Pedestrian Detection On Android System

With the development of computer technology and people's pursuit of a more convenient life,there are more and more application scenarios for pedestrian detection,such as pedestrian flow forecasting in scenic spots or shopping malls,personalized analys is of pedestrians,pedestrian traffic safety,self-driving cars,and finding older and children lost,etc.Under the rapid development of mobile equipment,more and more function is beginning to be developed in mobile devices.Thus,it becomes increasingly important to implement the pedestrian detection algorithm on the Android system.This thesis discussed a pedestrian detection algorithm implemented on Android system.The main innovations of this paper are as follows.First of all,this thes is introduces the Android operating system and specifically analyze the two Android vers ions used in this experiment.Describe the features of Android 4.4 and Android 6.0 systems and programming changes.This thesis presents the development of pedestrian detection applications on Android mobile phones and tests the applications as well.Two different Android phones were selected to install the application and test,finally thesis discussed the test results.Secondly,this thesis describes the features in detail which are often used in the field of pedestrian detection,inc luding Haar features,HOG features,and LBP features.After training and test over every feature,this thes is selects the better BPG features and LBP features to combine.BPG feature is a new feature discovered by scholars in recent years.Its essence is to extract the gradient features of the image.The LBP feature calculation method is used to calculate the BPG feature.The BPG feature,which preserves the characteristics of the HOG feature and combined with variable area statistics,makes it more appropriate to describe pedestrians in the image.This thesis describes the sliding window mechanism used in the detection process and the Adaboost algorithm is used to train classifier.In this thesis,the training process of cascaded classifiers is introduced in detail,and the structure of cascading classifiers is analyzed specifically.Meanwhile,specifically discussed components of weak classifier.Finally,this thesis implements the porting from C++ code to Java code.Image feature extraction and detection use C++ code.Android application development uses Java code.This thesis implements the application of C++ code in Java.At the same time,the OpenCV4 Android development environment is integrated in this experiment.By using the OpenCV library function,the amount of code in this artic le is greatly shortened.