A monocular approach to depth maps generation
In this paper, amonocularapproach for generation of the depthmaps of an object using computer vision is proposed. In this technique the object is immersed in a colour liquid and a single plan view is captured by a CCD camera. Due to the absorption property of the colour liquid, the intensity of the pixels of the captured image is modulated by the depths of the object along the camera axis. The depthmaps are generated from the grey level image by consecutive segmentation of the image into binary images at various thresholds. The depthmaps produced are parallel layers located in planes normal to the camera axis. The 2-D features of the contours contained in each layer are derived. The features of the layers are then combined to produce the 3-D model of the object. In order to employ the technique for object recognition a matching algorithm based on 3-D features of the object is developed. The technique is then verified by an experiment to recognize a set of objects. The approach can be used as long as the object is not damaged when placed in the colour liquid.