Illumination independent change detection for real world image sequences

Abstract

Change detection plays a very important role in many vision applications. Most change detection algorithms assume that the illumination on a scene will remain constant. Unfortunately, this assumption is not necessarily valid outside a well-controlled laboratory setting. The accuracy of existing algorithms diminishes significantly when confronted with image sequences in which the illumination is allowed to vary. In this note, we present two techniques for change detection that have been developed to deal with the more general scenario where illuination is not assumed to be constant. A detailed description of both new methods, the derivative model method and the shading model method, is provided. Results are presented for applying each of the techniques discussed to various image pairs. e 19x9 Academic press. hc.Detecting changes is fundamental to one's perception of the world. After all, the world we live in is dynamic, and the inputs to our senses are constantly changing. Many models have been proposed for detection of motion in the human visual system [13-171. In this note our concern is only change detection; motion detection may use change detection, but it is considered beyond the scope of this discussion. It is our aim is to develop robust change detection techniques for machine vision systems.It is not surprising that the process of change detection is fundamental to many machine vision applications. Systems that track moving objects [18], analyze cloud motion, mointor the growth of crops, or analyze traffic flow [2-41, are just a few examples of machine vision systems that use change detection algorithms. These algorithms provide the low level information that can be used by higher level algorithms to determine the information desired (the trajectory of an object, the growth of a tree, etc.). Therefore, for these systems to operate successfully, it is extremely important that change detection algorithms to accurate and robust.Change detection may take place either at the pixel level or at a higher level-by comparing features. We will address change detection at the pixel level. Our motivation in developing robust techniques for change detection at the pixel level is the possibility of very fast change detection for robotic applications.In this note, we present two new methods for change detection: the derivative model method and the shading model method. The derivative model method uses partial derivatives with respect to the pixel coordinates of a second order gray level surface model to compare regions and determine if a change has taken place. This

Change detection and background extraction by linear algebra

“…2) Object and Illumination Change: Skiftad [8] describes a test in which his algorithm would fail. In this test, a light gray ball is replaced by a dark gray ball of the exact same size and shape under the condition of constant illumination as illustrated in Fig.…”

“…In Section IV, state-of-the art models from Skifstad [8], Hsu [9], and Aach [10] are presented. These models can be categorized as follows.…”

A real time imaging system for assessment and control swine thermal comfort

“…Image processing techniques have been used in behavior-based thermal control systems (Geer et al, 1991;Wouters et al, 1990;Shao, 1997), and several image motion detection and segmentation methods have been developed (Yakimovsky et al, 1976;Skifstad et al, 1989;Gonzales et al, 1993;Hu and Xin, 2000). However, these methods focused on the feasibility and efficiency of image processing, computation time was not of direct concern due to their nature of off-line control.…”

“…There are several motion detection techniques in image processing field such as likelihood ratio method (Yakimovsky, Y., 1976), Fourier transform (Gonzales et al, 1993) and shading model methods (Skifstad et al, 1989;Hu and Xin, 2000) are two examples. Unlike image restoration problems, in our study we are interested in detecting animal movement in the captured image, rather than (ii) (12) are two binary images, % and y represent the pixel coordinates.…”

“…Skifstad et al, (1989) proposed a shading model method. This method was based on that image intensity is the product of illumination and surface shading coefficient, and in a small surface area, the surface shading coefficient and illumination are regarded constant.…”