https://github.com/lhiday-purdue/Project17-UAV-Ground-Detection-and-Tracking-Systems

Often a forest of trees has many different kinds of trees. In this project, we want to design a novel way to fly over a tree and determine what kind if tree it is, using a visual or biological sensor. Then, record the GPS point on the tree. At the end of a flight a map will show all of the GPS points for the specific variety of tree.

Team Member

Christian Ekeigwe -Purdue University
Daehyeon Jeong -Pusan National University
Jaeyoung Shim -Pusan National University
Jeonghwan Kang -Pusan National University
Seoungheong Jeong -Pusan National University

Projet Goal

This project aims to build a environmental operation system that can be to detect trees using machine learning object detection algorithm.

• OpenCV - One of the open source computer vision libraries is cross-platform and real-time image processing.
• Haar Cascade - It is an Object Detection Algorithm used to identify faces in an image or a real time video

System overview

• Use the Haar Cascades built into OpenCV to detect trees.
• Tree data set:It is a dataset for training the haar cascade model. There are a total of 107 positive images and 146 negative images. In the case of negative images, 1000 are recommended by adding additional images.detail
• Video capture program:This program is a program that captures images by dragging them from the video.detail
• Haar Cascade model:The xml of the learned cascade model is used. Through this, trees per frame are searched and the trees found are marked on the image and displayed as a video.detail
• When the object is detected, the raspberry pi stores the number of trees and the latitude and longitude received from the module from the GPS in MariaDB.
• Connect to the DB with another computer connected to the internal network to visualize the data.