Abstract: To utilize the texture diversity of clouds and land in remotely sensed images, a novel describing method for image targets identification with complex background was proposed. The properties and classification of object image describing immune primitives was computed by the correlation between antibody property and the specificity of amino acid residues. The affinity formula of the training image-s immune primitives was presented by statistical analysis, which bears an analogy with the lowest amino acids combinative energy according to the biological immune antibody coding principle, to achieve the finite dimension object image features- optimize combination. Furthermore, The methodology was employed in the cloud contamination area detection. The cloud antibody has been configured and the cloud antibody has been tested on 200 images and more than 97% of results were correct, which proofed the validity of immune describing method for object image recognition in complex background.