Image segmentation and classification is a tough when there are partial occlusion, reflection, and shadows in the image which increase false alarms. This paper describes the method for defining regions of interest in an image using ultrasonic sensors data which will be further used for image segmentation and classification. Sensors mounted in the view of the camera are calibrated for their position with respect to the camera to the relative distance between the human and robot. Depending on these calibration parameters, the image is divided into number of regions that is equivalent to number of sensors.