Esearch On Recognition Method Of Linear Pointer Instrument

Pointer instrument is widely used in military,medical,educational,industrial production and life because of its simple structure,convenient use,high reliability of identification,low cost and easy to be used in various environments.Although the development of digital display instrument technology is very rapid,the pointer instrument still occupies the main position in the industrial production process.At present,most of the acquisition of pointer meter data is manually recorded by humans,manually recorded and analyzed data,and the labor intensity is high,the efficiency is low,and the accuracy of the number of pointer meters identified is easily affected by external factors.Compared with the traditional way of acquiring instrument information manually,the application of computer vision technology greatly reduces the labor cost and improves the automation level of instrument detection,which has broad application prospects.At present,the automatic recognition task of pointer instrument is the focus on research in the field of image processing and pattern recognition.The key of recognition is how to extract the feature of pointer straight line quickly and accurately.Therefore,in order to avoid the above problems and improve the automation level of linear pointer instrument identification,a new method of pointer detection and indicator recognition is proposed in this thesis,the main research works are as follows:Firstly,image preprocessing.Before the image recognition work,the pre-processing operations such as grayscale processing,image filtering and edge detection should be performed on the recognized image,and the completed content and corresponding implementation effects of each part are analyzed in detail.Through comparative analysis,the appropriate algorithm is determined for the pre-processing process of automatic recognition of linear pointer instrument.Secondly,instrument pointer extraction and indicator recognition.The existing deficiencies in the existing recognition algorithms are analyzed.In the process of image feature extraction,the instrument pointers in the images cannot be extracted efficiently,which leads to the inefficiency of the recognition algorithm.Therefore,a method of pointer extraction for Freeman chain code instrument based on threshold processing is proposed in this thesis.Through the Freeman chain code operation on the image obtained after preprocessing,the instrument pointer feature in the image is obtained through threshold screening.The pointer deflection angle is calculated according to the direction parameters of the chain code,and the pointer indicator is identified by angle method.Through comparative analysis,the proposed algorithm is proved to be accurate,universal and efficient.Finally,correction and recognition of tilted linear pointer instrument image.Due to the limitation of some external conditions in the process of image acquisition,it is impossible to obtain the face image,which leads to the problem of image tilt.Various types of tilt are analyzed.Various types of tilt are analyzed.In the case of image distortion tilt,the feature points and matching points of the image are acquired,and the perspective transformation is used to correct the tilt once,so that the image becomes horizontal tilt.Then a new set of feature points and matching points are calculated according to the inclination characteristics of the instrument,and the perspective transformation is used for secondary correction to complete the inclination correction process of the pointer instrument.The different instrument images and related parameters verify that the proposed method can quickly correct the tilt image and accurately get the corresponding indication.