3D objects classification and retrieval are a growing research topic with many application in different areas, such as robotics, virtual/augmented reality, medicine and physics. In robotics, 3D objects can serve as robust landmarks for pose estimation and localization, so robust 3D object recognition methods are very important tools in this field. Co-occurrence matrices were used for texture classification on 2D images and 3D medical images. This work uses co-occurrence matrices for object classification, considering their 3D shape, and the method was tested on the University of Washington RGB-D dataset. This work uses only the depth information, since our main goal is to provide a 3D object recognition method, other attributes (like colors and 2D textures) can be easily integrated to this approach to further improvements in the proposed method. The obtained results are comparable to the other works on the same dataset and the source code is available online.