The initial motive of this paper is to recognize of algae which live in the river, in the lake, or in the sea and represent an important index of water quality. But one of the difficulties of recognition is due to their severe deformed shapes even in the same class. Therefore, the existing rigid templates are not appropriate for visual tasks such as recognition and representation for such a class of objects. For example flexible templates have been developed for recognition and segmentation of such objects, and "Active Shape Model", proposed by Cootes et al., is one of the templates. From their proposal, an object is defined by a set of points on the boundary and the statistical analysis for movements of such points acts as a constraint to the flexible template. The key point of this template is that any instance from the template can deform only in ways found in the training set. From this idea, recognition of an arbitrary posed object can be achieved by examining the degree of deformation of object from the mean shape representing the class. But in the Active Shape Model, it is emphasized that the landmark points on the contour must be placed manually, which is very time consuming. Therefore, this paper proposes "dual corner detection in scale-space" and dissimilarity function, and using this method it is possible to establish a deformable template automatically. This method is essential for recognition of unknown objects with arbitrary posed and various shape. In this way, automatic establishment of active model is achieved and experiment of recognition is done. The test result shows that this recognition performance is stable and its global recognition rate is 79.1%.