In this short note, a method is presented to classify metal, wood and glass objects by using the measurement of light detection and ranging (LiDAR). It is found that those objects can be easily distinguished if the incident angle, which is formed by a laser bin and a normal vector of the reflection surface, is near zero. However, for practical application, it would be quite difficult to guarantee the zero-incident-angle condition. Therefore, a bunch of data from arbitrary view were collected and used for training and classification. Experiments show that the incident angle affect the reflection intensity of the LiDAR significantly and therefore cause the performance degradation of object recognition.