非耕地系数反映了田间耕地与非耕地的分布情况，它的意义在于有助于获得较为准确的耕地面积。以影响非耕地系数的因素为指标对农区进行类型区划分是利用非耕地系数获取准确耕地面积的前提。首先以干旱区典型绿洲为研究区采用灰色关联方法探讨影响非耕地系数的主要因素，使用层次分析法综合影响非耕地系数的主要因素，并采用累计平方根法研究各个类型区分界点的确定。研究结果表明：影响非耕地系数的主要因素有土地利用的斑块破碎度、水网密度、农业人口密度、人均农业生产值等。将它们作为类型区划分的指标，并采用最优分层理论将类型区数目确定为8。利用累计平方根法确定各类型区的分界点，该方法使得分区结果较优，各类型区内的方差较小。该类型区划分结果反映了人们对土地利用的强度和各类型区的非耕地系数逐渐增加的趋势。

Non-cultivated land in cultivated areas is the key factor affecting monitoring cultivated land area using remote sensing technique on a large scale. In this paper, the conception of non-cultivated land coefficient was proposed to express the percentage of non-cultivated land in cultivated areas. Non-cultivated land coefficient shows the effects of human activities on the cultivated land. In addition, Precision of surveying results could be improved with non-cultivated land coefficient and the cultivated land area by remote sensing. Taking the typical agriculture area in the oasis in Yining County in Xinjiang as a study area, the paper designed 30 samples based on the Quick Bird data in 2008. Based on the platform of ArcGIS, spatial database was built to store the sampling data and calculate the non-cultivated land coefficient. It is time-saving to use the Quick Bird data to substitute for filed measurements. Firstly, the paper used grey correlation method to study the relationship between non-cultivated land coefficient and its influencing factors. Simplification of these factors was done by analytic hierarchy process. Based on the simplified factors, differences between type zones can be found to complete the agriculture division using accumulated ■ method. The results are as follows: the main factors influencing non-cultivated land coefficient were land use fragmentation, density of water network, agricultural population density and per-capita cultivated land. These factors were used as agricultural division index. In the paper, 8 zones are divided using the theory of optimal allocation. Finally, the results of agricultural division in the paper can meet the requirements of zone division that there is little difference between the samples in one division zone and quite obvious between two different division zones. Using accumulated method to find the thresholds of different type zones was very effective and convenient. The agricultural division, which reflects interference degree of human activities to land, is useful in guiding land use and land consolidation. Study on agricultural division based on non-cultivated land coefficient similitude in Xinjiang is quite useful for improving the cultivated areas by remote sensing.