《基于像素的纹理特征聚类及分割算法》由会员分享，可在线阅读，更多相关《基于像素的纹理特征聚类及分割算法（7页珍藏版）》请在金锄头文库上搜索。
1、 - 1 - 中国中国科技论文在线科技论文在线 Texture Feature Clustering and Segmentation Algorithm based on Pixel Zhu Hong, Zhang Guoying (Dept of Computer Science and Technology, China University and Mining Technology(Beijing), Beijing 100083) Abstract: As the coverage area of the sub-block is too large, the texture fe
2、ature clustering based on sub-block often produces the mosaic phenomenon of inaccurate boundary. In this paper, the texture clustering algorithm based on pixel extracts the texture feature vector of its central pixel point from sub-block. The image texture feature is standardized, then the normalize
3、d feature vector is clustered and the clustering result is used to realize the segmentation of complex image. Compared with the traditional segmentation methods such as gradient method, threshold method and so on, the segmentation boundary is accurate, and the phenomenon of under-segmentation and ov
4、er-segmentation also reduce significantly. Keywords:texture feature;co-occurrence matrix;clustering;image segmentation 0 Introduction During the process of ore crushing, the correct statistics of stones particle size distribution directly affects the production costs and quality. The traditional sie
5、ves method causes large error, low precision and bad real-time performance. The laser method has high precision but cost much1-4. Using image segmentation technique to realize stone segmentation and particle size distribution statistics has the advantages of good real-time performance, high precisio
6、n, low costs and so on, so this research has important practical value5-14. Image segmentation is the core of obtaining the physical characteristic parameters such as stone size, stone distribution and so on in the process of ore crushing. But during the crushing process, the image of ore is very fu
7、zzy, stones pile together, the structure characteristics of the stone is complex, the different surfaces of the same stone have different brightness, the edge of the stone is blur and not regular, and shape and color of different stones are similar, so recognizing the edge of a stone among so many s
8、tones is very difficult and there has not been a suit of standard segmentation algorithm so far. The commonly used segmentation methods are gradient method, threshold method, regional segmentation method and watershed segmentation method. Gradient method12 is suitable for the segmentation of stones
9、which are smooth and have uniform illumination. Threshold method has perfect segmentation result only in the conditions of different stones having different brightness. Traditional watershed method is often affected by stacked stone and illumination and causes under-segmentation and over-segmentatio
10、n. For the stone of small size, the efficiency of watershed method reduces largely and the shape information is not reliable. The improved threshold method 5-7raised by Wang Rongben is suitable for recognizing single stone and is not suitable for several stones, and the recognizing precision is low
11、under conditions of varying illumination and complex background environment. Combining gradient and threshold methods, Wang Weixing8 does not detect the amount and sizes of the stone, but directly computes average diameter based on edge density. This method is not suitable for the stones of large am
12、ount and small average size. Yang Qiang13 raised the region growth method based on fuzzy membership grade of different scales. This method can obtain perfect seed extraction effect for segmentation of stones which have changing illumination and are not regular. Literature 15-16 adopts the segmentati
13、on method based on texture features, and is suitable for segmentation of stones which - 2 - 中国中国科技论文在线科技论文在线 pile together and have non-uniform illumination. But when extracting texture features based on sub-block, because the coverage area of sub-block is too large, this method often causes mosaic
14、phenomenon of inaccurate boundary during image segmentation. In this paper, the texture clustering algorithm based on pixel extracts the texture feature vector of its central pixel from sub-block. The image texture features are standardized, and then the normalized feature vector is clustered. Accor
15、ding to the clustering results, the stacked stone image which has complex structure features, non-uniform brightness can be disposed adaptively, and a clear segmentation boundary can also be obtained. Experimental results show that this segmentation method can segment the stone image on the conveyor
16、 belt accurately in the crushing process, and provide important particle size characteristics and its distribution parameters for crushing control. 1 Texture Feature Extraction of complex For image processing, the texture features commonly used are Tamura texture feature, autoregressive texture model, directional feature, wavelet transform, co-occurrence matrix and so on. The common point of these texture analyzing methods is extracting the most important feature of special tex
