An Intelligent Method for Comparing Shapes of Three-Dimensional Objects

The article discusses a method for comparing the shape of three-dimensional objects, which was developed at the Peter the Great St. Petersburg Polytechnic University. The main method is the consecutive double conversion of the original three-dimensional object to a vector of parameters that are invariant to the initial position, rotation, and size of the explored object. At the first step, three-dimensional objects are transformed by a three-dimensional version of the chord method into a matrix of distances between random points on the surface of the object. Then the matrix is converted into a vector that forms a histogram of distances. The resulting vector is summed up with additional indicators and compared in pairs with objects stored in the database using the Siamese neural network, which at the output gives an estimate for the proximity of the three-dimensional objects' shape. The developed method is the basis of a system for comparing the shape of three-dimensional industrial designs, and is also used in the detection of cancer of the lungs and prostate gland.