The Algorithms Of Detecting Artificial Information In Natural Background

The detection of artificial information in natural background is a very important field in computer vision. It can be applied in many aspects.The algorithms of detecting artificial information in natural background are summarized here, including the methods of detecting artificial objects based on geometry feature, gray feature, polarization feature, fractal feature and probabilistic model. The methods of detecting artificial nature in natural nature background are introduced briefly as well.The algorithm of detecting artificial objects in natural background based on fractal feature is a kind of method based on background. Compared to the classical method based on target, the fractal method has a simpler procedure and a better performance in detecting, thus gets wide applications in this field.In this paper, grounded on the introduction of fractal model, several methods based on fractal feature are discussed in detail. It is the algorithm based on fractal dimension, the algorithm based on lacurity, the algorithm based on fractal fitting error, the fast algorithm based on fractal dimension and fractal fitting error, the algorithm based on fractal intercept and the algorithm based on multi-scale fractal feature. The algorithm of differential blanket counting and a little-target-detecting method based on fractal feature are introduced as well. But for the purpose of the application in reality, the algorithms based on fractal feature are still in need of deep investigation.In this paper, a new target-detection method based on color feature and fractal feature has been presented. Firstly, the algorithm of Unsupervised Optimal Fuzzy Clustering (UOFC) is used for color image segmentation. Because of the simplex and the uniformity in natural background, the regions with larger area are cleared out in order to quicken the speed of this method. Then the fractal feature is exacted from the left regions in order to remove the left natural background. In this way, the artificial objects are detected in the end. The experimental results show that this new target-detection method is accurate and efficient in detecting, especially in the field of multi-target detection in the complex background. It provides a new way to detect artificial objects in natural background.