Optimising colour and texture features for real-time visual inspection

The role of colour descriptors has been demonstrated to be quite remarkable in many visual inspection tasks. In some other tasks, texture measurements are needed because of unevenly coloured or achromatic surfaces. In many applications, colour and texture must be combined to achieve good performance. At the same time, the computational complexity of the methods must be kept as low as possible. In this paper, a methodology for combining and optimising colour and texture features is proposed. Feature sets are optimised using a deterministic and a randomised approach. Results are demonstrated in detecting and recognising defect types on wooden surfaces.