Object recognition based on GVF and SUSAN in Wireless Sensor Network

The role of image processing in image segmentation and object boundary recognition has been significant. It finds application in various fields including the emerging field of Wireless Sensor Network, where vision sensors are deployed. Vision sensors capture sequence of images from the sensor field and a sent to the base station for processing. Image Segmentation involves separation of the region of interest in an image, from its background, as a set of points forming contours. In Traditional Snake (Contour) algorithm, the boundary of image is considered as parametric curve. The process of finding an object boundary is an energy minimization process. In this work, a combined process of Gradient Vector Flow (GVF) Snake algorithm and Smallest Univalue Segment Assimilating Nucleus (SUSAN) approach has been implemented. Further, an Artificial Neural Network (ANN) system for object recognition based on the segmented image training has been implemented. The results obtained indicate that the combined approach of SUSAN and GVF Snake algorithm based segmentation process can improve GVF snake model´s precision to capture the boundary with sharp-angled corners. The combined approach gives larger capture range and stronger convergence ability to boundary concavities than the traditional snake method, thus increasing the accuracy and reliability of the system. Results also show that the proposed method is both time and space efficient.