In order to progress robust object recognition of robots, we present an algorithm for object exploration based on three-axis tactile data that is necessary and sufficient for evaluation of contact phenomena. Object surface contour is gathered by controlling finger position to keep a constant the normal force measured by optical three-axis tactile sensors as two fingertips slide along the object surface. In this algorithm, when the robot grasps the object, time derivative of shearing force is checked to judge initial contact state because it is more sensitive than normal force. After contact between fingertips and the object, normal force is adjusted to keep a constant with a tolerant value between upper and lower thresholds. In the verification test, shape exploration experiments are conducted using sinusoidal shape hard object of wood. Experimental results show that this tactile sensor is capable of gathering the surface shape information of an object having concave or convex portion because the finger position controlled by normal force obtained by this tactile sensor follows the object surface.