Automating Snakes for Multiple Objects Detection

Abstract

We propose a novel snake initialization as well as validation technique to automate snake/active contour for multiple objects detection. We apply a probabilistic quad tree based approximate segmentation to find the regions of interest (ROI) in an image, evolve snakes within ROIs and finally classify the snakes into object and non-object classes using boosting. We propose a novel loss function for boosting that is more robust to outliers and we derive a modified Adaboost algorithm by minimizing the proposed loss function to achieve better classification results. Extensive experiments have been carried out on two datasets: one has importance in oil sand mining industry and the other one is significant in bio-medical engineering. Performances of both proposed snake initialization and validation have been compared with competitive methods. Results show that proposed algorithm is computationally less expensive and can delineate objects up to 30% more accurately as well as precisely.

Segment of Multiple Objects Based on Parameter Active Contour Model

“…The contour might contain an object if its reconstruction error is less than a given threshold. Reference [9] does not sow randomly in the image's global area, and determines the area in which the objects probably exist at first. Second, the seed points are sowed in the area.…”

“…In the second phase, segmentation of multiple objects in the image is executed by GVF. In [8][9], a seed point corresponds to one object, and this is the initial contour at the start of evolution. Then these seed points evolve into contours that contain objects under the expansion strategy.…”