Based On Random Forest And Kalman Filter Tracking Method Research Of The Human Body

Since therobot was invented, humanity’s greatest dream is to make robot be able to replace their own work in various areas of their lives. The invention of the service robots is to fulfill this dream. As a substitute for human, service robots should have the same capabilities as human.One of the most important capabilities is to identify and track the human body in complex environments.The current human body recognition technologies are:the human body thermal radiation identifying by infrared sensors, the human body recognition by machine vision. Due to the nature of the infrared sensors, the former one has shortcomings such as great interference by environmental background radiation, and because of relying on visual, the latter one is affected by light.With the maturity of laser radar technology, the price of the sensor is declining so that it becomes the standard of service robot configuration. Moreover not affected by environment makes it the sensor method of human body identification and tracking.This paper proposes a composite system which relies on laser radar as data sources, uses lower body statistical characteristics to train random forest classifier to identify laser radar data, and then uses Kalman Filter to track the human body lower limbs that have been identified. This paper mainly completed the following work:Firstly, it puts forward a method based on the laser radar information and random forest classifier body lower limb identification. This method considers laser radar feedback information as input medium, relies on the human body lower extremity to choose different statistical characteristics of age, sex, the clothes of the lower limbs laser radar data training a highly available random forest classifier. To use this classifier can identify more than99%of the lower body images. It forms a completion of human lower limb identification system, by matching the identified lower limbs according to the human body stride and feet wide.Secondly, according to the information of position provided by human body lower limbs recognition system, the paper implements a discrete Kalman Filter lower limb of the human body tracking system. This tracking system can effectively predict the position of lower limbs in the movement of human body so as to realize the human body tracking.This paper combining with the above two systems achieved a highly available service robot sports Human Movement Tracking System, using which in daily environment to complete all kinds ofexperiments about human boy tracking.