In order to recognize the random crack in the X film image of the metal cast automated,this paper proposes a new algorithm to recognize the crack of metal.To realize the method,it constructs the ideal probability model according to the character of crack image and transform the model to the ideal template.It computes the Bhattacharyya coefficient between the ideal template and the original image in order to construct the Bhattacharyya coefficient matrix and transform the matrix to a grayscale image.The complete and accurate outline of the crack can be got by morphological operating.Experimental result shows the outcome of the algorithm is both adaptive and accurate.