A Method for Early Prediction of Pedestrian Flow Based on Spatiotemporal Graph Convolution
A prediction method and pedestrian flow technology, applied in the field of pedestrian flow prediction, to achieve the effect of high prediction accuracy
Active Publication Date: 2022-04-19
HANGZHOU INNOVATION RES INST OF BEIJING UNIV OF AERONAUTICS & ASTRONAUTICS
View PDF6 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0005] In order to solve the problem that the current pedestrian flow forecast can only be trained and predicted by setting a fixed step size of historical data in advance, the present invention provides an early prediction method of pedestrian flow based on spatio-temporal graph convolution, which can not only be based on Historical data adaptively selects an appropriate step size for prediction before prediction, and the prediction accuracy is higher than that of a model with a fixed step size
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0021] The present invention will be further elaborated and described below in combination with specific embodiments. The technical features of the various implementations in the present invention can be combined accordingly on the premise that there is no conflict with each other.
[0022] The scheme adopted by the present invention to solve its problems will combine figure 1 To illustrate: replace the matrix multiplication in the gated recurrent unit with the diffusion convolution of the undirected graph, input the historical data into the encoder composed of the gated recurrent unit to extract features, use the hidden state output by the encoder to initialize the decoder and perform Iterate to get the final hidden state as the result output. The controller regards the result as immutable data and maps it to the stop probability, and establishes a Bernoulli distribution according to the stop probability, and controls whether the step size of the required historical data sho...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses a method for early prediction of pedestrian flow based on spatiotemporal graph convolution, which comprises: acquiring training samples of pedestrian flow, constructing an encoder-decoder structure applying undirected graph diffusion convolution, constructing and controlling the step size of historical data The controller for iterative training of a spatiotemporal graph convolution-based early prediction model: obtaining the results of early prediction of pedestrian flow. In the present invention, the real-time data samples collected by the sensor are input into the model, the sample features are extracted by the encoder, the output of the encoder is used to initialize the decoder, and the controller judges whether the current moment satisfies the optimal step size according to the hidden state of the final output of the decoder. If the conditions are met, the final predicted value is obtained by the fully connected layer, that is, the number of pedestrians passing through each sensor area in a certain period of time in the future. The present invention not only can adaptively select a suitable step size for prediction before prediction according to historical data, but also has a higher prediction accuracy than a model with a fixed step size.
Description
technical field [0001] The invention belongs to the field of pedestrian flow prediction, and relates to an early prediction method of pedestrian flow based on spatio-temporal graph convolution. The method can adaptively select a step size of appropriate historical data for early prediction of pedestrian flow. Background technique [0002] Today, there is an increasing need for effective regulation and evacuation of pedestrian flow in public areas. Without proper guidance and regulation, congestion can occur as pedestrians gradually gather. Especially in emergency situations, there may be some competition between pedestrians, which may turn into a stampede. Therefore, research on pedestrian flow prediction is of great significance to public crowd safety. Pedestrian flow prediction is to predict the number of pedestrians passing by the sensor within a certain period of time. [0003] Considering that nodes with similar Euclidean distances may have greater differences in pede...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06Q10/04G06Q50/26G06N3/04G06N3/08
CPCG06Q10/04G06Q50/26G06N3/08G06N3/044G06N3/045
Inventor 邵巍周家柳彭紫妍罗喜伶
Owner HANGZHOU INNOVATION RES INST OF BEIJING UNIV OF AERONAUTICS & ASTRONAUTICS