Pixel-Based Colour Image Segmentation Using Support Vector Machine for Automatic Pipe Inspection
This paper presents a new approach to image segmentation of colour images for automatic pipe inspection. Pixel-based segmentation of colour images is carried out by a support vector machine (SVM) labelling pixels on the basis of local features. Segmentati
- PDF / 536,137 Bytes
- 5 Pages / 430 x 660 pts Page_size
- 90 Downloads / 162 Views
Abstract. This paper presents a new approach to image segmentation of colour images for automatic pipe inspection. Pixel-based segmentation of colour images is carried out by a support vector machine (SVM) labelling pixels on the basis of local features. Segmentation can be effected by this pixel labelling together with connected component labelling. The method has been tested using RGB, HSB, Gabor, local window and HS feature sets and is seen to work best with the HSB feature set.
1 Introduction In this paper we propose an approach to image segmentation for automatic pipe inspection based on colour images. The PIRAT system [1] was made up of an inspection device providing range images (in which the pixel values are distances) together with an interpretation system using neural networks and other AI techniques [2]. Here we consider inspection systems based on colour images. We assume that the inspection system has the ability to combine together all of the obtained images to obtain a single unwrapped pipe image such as that of which part is shown in Figure 1 (for example the Panoramo system of the Ibak company has this capability).
2 The Segmentation Method There are a number of approaches to colour image segmentation including histogram thresholding, feature based clustering, region-based approaches, edge detection approaches, fuzzy approaches and neural network approaches. The segmentation method used in the PIRAT system was successful. Therefore it seems reasonable to try and apply it to the case of colour images. This method used in PIRAT is a supervised method which means that the segmentation system is trained by using training examples which are labelled with their correct classifications. The method of segmentation involved feature extraction followed by classification of the feature vector [3]. In the PIRAT project differential M.A. Orgun and J. Thornton (Eds.): AI 2007, LNAI 4830, pp. 739–743, 2007. © Springer-Verlag Berlin Heidelberg 2007
740
J. Mashford, P. Davis, and M. Rahilly
Pipe joint Longitudinal crack
Pipe connection
Longitudinal direction Fig. 1. A section of an unwrapped pipe image
geometric features were used and the classifier was a nearest neighbour classifier. Differential geometric features are mainly only applicable to range images. Therefore other features will have to be used. A number of authors have used a feature vector consisting simply of the RGB (red, green and blue) values of a pixel as inputs to a classifier [4]. We have also used as a feature set the H, S and B components in the HSB (hue, saturation and brightness) colour space [5]. In order to obtain a segmentation system which is independent of lighting conditions we have also tried a feature set consisting of the H and S components of the HSB colour space. In addition, a slightly more sophisticated feature set has been implemented in which the max, min and average are computed for RGB values in a window (of size say 7x7) about each pixel together with the max, min and average of an intensity feature I given by I = √
Data Loading...
