We propose a novel solution of flexible cable handling by an industrial robot. Our method uses two different types of 3-D sensor and object recognition. The first one covers wide view area and measures cable using a motion stereo based method and estimates cable shape based on regression to the measured 3-D points. The second one uses our original 3-D sensor “Micro 3D” using structured light and focuses to narrow area to do highly accurate measurement of a connector attached at the end of the cable. We use depth edge detection and regression to the dense 3-D points of clouds for the pose estimation of connectors. To confirm the validity of proposed method, we constructed a prototype system in which a robot attached both 3-D sensors measures and handles the cables with branch structure to insert them at notched holding points of the kitting jig. We confirmed that our prototype system can assemble 200 sets of cables without termination caused by errors.