Abstract—This paper presents a new fast and robust goal detection system for the Nao humanoid player at the RoboCup standard platform league. The proposed methodology is done totally based on Artificial Vision, without additional sensors. First, the goals are detected by means of color based segmentation and geometrical image processing methods from the 2D images provided by the front camera mounted in the head of the Nao
robot. Then, once the goals have been recognized, the position of the robot with respect to the goal is obtained exploiting 3D geometric properties. The proposed system is validated with real images by emulating real RoboCup conditions. Index Terms—RoboCup and soccer robots, Artificial Vision and Robotics, Nao humanoid, Goal Detection, Color Segmentation, 3D geometry.

A. Detection Based on Geometrical Relations
The first proposed method is intended to be robust and fast in order to overcome some of the usual drawbacks of the vision systems in the RoboCup, such as the excessive dependency of the illumination and the play field conditions, the difficulty in the detection of the goal posts depending on geometrical aspects (rotations, scale,. . . ) of the images captured by the robots, or the excessive computational cost of robust solutions based on classical Artificial Vision techniques. The proposed
approach can be decomposed into different stages that are described in the next subsections.
1) Color calibration: The first stage of the proposed method consists of a color calibration process. Thus, a set of YUV images acquired from the front camera of the Nao robot is segmented into regions representing one color class each.
Fig.2 shows an example image captured by the Nao robot containing a blue goal.
The segmentation process is performed by using a k-means clustering algorithm, but considering all the available centroids as initial seeds. Thus, in fact, seven centroids are utilized, corresponding to the colors of the ball (orange), goals (yellow and blue), field (green), robots (red and blue) and lines (white).
sensors-13-14954f7-1024

The first problem addressed in this project is the segmentation of the image to separate the object (in this case the goal) from the background. As the RoboCup rules specify that goals are painted either sky blue or yellow, a simple segmentation solution based on color has been chosen.  It can be assumed that normally no other objects of similar colors will be present in the picture taken from the camera. If a similar color object is indeed present, it will corrupt the segmentation result, and depending on is size or shape it can render the method inoperative.

Therefore, proper tuning of the segmentation parameters is mandatory to avoid these situations. The following figure shows the result of an ideal segmentation of a camera image.

Ideal segmentation (right) of the goal in the camera image (left).Edge_detection

Edge detection
Once the image is segmented to separate the goal from the background, one can attempt to extract the edges of the resulting binary image. This is done to reducethenumberoflinesdetectedinthefollowingHoughtransformstage, while retaining the structural information of the goal.

Ideal edge detection (right) of the segmented image (left).

Edge_detection1

 

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s