Pattern directionality is a key characteristic of landscape pattern

it both reflects and affects ecological flow within the landscape

and it is very important to realize fast and accurate auto-extraction of the pattern directionality information

especially for landscape heterogeneity analysis and driving forces analysis.Qian-an county lies in the northwest of Jilin Province in China

where the climate is characterized as temperate semiarid continental monsoon

the evaporation greatly exceeds the precipitation.Wind

especially the strongly prevailing wind in winter from northwest to southeast plays an important role in shaping some local landscapes.For most time of the year

the weather is extremely dry and the soil is sandy and uncovered.There are many lakes

known as Qian-an Group Lakes in Qian-an county.The lakes are typically small

and the water in the lakes is salty

seasonal and shallow

commonly about 1—2m deep.The lakes are mostly in the shape of a raindrop

with gently sloping lake beds in each lake’s northwest（the windward side）

and a steeply sloping lake beds in the southeast（the down wind side） with dune or loess piling up on this side.Wind is considered to be one of the most important forces driving the pattern formation of the lakes

i.e.

the current shape of the lakes is the result of windy erosion upon the tectonic basins.In order to provide a quantitative and visual basis for the above mentioned conclusion conveniently and objectively

based on object-oriented method Qian-an Group Lakes are extracted by multi-resolution image segmentation from Qian-an TM image.To achieve the high classification accuracy

a perimeter-area shape complexity metric called the Square Pixel Metric（SqP） is used

then vector analysis theory on landscape pattern（VATLP） is applied to the homogeneous objects to extract vector information of group lakes.Where

vector analysis theory on landscape pattern（VATLP） is developed by Shuqing Zhang et al.（2006） based on planar characteristics defined in mechanics such as centroid

moment of inertia

product of inertia and principal axes

and its application helps us better understand the pattern formation of the lakes and the related ecological processes that occur on both ends of the major principal axis

e.g.

the transportation of energy and materials flow

and the ecological gradients.The experimental results show that our method achieves the high segmentation accuracy（producer’s accuracy: 97.8%

user’s accuracy: 97.5%）

and extraction of directionality information is accurate and convenient.This can establish the groundwork for the further geographic analysis.This paper offered two-aspect idea: one is the use of landscape pattern metrics in image classification

the other is the further application of vector analysis theory on landscape pattern （VATLP）.