Abstract: To improve the on-line measurement speed of complex products' inner structure, according to its features this paper puts forward a method of feature compression and identification on the basis of direction projection. Firstly the image is projected and transformed into one-dimension projection. Then through differencing quantization, they are qualified into zero and one alphabetic string, which is looked as final identification features. Finally the products will be tested through KMP algorithm and NC matching method. The experiments prove that this method can characterize quickly with high efficiency, which is a new try in test and identification of complex products.