Зарегистрироваться
Восстановить пароль
FAQ по входу

Bhanu B., Lee S. Genetic Learning for Adaptive Image Segmentation

  • Файл формата pdf
  • размером 4,29 МБ
  • Добавлен пользователем
  • Отредактирован
Bhanu B., Lee S. Genetic Learning for Adaptive Image Segmentation
Springer, 1994. — 209.
Image segmentation is an old and difficult problem. It refers to the partitioning of an image into meaningful components. Generally, it is the first task of any automated image understanding process. All subsequent tasks, such as feature extraction, object detection, and object recognition, rely heavily on the quality of the segmentation.
Currently, there are a large number of segmentation techniques that are available. However, these techniques rarely demonstrate the robustness required for practical applications of image understanding, such as autonomous vehicle navigation, target recognition, photointerpretation, etc. The difficulty arises since the segmentation performance needs to be adapted to the changes in image quality. Image quality is affected by variations in environmental conditions, imaging devices, time of day, etc. Thus, one of the fundamental weaknesses of current image segmentation algorithms is their inability to adapt the segmentation process as real-world changes are reflected in the image.
Only after numerous modifications to an algorithm's control parameters can any current image segmentation technique be used to handle the diversity of images encountered in real-world applications. While there are threshold selection techniques which adapt to local image properties in a single image for image segmentation, these techniques do not adapt local thresholds from frame to frame so as to compensate for changes in images caused by variations in the environmental conditions. Also, they do not accomplish any learning from experience to improve the performance of the system over time. To date, no segmentation algorithm has been developed which can automatically generate an "ideal" segmentation result in one pass (or in an open loop manner) over a range of scenarios encountered in practical applications. Any technique, no matter how "sophisticated" it may be, will eventually yield poor performance if it does not adapt to the environmental variations. Therefore, in this research we attempt to address this fundamental limitation in developing "useful" computer vision systems for practical scenarios by developing a closed-loop system which automatically adjusts the performance of the segmentation algorithm. The system is based on changing the control parameters of the segmentation algorithm such that it will be operational across a wide diversity of image characteristics and application scenarios. It is noted that the performance of the adaptive segmentation system is limited by the capabilities of the segmentation algorithm, but the results will be optimal for a given image based on the evaluation criteria that have been defined.
This book presents the first closed-loop image segmentation system that incorporates genetic algorithms to adapt the segmentation process to changes in image characteristics caused by variable environmental conditions, such as time of day, time of year, weather, etc. The goals of the adaptive image segmentation system presented in this book are to provide continuous adaptation to normal environmental variations, to exhibit learning capabilities, and to provide robust performance when interacting with a dynamic environment.
The research is directed towards adapting the performance of a well known existing segmentation algorithm (Phoenix) across a wide variety of environmental conditions which cause changes in the image characteristics. The book presents a large number of experimental results that demonstrate (a) the ability to adapt the segmentation performance in both indoor and outdoor color imagery, and (b) that learning from experience can be used to improve the segmentation perf()rmance over time.
Introduction
Image Segmentation Techniques
Segmentation as an Optimization Problem
Baseline Adaptive Image Segmentation Using a Genetic Algorithm
Basic Experimental Results – Indoor Imagery
Basic Experimental Results – Outdoor Imagery
Evaluating the Effectiveness of the Baseline Technique – Further Experiments
Hybrid Search Scheme for Adaptive Image Segmentation
Simultaneous Optimization of Global and Local Evaluation Measures
Summary
  • Чтобы скачать этот файл зарегистрируйтесь и/или войдите на сайт используя форму сверху.
  • Регистрация