Neural networks for color image segmentation: Application to sapwood assessment

This paper presents a method for detecting sapwood in hard wood such as cherry and maple. In the wood industry most applications need aesthetical boards. Thus the sapwood area on the board has to be detected as a defect region and be removed. To achieve this process, we classify the regions of the wood into two groups, by using neural networks techniques: sapwood is classified as a defect region while heartwood is considered as a good region. The use of neural networks by properly tuning the input vector provides a high defect detection rate with a low false positive rate.