[attachment=14377]
ABSTRACT
Machine vision or computer vision is the study of methods that machines understand multi-dimensional data, such as images or a sequence of images, three-dimensional volume data, and single to multi-variable signals. Whereas computer vision is mainly focused on machine-based image processing, machine vision most often requires also digital input/output devices and computer networks to control other manufacturing equipment such as robotic arms. Machine Vision is a subfield of engineering that encompasses computer science, optics, mechanical engineering, and industrial automation. Many fields can be related to machine vision, such as mathematics, physics, artificial intelligence, image processing, biology, signal processing, etc.; the application of this field is diverse. In this paper, we propose a robust and efficient system for model based object recognition based on wavelets and multiresolution analysis. This framework is suitable for autonomous vehicle navigation. Since our approach is simple and effective, it can be used in real time processing or expanded for more complex applications.
1. INTRODUCTION
Computer imaging has been an active area of research since the 1960s [1]. However, limitations of system resources hindered the development until more advanced technology could be introduced. In the 1960s researchers from Massachusetts Institute of Technology, Bell Labs, and the University of Maryland started to apply their techniques in medical imaging, satellite imagery, photo processing, and other basic imagery manipulation. In the 1970s, the first work with edge detection was done followed closely by physics-based analysis starting in the 70s and extending into the 80s. The 1980s also featured the first work done in processing images based on motion; however, it wasn’t until the 1990s and 2000s until real-time, interactive computer vision could be manipulated [8, 9].
Computer imaging can be separated into two different but overlapping fields: computer vision and image processing. A few prominent applications are medical image processing, automatic fault detection, the manufacturing process, military applications, robotics, autonomous vehicle navigational systems, and space exploration. The purpose of medical image processing is to extract information from images, such as tomography or MRI images for diagnosis purpose. For example, this would extract information about tumors or organs so that medical professionals can diagnose the patient’s condition. One example of computer vision is battlefield awareness. Various sensors can be used to make decisions about enemies’ position that can be used to make strategic decisions.
It has many potential applications in real life problems such as autonomous car driving, military autonomous vehicle navigation, aircraft landing, space exploration, and forest fire and rescue systems. Recently, four robotic vehicles succeeded in making it to the finish line in a race across the Mojave Desert sponsored by the Pentagon. These vehicles have set a technological landmark in autonomous vehicle navigation. In this paper, we provide a novel approach for a robust and efficient system for model based object recognition based on wavelets and multiresolution analysis. Since our method is robust and efficient, it can be implemented in real time.
Following this introduction, we provide a brief background on Multiresolution Analysis. Section 3 is our methodology, section 4 contains the results, and section 5 is our conclusion.
2. MULTIRESOLUTION ANALSYIS
In the last decade, the wavelet transform has become a cutting edge technology in the signal and image processing field because of its simplicity and elegance of the notion of scales [11]. We can slightly expand the wavelet concept into multiresolution analysis which is important for our task, such as image segmentation, object recognition, and other preprocessing techniques. Further, these tasks can be performed in a successive approximation manner starting on the coarse version until the finer version