Object learning and detection using evolutionary deformable models for mobile robot navigation

Abstract

SUMMARYDeformable models have been studied in image analysis over the last decade and used for recognition of flexible or rigid templates under diverse viewing conditions. This article addresses the question of how to define a deformable model for a real-time color vision system for mobile robot navigation. Instead of receiving the detailed model definition from the user, the algorithm extracts and learns the information from each object automatically. How well a model represents the template that exists in the image is measured by an energy function. Its minimum corresponds to the model that best fits with the image and it is found by a genetic algorithm that handles the model deformation. At a later stage, if there is symbolic information inside the object, it is extracted and interpreted using a neural network. The resulting perception module has been integrated successfully in a complex navigation system. Various experimental results in real environments are presented in this article, showing the effectiveness and capacity of the system.

A survey on image segmentation using metaheuristic-based deformable models: state of the art and critical analysis

“…In terms of applications, a broad variety of problems has been tackled, ranging from face/body recognition [158,159,161,173,[175][176][177] and car/road-sign localization [155,174,178,182,183] to medical image segmentation [152-154, 156, 157, 160-162, 164-172, 179-181, 184], or mobile robot navigation [163]. In particular, in this section we have discussed Bayesian approaches [155,166,168,169], the online definition of a deformable model for mobile robot navigation [163], and approaches focused on the segmentation of objects using sophisticated metaheuristics (a generalization of the CMA-ES on vector spaces to Riemannian manifolds [153], as well as an evolutionary version of simulated annealing [167]), or the appropriate use of evolutionary algorithms in the solution of very difficult biomedical problems [154,157].…”

“…The approaches described in this section aim to optimize the positions of the control points of the deformable model, i.e., metaheuristics are employed to "guide" the movement and deformation of the deformable model [153,[155][156][157][158][159][160][161], to optimize/tune different parameters of the segmentation method [162,163], or, most frequently, to optimize the weights of the main modes of variation found in the training set and the parameters of a (usuallly affine) transformation [152,154,[164][165][166][167][168][169].…”

“…In [163] is placed over the model diagonals to exploit the frequent symmetries which can be observed in the objects used as landmarks.…”

LSM: A layer subdivision method for deformable object matching

“…This method is an extension of the free-form deformation (FFD) technique. More recent research can be referred to [3,9,11,16,18,21,25,26,28,30,31].…”

Decoupled Active Contour (DAC) for Boundary Detection

“…The task of locating exact boundaries of objects in cluttered and noisy environments has many applications in object tracking [1], content based image and video retrieval [2], [3], robotics [4], image composition [5], [6] and biomedical engineering [7], [8]. Energy minimizing splines, also known as deformable snakes or active contours [9], [10], [11], [12], [13], are the key approaches in the computer vision literature for such boundary extraction problems.…”