Based on the sample space of farming area measurement method combining the advantages of remote sensing and sampling theory, spatial sample method has been widely applied for crop planting area. In the paper, a stratfied sampling method based on grid was introduced for such purposes. Firstly, with the 0-1 image, the effects of the factors of grid size, strata layer number on sampling accuracy, sampling variance and sampling ratio were analyzed. Secondly, binary image classification defining as result for crop area, random mix error was adopted for different kinds of abundance (10%, 20% ……, 100%) of winter wheat. The effects of the factors of grid size, strata layer number and classification error on sampling accuracy, sample ratio were analyzed. Finally, the layer number defined as six was brought up, and under classification error lower 40% was good for crop area estimation with spatial sample method. In a word, the research has developed a good method for optimizing spatial sampling.