Overview

The package aims to detect the UAV during UAV automonous take-off and landing. In another word, it is to locate the the center of the UAV in each frame. An initial closed curve is set according to previous detection results, and then Chan-Vese model-based segmentation algorithm is employed to detect object region. Finally, with the segmented object region, the UAV center is located.

Quick Start

Parameters

Setup initial parameters in launch/CV_detection.launch

<param name="initial_x" type="int" value="320" />
<param name="initial_y" type="int" value="270" />
<param name="initial_r" type="int" value="20" />

The coordinate (initial_x, initial_y) is the center of the initial curve (circle) and the initial_r is the radius

Usage

Installation

cd ~/catkin_ws/src
git clone https://github.com/micros-uav/micros_cv_detection
cd ..
catkin_make

Running

Open a new consol and run a video node

Note:run this video node under ~/catkin_ws directory Run detection node also in ~/catkin_ws directory

source devel/setup.bash
roslaunch cv_detection cv_detection.launch

Result demo

https://cloud.githubusercontent.com/assets/11674154/9218708/384ea87e-4107-11e5-9685-ef04869a1113.png

* The image is a single frame of the UAV autonomous landing video

* The red circle represents initial curve

* The green point is the UAV detection result

Note:the package is inspired by and adapted from [1]. The details about the Chan-Vese model can be found in [2].

Reference

[1] D. Tang, T. Hu, L. Shen, D. Zhang and D. Zhou. Chan-Vese model based binocular visual object extraction for UAV autonomous take-off and landing. International Conference on Information Science and Technology, 2015, 67-73.

[2] T. F. Chan and L. A. Vese. Active contours without edges. IEEE Transactions on Image Processing, 2001, 10: 266-277.

Wiki: micros_cv_detection (last edited 2016-10-20 03:10:16 by tangdengqing)