An Approach for Picking T-Shape Workpiece Based on Monocular Vision

An approach for acquiring the distribution of top workpieces of T-shape workpiece based on monocular vision is proposed in this article. Monocular vision is awfully sensitive to the variation of profile caused by factors such as shelter, reflection and shade when it is applied to sort T-shape workpieces. The real integrated distribution of top workpieces can not be detected efficiently by using the whole profile as template. Using face feature of T-shape workpiece as recognition template is introduced to improve recognition rate and robustness of top workpiece recognition. In addition, we introduce a method to sort authenticity of recognized workpieces based on degree of aggregation and representativeness of face feature, which provides priority reference for eliminating repeated recognition and misrecognition generated by recognition. Then unreal workpieces can be eliminated by volume overlap ratio according to rank of authenticity. Finally, a test is performed on 50 groups randomly stacked T-shape workpieces, which shows that the designed top workpiece recognition method is superior to the traditional integrated profile recognition method.