Extracting Image Features for Visual Servoing

button3Extracting Image Features for Visual Servoing

By: Mohd Faid Bin Yahya (PhD Student)


The basis of visual servoing requires 2 set of images: actual and desired. An actual image is an image which is acquired by a camera at the moment visual servo algorithm starts. Wherea desired image is an image where a robot have already stored in its memory. The robot has to adjust its current pose so as the error between the actual and desired images are minimized. In order to extract features in image, there are several ways to do it. In one way, there are 6 stages to perform so that the features can be extracted. Fig. 1 below shows diagram of all the stages.

3a oct15

Fig. 1.6 stages to extract image features.

Stage 1: acquiring image can be done using a camera. The camera must be able to provide good resolution and frames per second of at least 10.

Stage 2: usually, image captured by a camera comes in RGB format. The image needs to be converted to other format (for example HSV) so that specified color features in the image can be extracted.

Stage 3: the color range of the objects needs to be specified so that thresholding is realizeable by differentiating the background and the object features that the robot needs to separate.

Stage 4: once the features of the object had been separated, they will be noises left behind and these noises can be eliminated using morphology opening.

Stage 5: then, by calculating image moments, the radius and the center of the objects can be acquired.

Stage 6: finally, the features can be sorted so that it can be properly compared with desired image.

Fig. 2 below shows example of all 6 stages being implemented and Fig. 3 shows desired image. Finally, Fig. 4 shows the comparison between actual and desired image.

3b oct15

Fig. 2 All 6 stages in extracting features: (a) stage 1 (b) stage 2 (c) stage 3

(d) stage 4 (e) stage 5 (f) stage 6


3c oct15

Fig. 3 Desired image


3d oct15

Fig. 4 Compared image between desired and actual images