Automatic segmentation of pallet images using the 2-D wavelet transform and YUV color space

Segmentation is an important stage in automatic digital image processing. A special case of segmentation is to segment objects from their background. Among different segmentation algorithm for object detection, learning based approach is widely applied. In steel industry, pallets are moving on a rail. They have high resolution details in their structure and the image of a pallet taken by a camera in real time suffers from severe noise and illumination variations. The purpose of this paper is to segment the pallet from a frame of a sequence of video images, such that the pallet is segmented without degradation of resolution. We use the pallet image in YUV color space together with wavelet transform (WT) for detection. For classification Support Vector Machine (SVM) is incorporated to the images. It is shown that the above procedure segments the pallets successfully without degradation of resolution.