Accurate recognition and location of randomly placed industrial parts is the key technology to realized intelligent bin-picking of industrial robot. An approach to recognition and location of target object in randomly placed industrial parts using monocular camera. Efficient recognition of industrial parts with distinct profile could be realized when the geometric information based on 3D CAD model is used to train the temple. The most appropriate temple was obtained after the combination of a pyramid approach with hierarchical model views and the matching mode characterize by normalized similarity measure. The tested objective is a metal couple cage with distinct profile and reflective feature. Some experiments were carried out for randomly placed parts. Experimental results show that the target can be recognized accurately in randomly placed parts. The average location errors of target parts in X, Y and Z direction are 0.948 mm, 1.078 mm and 2.175 mm.