| Summary: | 碩士 === 南台科技大學 === 資訊工程系 === 99 === According to much information, the methods of image tracking have already developed on many excellent algorithms and applied in various fields. However, how to quickly find the objects in the image is the key point. In this study, through the kernel-basis to find targets, Kernel function can Process some of the smooth area in image, exploring different shape and colors of objects. We discuss the complete degree in different kernel functions, for example, searching speed and difference between objects and goals.
In this study, we use different object images and use these objects as the base of searching, and then under complex background, different kernel functions are adopted to test, and the targets spread in different locations in the image might be found. The highest point of the kernel function is in the center and fall disastrously to zero. We calculate the color probability density of the object model and target. Then through the Bhattacharyya coefficient we test the similarity of the object and the target. Next we give a suitable threshold to find ideal object location. In the process, we use color three elements (RGB) to judge, respectively, and to get color similarity of the object, then to find the various objects in the image, which is favorable for us to find the object that we are interested in.
In this study we hope to use the simplest way to make pre-processing for tracking objects. This result can be applied to more applications, for example, tracking the unknown objects with camera and can more quickly detect in a monitor. In addition, the method can be used as the basis of tracking and it is favorable for future analysis.
|
|---|
