Strengthen accuracy of feature recognition via integration of ant colony detection and adaptive contour tracking

Reliable feature recognition is necessary in broad fields of computer vision and image processing. Edges often act as primary artifacts of visual data. Edge detection is to mark sharp changes of the intensity or brightness of digital images. Canny edge detection and ant colony optimization detection are two essential edge detection approaches. The former is susceptible to noises presented on source images. The information loss occurs when Gaussian smoothing is used to improve connectivity of Canny edge detection. Edges can be also detected via other approaches. To avoid edge suppression and feature deformity, ACO has been proposed for edge and contour detection against false detection, by which more intrinsic information will be extracted. The evolutionary computation oriented ACO scheme is a promising approach for feature capturing without the necessity of smoothing filters. It is among the most effective approaches for edge detection. However, it may give rise to broken pieces of numerous true edges occasionally. To further improve accuracy, contour tracking schemes are needed to achieve stable feature recognition. Some intelligent schemes are too complex to handle in real time, so a simple adaptive contour tracking scheme has been proposed which is combined with enhanced ACO schemes. This technology integration will result in the sufficient true edge representation together with well connected linkage, which can be easily extended to contour detection of binary, grayscale and true color images. Using quantitative metrics, an objective study is made to evaluate performance outcomes based on integration of the ACO schemes and adaptive contour tracking.