The single-angle feature coding identity authentication method cannot meet the current demand in terms of data capacity because of the increasing number of cattle. A cascade structure is used to detect the cattle's face and then estimate the angle of the cattle's face, which build a solid feature base for multi-angle feature coding. The result of experiments shows that the cascade structure can obtain a higher accuracy in both the face detection and attitude angle estimation tasks.