GPS Track Stay Point And Stay Area Recognition Method

With the rapid development of the global positioning system(Global Positioning System)GPS technology,the trajectory data has increased a lot,making it possible to mine the deep semantic information contained in the trajectory information.As an important research part of trajectory data,trajectory recognition has a wide range of application scenarios in actual research.It can provide technical support for companies to implement map services,travel recommendations,etc.;provide government departments with a basis for decision-making to achieve urban traffic management,urban Planning and travel recommendations for residents.At present,many studies only focus on the recognition of trajectory points and the semantic recognition of trajectory points.They simply study the trajectory points and do not study the wider area;while in the area trajectory research,they only study a single type of trajectory data,and do not study mixed Type of trajectory data.In view of the above problems,this paper mainly conducts trajectory recognition research on staying point recognition and area trajectory recognition.The main research contents are as follows:(1)Identification of staying points.The past stay point recognition method usually uses spatial position information and time converted by the latitude and longitude of the track point to perform clustering.Since the data features used in clustering and constraining are separated,the effect in the constraining stage is not ideal.In this paper,aiming at the shortcomings of the recognition of staying points in the past,a staying point recognition algorithm based on spacetime constrained density clustering is proposed.The density clustering method using the same trajectory data feature speed and distance in the clustering stage and the constraining stage is The algorithm uses the trajectory data feature conditions such as speed and distance in the clustering part and the constraint part,which improves the previous situation of using only time and space for clustering and using other conditions for the constraint mode,so this can effectively handle the entire stay The point recognizes that some data features are separated.(2)Area trajectory recognition.On the basis of the recognition of staying points,deeper types of trajectory information are recognized.Deep-level trajectory data information requires a deep-level identification method.Because the trajectory feature data is one-dimensional and the trajectory data has time series features,the deep learning method based on Long Short Term Mermory network(LSTM)Track recognition.The LSTM deep learning method can obtain the previously learned content,effectively filter the previously useful information,and remove the redundant information before,so it can effectively grasp the context of the individual travel trajectory.The LSTM deep learning method has other different improved models.The improved models have their own unique advantages.For example,the LSTM derived model BiLSTM bidirectional model performs forward and backward learning at each BiLSTM layer,which can effectively obtain the delicate trajectory data before and after.Information,for the original model has the advantage of obtaining all the information features;GRU model is a simplified LSTM model,GRU has fewer logic gates,but without sacrificing prediction accuracy,it improves the efficiency of the model.In addition,the LSTM-based attention mechanism model can pay attention to key information features.When learning the LSTM model,it pays attention to the weights that affect the model more,and uses more resources on important features,thus improving the accuracy of model recognition.rate.In the experiment,GPS trajectory data was used for research,first,trajectory preprocessing,then recognition of staying points,and finally regional trajectory recognition.The experiment analyzes various methods and improves the research on this basis.The experimental results show that the method provided in this paper has high recognition accuracy.