Vision-Based Target Tracking For Unmanned Aerial Vehicle Surveillance
This Research paper provides knowledge about vision-based target tracking of faces using Python Open Computer Vision library which is coordinated by Raspberry Pi with an onboard affixed camera on the UAV. An Unmanned Ariel Vehicle (UAV) with flying capabilities was developed using KK2.1.5 Motherboard, Electronic Speed Controllers, Propellers of 4.5 degrees of pitch coordinating with 1000KV Brushless DC Motors for rotation, Camera with 1080p wide-angle lens, and various other required components. The UAV is controlled by a CT6B transmitter and receiver. The load characteristics of the UAV were simulated in Ansys Software. The Camera installed on the UAV coordinates with the program in the raspberry pi using the VNC Viewer application in the PC with the interface of the WiFi network. In the starting stage, the images are captured in numerical like 1,2,3 and so on; and in one of the programs a name can be set for each image file number. When the images are captured by the camera, they are converted into grayscale, cropped to the frontal face, and saved in the dataset folder. Local Binary Pattern Histogram (LBPH) Algorithm was used to develop this project. Using LBPH, the images were taken will be cropped to the perfect face in a rectangular shape which is then used for the tracking. This program is mainly used for known faces which the number can be set according to the program. The face will be recognized with the help of the “haarcascade_frontalface_default.xml” file that has all the required data for recognizing the frontal face. There will be a “TrainingData.yml” file which is saved in the Recognizer folder which consists of the data required to train an image to be tracked.