This paper firstly describes that the 3rd-order autocorrelation function for a picture of heaped chips is useful for identifying the feature on the shape of a separate chip. It has been actually shown that, for typical five kinds of chips with curled (bigger), curled (smaller), spiral (slim), spiral (thick) and powdery shapes, the 3rd-order autocorrelation function could extract the feature of each type of chips. In the experiment for each 10 sample images of the five kinds of chips, 49 samples in those 50 samples above-mentioned have been successfully identified. Lastly, a hardware for calculating the 3rd-order autocorrelation function is shown.