This thesis presents an integrated approach of a vision sensor for bin picking. The vision system that has been devised consists of three major components. The first addresses the implementation of a bifocal range sensor which estimates the depth by measuring the relative blurring between two images captured with different focal settings. A key element in the success of this approach is that it overcomes some of the limitations that were associated with other related implementations and the experimental results indicate that the precision offered by the sensor discussed in this thesis is precise enough for a large variety of industrial applications. The second component deals with the implementation of an edge-based segmentation technique which is applied in order to detect the boundaries of the objects that define the scene. An important issue related to this segmentation technique consists of minimising the errors in the edge detected output, an operation that is carried out by analysing the information associated with the singular edge points. The last component addresses the object recognition and pose estimation using the information resulting from the application of the segmentation algorithm. The recognition stage consists of matching the primitives derived from the scene regions, while the pose estimation is addressed using an appearance-based approach augmented with a range data analysis. The developed system is suitable for real-time operation and in order to demonstrate the validity of the proposed approach it has been examined under varying real-world scenes.