Anti-bounce Bidirectional K-level Tolerant Steel-bar Counting System Research

The vision-based online steel bar counting system co-researched by Lianyuan Iron and Steel Group CO. Ltd and Central South University has achieved considerable success in some bar plants. However, a number of technical issues needed to be addressed urgently has also revealed in the use of the real production process.For the low recognition rate in the bar counting system, first in order to solve the problem of uneven division caused by asymmetric illumination of the bar ends, the thesis uses severe fuzzy image of sequential filtering as a threshold image to segment image by multi-thresholds method and eliminate the shortcomings; second in order to solve the problem of unsuitability of burr in the center enhanced algorithm, the thesis uses moderate fuzzy to eliminate non-circularity before the algorithm starts and after the algorithm to erase flash burr to increase each block mass' independence and coherency; last to improve the drawback of classical bucket-clustering is not a best clustering and can not guarantee accomplishment of the clustering in the multiple clustering, the thesis propose to use bucket-clustering when crumby and use minimum-distance clustering and guarantee the accuracy of the steel bars' center.In order to solve the problem of bounce of steel bar generated by chain moves jiggly, the thesis proposes a synthesis method of projection matching and characteristic points pattern based on original method such as universe slippage matching, sectionalize slippage matching and integral matching and so on according to steel bar's counting feature. The test proves that these improvements can guarantee the accuracy.In order to solve the special situation of Chain bed two-way movement in some plants, the paper proposes bidirectional K-level counting model based on original K-level unidirectional counting model. When the chain bed moves forward the count value increases, otherwise the count value decreases. In this way it avoids the situation that the identical steel bar is counted repeatedly because of multiple passing the counting system.The test proves that the accuracy of recognition as well as the tracking of the on-line visual counting system has been improved and meanwhile achieves the bidirectional counting goal after improvement.