Circle detection on images using genetic algorithms

Characterization of multiphase systems occurring in fermentation processes is a time-consuming and tedious process when manual methods are used. This work describes a new semi-automatic methodology for the on-line assessment of diameters of oil drops and air bubbles occurring in a complex simulated fermentation broth. High-quality digital images were obtained from the interior of a mechanically stirred tank. These images were pre-processed to find segments of edges belonging to the objects of interest. The contours of air bubbles and oil drops were then reconstructed using an improved Hough transform algorithm which was tested in two, three and four-phase simulated fermentation model systems. The results were compared against those obtained manually by a trained observer, showing no significant statistical differences. The method was able to reduce the total processing time for the measurements of bubbles and drops in different systems by 21–50% and the manual intervention time for the segmentation procedure by 80–100%.

Semi-automatic image analysis methodology for the segmentation of bubbles and drops in complex dispersions occurring in bioreactors

Pattern recognition is one of the main research areas in the computer vision community. The basic problem consists in detecting and recognizing one or several known patterns in a data stream. Patterns can be specified in the raw data space or in any feature space suitable for the analysis of the input data. In particular, visual pattern recognition deals with applications where the pattern description is specified in terms of visual information and where input data comes from any kind of visual sensor (Chen & Wang, 2005). A common classification scheme divides pattern recognition according to the way the pattern is defined. There are then structural and statistical techniques. With respect to the problem solving approach, object location techniques are divided into two types of methods: i) deterministic methods like Hough transform, e.g. (Yuen et al., 1990), geometric hashing and template or model matching, e.g. (Iivarinen et al., 1997; Jones et al., 1990) and ii) stochastic techniques, including RANSAC (Fischer & Bolles, 1981), simulated annealing and genetic algorithms (Roth & Levine, 1994). Geometric shapes are very useful in a number of tasks because they are often present in human-made environments. They are also widely used as a part of man-designed symbols. To recognize this kind of shapes, many methods have been developed. In particular, circle and ellipse detection problems have been widely studied in the shape recognition community. Most approaches use Hough transform-based techniques. For instance, Lam and Yuen (Lam & Yuen, 1996) have proposed to use a hypothesis filtering approach to a Hough transform to detect circles in images. Yuen and Lo (Yuen & Lo, 1994) have posed the circle detection problem as a multi-resolution problem for the Hough transform. A traffic sign detector proposed by Mainzer uses a circle detector (Mainzer, 2002a,b). Shape classification using a soft computing approach is addressed by Rosin and Nyongesa (Rosin & Nyongesa, 2000). Genetic Algorithms (GA), proposed by Holland (Holland, 1975) in the 60s, are a family of algorithms where we apply an artificial evolution process. The purpose of the evolution is to produce a computational individual which is the best fitted to solve a specific problem. GA have been extensively used to solve optimization problems. Genetic Algorithms (GA) have been already used for several pattern classification tasks. In order to apply them to a pattern recognition problem, we need to pose the pattern recognition problem as an optimization problem. We need then to associate key elements in both approaches. Bandyopadhay

/pdf/parametric-circle-approximation-using-genetic-algorithms-1u2vrr83w7.pdf

Parametric Circle Approximation Using Genetic Algorithms

In this paper, the identification and classification of Viewer Age Range Smart Signs, designed by the Radio and Television Supreme Council of Turkey, to give age range information for the TV viewers, are realized. Therefore, the automatic detection at the broadcast will be possible, enabling the manufacturing of TV receivers which are sensible to these signs. The most important step at this process is the pattern recognition. Since the symbols that must be identified are circular, various circle detection techniques can be employed. In our study, first, two different circle segmentation methods for still images are analyzed, their advantages and drawbacks are discussed. A popular neural network structure called Multilayer Perceptron is employed for the classification. Afterwards, the same procedures are carried out for streaming video. All of the steps depicted above are realized on a standard PC.

/pdf/detection-and-classification-of-viewer-age-range-smart-signs-2ekuo0g293.pdf

Detection and Classification of Viewer Age Range Smart Signs at TV Broadcast

Many structures in aerospace, semiconductor and precision engineering are multi-layer in nature. Examples include Low
Temperature Co-Fire Ceramic (LTCC), PCBA, stacked IC, Through-Silicon-Via and composite materials for aircraft
wings. Segmentation of each internal layer in any orientation is essential for layer alignment as well as delamination,
disbond and warpage analysis. In this paper we propose a RDHT (Reduced Dimension Hough Transformation) for
automatic layer detection. Instead of segmenting internal surfaces at voxel level, correlation based edge operator is
applied to extract features in 3D space whereby the likelihood of any planar structure is associated with the number of
features on a specific plane. We use Randomized Hough Transform to map 3D features in three one dimensional
accumulators plus one verification accumulator to reduce Hough space dimension. The RDHT has been successfully
applied to various objects to reveal internal planar structures. For a CT result with a 512×512×512 cube, the feature
detection takes 30 seconds and the subsequent layer separation takes 12 seconds (laptop with Intel dual core 1.6G). We
demonstrate that the algorithm can segment all 16 layers of a stacked IC with an accuracy of 0.5 voxel.

X-ray CT image segmentation: automatic sandwich structure layer separation using reduced dimension Hough transformation

An efficient circle detection method using a coarse-to-fine strategy is proposed. The Hough transform is performed using a rough Hough space within a small window around a feature point which is selected using local connectivity information. Hough cells which have sufficient number of votes are retained, whilst all others are discarded. These Hough cells are refined using a finer quantization and then a larger window around the feature point is employed for another Hough transform. This process is repeated till sufficient evidence for the existence of a circle is obtained. >

E.K.H. Lo

Papers

A coarse-to-fine approach for circle detection