| The Steel Bars Automatic Counting and Splitting System made by Central South University for Liang Yuan Steel Group has been used in production line successfully. However the system can not cope with some special problems as follows: 1) target-occlusion problem caused by slanting and shrinking lying of steel bars; 2) target-unrecognized problem for dark bars; 3) status detecting problem when splitting at steel bars stacking position.To solve the above problems, this paper proposed multi-camera solution and projection-integral algorithm to detect splitting status quickly. In this solution steel bars are captured and recognized from top and front angles, then the recognized results are fused to solve occlusion and dark bars problem.Concerning steel bars recognizing, this paper focused on studying how to recognize bars from top view. The edges of bars end sect are firstly detected, and then a quick filter algorithm, in which the filter orientation can be adjusted, is adopted to filter noise. Finally edges are clustered to points as bars center using a bucket clustering algorithm.As regard to recognizing results fusion, this paper deduced mapping relation between pictures of two views, set down fusion rules, and set up a fusion math model according to that rules. Experiments validated the validity and feasibility of that fusion model. Finally this paper improved the K-time counting algorithm to rectify fusion result, aimed at solving misrecognitions.As to the bars splitting problem, this paper particularly presented the theory and steps of projection-integral algorithm to detect splitting status quickly, discussed the limitation of this algorithm, and how to copy with the limitation. Finally experiments from simulating system and real system are presented and show that the algorithm works well even in complex situation.Experiments show that the structure and algorithms designed in this paper are feasible and useful, which make good contribution to the system extending. |
PDF Full Text Request