Point cloud filter method and apparatus

A method comprising:

• receiving a point cloud data set, the point cloud data set comprising individual points Pi;
• removing redundancies in the point cloud data set;
• storing the point cloud data set in a filter cache;
• for every point Pi in the point cloud data set, determining a plurality of points Pj that belong to a defined area A around a respective one of the individual points Pi;
• determining a covariance matrix S using the plurality of points Pj;
• determining from the plurality of points Pj, using the covariance matrix S, one or more eigenvectors and one or more eigenvalues corresponding thereto;
• generating a position correction vector Fi by applying a first filtering function using a Gaussian filtration for each point Pi;
• generating a revised position correction vector Gi by applying a second filtering function, the second filtering function defined by a one direction filtering which is perpendicular to a surface represented by the point cloud data set;
• determining, using the revised position correction vector Gi, if all the individual points Pi in the defined area A represent a surface and
• if so, applying a surface detection function followed by applying a correction factor to the individual points Pi to generate a filtered point cloud data set;
• and
• otherwise, applying the correction factor to individual points Pi to generate the filtered point cloud data set.