Research On Classification Algorithms For Wooden Board Defects Based On Single Threshold Segmentation Of Artificial Bee Colony And SRC Theory

Wooden boards are widely used in wooden furniture and floor manufacturing fields.The defects of wooden boards,such as live knot,dead knot,crack and slipknot,are important factors that affecting the quality and grade of the material.In this case,how to effectively recognize and classify the defect has become the key to improve the utilization of the wooden boards.Traditional defect detection system mainly relies on artificial selection which based on the naked eye to recognize defect,the detection standard is not unified,which is also easily to be influenced by subjective factors.Design and develop the intelligent on-line detection and recognition method based on computer vision has a deep meaning for research and production.A wooden board defect detection method based on advanced artificial bee algorithms and SRC theory is proposed in this thesis.At the stage of detecting defects,advanced artificial bee colony algorithm is used to get the optimal segmentation threshold,which could realize both global and local search simultaneously during each iteration,aiming at such shortages as easily falling into local optimum situation,precocity and slow convergence,the artificial bee colony(ABC)algorithm was improved.After improved,the time-varied search parameter is added to the search equation of the follower,so that the search radius can be adjusted.When the local search is performed,the radius will decrease with the iteration,and when the local search is jumped out.The search radius will increase the adaptively.In addition,after the employed bees conversed into scouters,the strategy changed to give up the worst nectar selection,which could enhance the search ability.In the process of wooden board defects classification based on sparse representation-based classifier,the compress sensing theory is applied in defects classification field,which could transform the defect classification problem into the problem of obtaining the most sparse coefficient solution.It can decrease the computation and improve the classification accuracy.Experiments verified that improved artificial bee colony algorithm proposed in this paper,the search ability and the convergence precision has been improved obviously,which can avoid the disadvantage of easily falling into local optimum.The optimal segmentation threshold could achieve the segmentation of the wooden boards effectively.After principal component analysis,select several key feature vectors as the input of SRC classifier,accuracy rate can be improved to over 90%,which is reliable and feasible,and because SRC classification is based on a small amount,independent sparse vectors to achieve the classification,so it has good noise resistance,which can be widely used in the actual production.