Research On Wi-Fi Indoor Positioning Technology Based On Location Fingerprint

With the continuous improvement of people's pursuit of material living standards,mobile terminals are increasingly updated with the development of the Internet of Things,and the social division of labor is increasingly refined.People spend more than 70% of their time in indoor activities,and various indoor positioning technologies based on location perception are booming.Among the many indoor positioning technologies,the advantages of Wi-Fi indoor positioning technology are gradually recognized by people because the overall investment cost is lower,the operation is more convenient,and it is widely used in many indoor places such as office buildings,shopping malls,and libraries.Because Wi-Fi's indoor positioning technology has a meter-level error level,people continue to put forward higher requirements for the accuracy and function extension of Wi-Fi-based indoor positioning,and continuous improvement of positioning algorithms has become a research focus and problem.This paper mainly studies the content of Wi-Fi fingerprint database location technology based on convolutional neural network(CNN).This paper mainly studies from the following three aspects.At first the data is preprocessed,then the fingerprint library and online positioning algorithm are established,and finally the moving target detection results of YOLOv3 technology are combined to achieve personnel reduction steps.Because the validity of the offline fingerprint database lies in whether the data used is closer to the true value when establishing the offline fingerprint database,this paper first does a corresponding research on the data preprocessing.The data collected offline does not directly obtain the data available at a certain point through software,etc.,but directly obtains the original data in the server for further processing.On the premise that the original data is not modified,the data preprocessing is completed according to certain rules to minimize the proportion of data accuracy in positioning errors.This paper proposes a method for Wi-Fi positioning using two fingerprint libraries.The CNN is used to train the imaged RSSI values to achieve a model that can accurately implement classification based on the classification of fingerprint points to form a large-scale fingerprint database.We utilize the advantages of a large range of fingerprint database to determine the accuracy of the category,and limit the final location of Wi-Fi positioning with high accuracy around a fingerprint point.And then the "2?criteria + JNB classification + weighting" three-step data processing method is used to effectively remove Noise to form a small-scale fingerprint library.A small-scale fingerprint database is adopted to determine several candidate fingerprint points and increase the possibility of the final measured result at the real location.The two fingerprint library finally determine the Wi-Fi positioning coordinate result through the principle of "the distance is inversely proportional to the weight".The test results show that most positioning errors are between 0.5-1.5m.This paper will also apply the Wi-Fi indoor positioning coordinate results to the statistics of the number of people in the measured space.In the process of using Wi-Fi information to count MAC address types to count the number of people,there is a situation where one person carries multiple mobile terminals,etc.In this paper,the Wi-Fi positioning results and the YOLOv3 algorithm target detection results are combined to achieve the distinction of MAC address categories.The experimental results show that under the condition of less personnel overlap,the verification function of the MAC address category can be realized to accurately realize the counting of people and prevent the mistake of counting people based on the type of MAC address due to the influence of one person carrying multiple mobile terminals.