Research On Object Recognition And Localization For Maintenance & Inspection Robot For Live Working

Live working is one of the important methods for testing,maintaining and improving electric power equipment,and it is of great importance to improve the reliability of power supply,reduce outage costs and ensure the safety of power network.Live working robot technology is one of the important trends in the research field of safety operation in live working,and vision guidance is the key technique to guarantee its autonomous operation.Based on the technical requirements of isolating circuit-breaker's power-on maintenance operation,object recognition and location method towards the metal fitting is the emphasis in the thesis.The main research contents are as follows:1.Based on the analysis of the working environment and the characteristics of the metal fitting,a vision system intended to operate under high voltage environment is proposed.Meanwhile the detailed system design and localization strategy is discussed.By experimental test,the equipotential bonding strategy proves feasible under high voltage.2.Most methods for the recognition of text-less objects are shape-based methods based on gradients or edges.However,the methods based on image gradient need a lot of templates while the edges-based methods have a high demand for the quality of edge detection.In this paper,a new method based on linear feature of edges is proposed,which has three contributions.Firstly,a scale-,rotation-and translation-invariant descriptor is created,which results in detecting objects with fewer templates.Secondly,the method has lower demand for the integrity of the edges.Finally,it can adapt to small deformation of the object.Extensive experimental results verify the effectiveness of the method.3.The industrial image often exists non-uniform illumination problems caused by poor light environment and especially when lighting disproportionation occurred on the metal fitting's surface.A robust image segmentation method based on the method of Otsu and the idea of local threshold is proposed to tackle those problems,where the calculation of the key variables is discussed in detail.Finally the location can be obtained by ellipse fitting and calculation.The test result shows that the new method can perform well to image segmentation in non-uniform illumination background situation,especially in the highlight reflection situation.And the method has a good positioning accuracy which meets the requirement of the subject.