Beijing 2008 
431 
Data using 
Processing 
Aug 24- 
STOCHASTIC MOTION AND THE LEVEL SET METHOD IN SEMI-AUTOMATIC 
BUILDING DETECTION 
ntation based on Qiu Zhen Ge a b *, Li. Qiong c , Zhang Chun Ling e , Xin Xian Hui d , Guo Zhang f 
Engineering and 
a Key Laboratory of Geo-informatics of State Bureau of Surveying and Mapping, Chinese Academy of surveying and 
mapping, Beijing, China, 100039 
b Institute of Computing Technology Chinese Academy of Sciences, Beijing, China, 100080- qiuzhenge@sina.com 
c China Institute of Geotechnical Investigation and Surveying, Beijing 100007, China - liq@cigis.com.cn 
d Tian Jin Institute of Hydro graphic surveying and charting ,Tian Jin, 30061, China - xin_xianhui@163.com 
c He Nan Bureau of Surveying and Mapping, Zheng Zhou 450052, China - zhangchunling06@sina.com 
f State Key Library of Information Engineering in Survey, Mapping and Remote Sensing,Wuhan University, Wuhan, 
430079,China 
- guozhang@lmars.whu.edu.cn 
Commission VI, WG III/4 
KEY WORDS: Active Contour, image segmentation, graph portioning, watershed method, weight dissimilarity 
ABSTRACT: 
Image segmentation is defined as partitioning an image into non-overlapping regions based on the intensity or texture. The active 
contour methods with comes from the basic ideas of Stochastic Motion and the Level Set Method provide an effective way for 
segmentation, in which the boundary of an object usually with large image gradient value is detected by an evolving curve. But, 
these methods have limitations due to the fact that real images may have objects with complex geometric structures and shapes, and 
are often corrupted by noise. Developing more robust and accurate active contour methods has been an active research area since the 
idea of the methods was proposed. In this paper, we propose a new active contour method and apply the method to remote sensing 
image segmentation. This new method uses combination of boundary-based modelling and region-based modelling. The new 
method is more efficient and effective, especially in detecting structures with noise. 
1. INTRODUCTIO 
Extraction of geospatial semantic information from digital 
images is one of the most complex and challenging tasks faced 
by computer vision and photogrammetry communities. 
Semantic information such as buildings.in particular is required 
for varieties of applications such as urban planning, creation of 
urban city models. 
For decades, manual extraction of semantic information in 
urban areas has been by conventional photogrammetry using 
aerial photos. This method is tedious, expensive, and requires 
well-trained people. 
In recent years we have experienced the lots of emergence of 
high-resolution space borne images, which have disclosed a 
large number of new opportunities for large-scale for urban 
areas topographic mapping, and high efficiently extraction of 
semantic information is a key technology problem, the most 
potential way to deal with it is semi automatic reorganization. 
The insight which we brought here is that there are many 
foreknow information about semantic information, especially 
about buildings, roads ect. We can use them to make 
recognition. 
Based on recent work on Stochastic Partial Differential 
Equations (SPDEs), this paper presents a simple and well- 
founded method to implement the stochastic evolution of a 
curve to detect structured and unstructured urban settlements 
areas from high-spatial resolution panchromatic imagery. 
In this paper we focus on building detection on high resolution 
remote sensing image, leading to what we call Stochastic 
Active Contours. The active contour method (ACM) for image 
segmentation was proposed in late 1980’s. In ACM, a curve is 
evolved towards the object boundary under a force, until it 
stops at the boundary in which the curve moves to minimize the 
energy. These various energy functions used in ACM can be 
roughly categorized as boundary-based modelling, region-based 
modelling or a combination of both. 
In boundary-based modelling, boundaries are characterized by a 
so-called “descriptor”: 
y(T)= 
The key idea is to find contour minimizing cost function J , at 
mean time the curve has the minimum length in the metric 
defined by function k . As the J is a line integral in the curve 
r so it only uses the information along the T . 
In region-based modelling region is characterized by a 
“descriptor”: 
Corresponding author. This is useful to know for communication with the appropriate person in cases with more than one author.