Design And Implementation Of Urban Road Health Monitoring System Based On Group Intelligence Perception

Roads play a great role in the transportation of our country,and are critical infrastructures of national economy.The road health is important to the safety of transportation and the overall urban planning.However,the main exiting way of road health monitoring is using specialty instruments with high costs,which need relevant specialized personnels accomplish measuring spending lots of time and energy.This article proposes a road health judgement method using smartphone built-in sensors,getting road sensory datas from numerous vehicle driving on the road with the method of crowd Sensing.The main contributions of this article are as following:1.Propose two kinds of selection algorithm of sensors monitoring road health.Determine which sensors are used in raod health monitoring according to the sensor selection algorithm.The first algorithm is based on principal component analysis(PCA)dimensionality reduction.Firstly,solve the correlation coefficient matrix of the collected sensor data vectors,then solve its eigenvalues and corresponding eigenvectors,associate sensors with values of each feature eigenvector,fetch the-sensors of which corresponding eigenvectors values are bigger than threshold ?.Finally,intersect the sensors fetched each time,and get difference set of sensors selected under two kinds of road conditions.The second algorithm uses comparison method of variances differences of sensor datas,and fetches sensor datas that have greater influences under different road conditions.This method is simple and easy to implement.Sensors selected by the two algorithms are same with artificial selection.2.Using proposing method of variances differences,get an accuracy rate over 90%to testing road health.Propose a slope method based on this.The method mainly uses slopes changes of Z-axis datas of the selected accelerometer sensor to judge whether the road is flat.The method is real-time high efficiency,simple and high accuracy comparing with traditional methods,and can classify the testing datas 100%.