An extraction algorithm based on color and texture was developed to realize segmentation between downy mildew and cucumber plants in greenhouse. A light analysis model was established by comparing the distribution of cucumber images in RGB, HIS and YCbCr color space, which was beneficial to the recognition of disease in variable illuminations. Combination model of Cband Cr elements was built based on the mean difference of lesions and interference informations in Cb and Cr, which included leaves, poles and soil, and extracted target rapidly. A detecting expriment was carried out on 30 images with downy mildew, which were taken in a changing greenhouse enviornment. The results indicated that the accuracy rate of segmentation was 90.6%. 