Font Size: a A A

Research On Insulator Nsdd Method Of Railway Catenary Based On Visible Image

Posted on:2020-03-21Degree:MasterType:Thesis
Country:ChinaCandidate:L SunFull Text:PDF
GTID:2392330623458136Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
Affected by external factors such as regional climate,environmental degradation and railway transportation conditions,the natural pollution of railway catenary insulators cannot be ignored.Under certain environmental conditions,excessive insulation of insulators will cause “pollution flashover” events,which will make damage to the railway catenary and result in an incalculable economic loss.Among the existing technical solutions or products applied to insulator maintenance work,charged water washing vehicles are valued and used by many railway bureaus because of their unique advantages such as simple operation,good cleaning effect and low working cost.However,charged water washing vehicles rely on manual operations,and there are still problems such as low work efficiency,insufficient manual detection accuracy,and safety problems under high voltage conditions.With the development and application of science and technology in various fields,as well as the improvement of the safety performance requirements of contact networks,the intelligent transformation of water washing vehicles is also urgent.In order to achieve automatic water flushing work,on-line detection of insulator contamination is essential.In the existing related research,most of the researches focus on improving the detection accuracy and lack of consideration of practical application.Most of the detection methods cannot obtain results before the "pollution flashover" accident occurs.Equivalent salt density and the non soluble deposit density are currently the most authoritative methods for describing the degree of contamination.The equivalent salt density detection method has high accuracy but is difficult to achieve online detection.The non soluble deposit density method has slightly lower detection accuracy,but due to its high practicality It is favored because of its stability,and because the focus of the non soluble deposit density method is visible dirt on the surface of the insulator,it can be well combined with the rapidly developing image processing technology.In view of this,an on-line detection method for the equivalent value of non soluble deposit density dense insulators based on indirect detection of visible light images is proposed,and some practical problems appearing in applications are solved in combination with actual industrial control to achieve high detection accuracy.The main research contents are as follows:1)Analyze the pollution status of the railway contact net insulator and combine theactual working conditions of the water washing vehicle,and select the visible light image method as the main method for automatic identification and detection.2)In view of the classification problem of insulators that are neglected in the current application of technology in this field,the model classification work before the contamination detection is proposed by using the visual word package model,the synonym distribution method is introduced,the sample is classified by SVM,and the classification is tested by experiment.Effect.3)When dealing with the problem of insulator contamination detection,extract the color feature quantity and the texture feature quantity for statistical calculation,perform feature fusion and dimensionality reduction to improve the classification learning effect,and calculate the best description of various typical insulators as reference pollution.Degree.4)Adding a fuzzy factor to form a double hypersphere data domain description method,forming a fuzzy interval,and adjusting the decision-making scheme to reduce the missed alarm rate caused by the second classification.At the same time,the false alarm samples are analyzed,and corresponding detection methods are developed for some special samples to reduce the false alarm rate.
Keywords/Search Tags:Insulator contamination, Visual Word Package Model, Color feature, Texture feature, Detection accuracy
PDF Full Text Request
Related items