, so that the diffusion can adapt to different vector distributions, and the result is more Reasonable.

Trajectory flow field case

The above is a brief process of the flow field generation algorithm we proposed for massive trajectory data. The following figure is the effect of the method based on the visualization of mobile phone signaling data from 8:00 to 8:10 in the morning of August 14, 2017 in a city:

In this case, we use particle flow to represent the flow field data. The number of particles in the grid is the number of trajectories (the number of moving objects), the moving direction of the particles represents the moving direction of the moving objects, and the color and speed of the particles both represent The moving speed of the moving object, where the higher the speed, the closer the color is to blue, and the lower the speed, the closer the color is to red. At the same time, we provide some controls to adjust parameters such as flow field direction, trajectory number threshold and grid number for interactive query. In this way, the amount of drawing is greatly reduced and the visual overlap is reduced, enabling a clear observation of the movement of people in the city.

DataV not only provides support and services for customers from all walks of life in the field of big data screens, but also actively explores the research on visual analysis of big data. Whether it is the recently launched smart venue solution or the flow field generation method for massive trajectories, they are all new attempts. We will not forget our original intention and continue to work hard to bring you better visualization products.