Optimizing Mathematical Morphology for Image Segmentation and Vision-based Path Planning in Robotic Environments

Abstract

Robotics advances have generated an increasing interest in new research projects and developments. Nowadays this science has several new applications characterized by working in non-structured dynamic environments. As a result, the research on this emerging area is growing, and, specially, vision algorithms are constantly being improved. In many cases, navigation needs real-time answers. As robots often work in dynamic environments, it would be desirable that the system takes a decision and applies it before external conditions change. Much work has been done on solving the problem of planning shortest paths between different locations within an environment (also known as a workspace) scattered with obstacles. For these solutions, the obstacles are usually considered as solid objects, and a collision-free path (of possibly shortest distance) must be found to navigate around them. However, not all path planning applications can be modelled as such a problem. On the other hand, Mathematical Morphology (MM) is a useful tool in image analysis, commonly used to extract components of the image, like contours, skeletons and convex forms. Although there are some approaches that take into account topographical maps in order for a robot to navigate through a workspace, few approaches actually deal with Mathematical Morphology operations. In this chapter we will focus on some of the research that we have completed in this field in the last few years. This way, two different robotic MM-based applications are discussed: • Path planning, which is strongly influenced by the precision of the acquisition process. Thus, it can be modified both by the quality of the information obtained from the environment, and the attributes of the system and the environment in which it works. Here, we shall refer to vision-based path planning. • Image segmentation, which is an essential part of any intelligent system, since it is necessary for further processing such as feature extraction or object and face recognition, among others. The research work here described has obtained very good experimental results and would contribute to the development of practical recognition and path planning systems. The use of vision improves the system, since once the visual information has been interpreted in