Cloud adaboost feedback training machine for outside available parking spaces query service

A cloud based outside available parking spaces query service system is proposed. The concept of user pay the fee is considered to create the service system. The data available on the system database are collected from both of the active and passive ways. For the active way, detectors installed on a periodically routing vehicle will provide the data. For the passive way, the data is from the feedback of the system users. To increase to recognition rate, a detection method called bi-block size double hits method (BBDH) is proposed to get the rid of the mistaken target area of detection. The system classifier is composed of two strong classifiers based on Adaboost algorithm, each with different detection block sizes. The classifier will be retrained based on the more positive samples to improve the recognition rate of the whole system. The late version of the retrained classifier also used to update the old one on the detector of the routing vehicle. It is shown that the proposed method can improve the detection rate from 73% to 82% and the processing rate up to 20 fps. The entire system has the ability of learning and finally provides the public real-time parking location queries service, could be an effective location based service(LBS) solution to the problem for finding parking spaces of crowded urban areas in city.