Transportation Mode Recognition Based On AML-BP Algorithm

Transportation modes recognition is a new research field in intelligent transportation system, travelers’ transportation modes are figured out mainly through the data such as: speed and distance, travelers’ transportation modes can be used for investigating the transportation habit of people in some region, analysis their travel characteristics to help municipal decision-making, and helping to provide the most reasonable path for different transportation mode. According to the mobile Internet technology and machine learning, this paper propose a transportation mode recognition method which is based on AML-BP algorithm.This paper choose walking, taking bus, biking, car driving and electric motorcycle which are the main transportation modes in the city as the transportation mode we need to figure out. What we have done are as follows: first of all, this paper make an introduction to what others have done on transportation mode recognition topic at home and abroad, in the latter half of this section, we list the steps of our research procedure: the collection of transportation data, the analysis of transportation data, the research and improvement of the pattern recognition algorithm and the implementation of transportation mode recognition system; secondly, we develop an APP on Android platform to collect speed and accelerator data for each transportation mode; simultaneously, we use kinematics and statistics knowledge to analysis the collected data, select the travel features which can differentiate each transportation mode as the transportation mode parameter; following the preceding work, we use BP neural network as the transportation mode recognition algorithm, to address its drawbacks, we put forward the AML(Adaptive Momentum and Learning rate) based BP neural network, then we test the improved algorithm and make a comparison with some other algorithms; finally, basing on the research result, we use the Android platform technology, dynamic link library and the Baidu cloud service to develop an APP for transportation recognition on Android platform.With the characteristics of transportation modes and the improved BP Algorithm, this paper gets better result in transportation modes recognition with the comparison to some other methods. Then we apply this algorithm to an Android APP, with the real environment testing, it has stable operation and achieved the design requirements, the application has good recognition effect on transportation modes.