Visual-based Liquid Level Automatic Recognition And Tracking Design

Level measurement technology has been widely used in petrochemical,machinery manufacturing,industrial production,national defense science and technology construction and other fields.Traditional level measurement methods have some limitations.For example,some contact level gauges are not suitable for the measurement of corrosive and toxic liquid levels and some non-contact level gauges,such as ultrasonic level gauges,are greatly affected by fluid level fluctuations and are expensive to produce.Therefore,for the problem that the traditional liquid level detection method can not be effectively applied to some complex and high precision requirement occasions,many new liquid level measurement methods have been studied.At the same time,the liquid level measurement methods based on image processing stand out and develop well because of their good real-time performance,high measurement accuracy,strong environmental adaptability,long service life,safety,reliability,and convenience in cooperation with other devices.Firstly,the overall hardware system for liquid level automatic identification is designed,and then the liquid level identification algorithm is deeply studied.The liquid level recognition is mainly divided into image preprocessing,steel ruler scale line identification and transparent pipe water level line identification.The main work of this paper is as follows:1.The overall design of the vision-based liquid level automatic identification system is completed,including the design of the visual level measurement system of the visual level,the design of the liquid level automatic tracking system,and the selection and design of the main control console of the system.A liquid level automatic tracking system is specifically designed to provide feedback to the servo motor through real-time detection of the liquid level and control the CCD camera to achieve automatic tracking.The industrial requirements of the liquid level measurement system include: measurement accuracy,real-time recognition of frame rate of 45frame/s or more.2.The acquired level map needs to be pre-processed: grayscale processing of the image,filtering,denoising,and binarization are included.Bilateral filters are used during the pretreatment in this paper.And it is found that the local adaptive threshold segmentation method can effectively segment the scale area of the steel scale,which paves the way for the subsequent identification.3.In the recognition of the ruler scale value and the scale line,because the ruler scale value belongs to the rule print number,the digital identification algorithm of the fixed structure is studied,and it is concluded that the template matching algorithm has a unique advantage in this respect.Therefore,the digital recognition method based on template matching is adopted,the specific identification process is designed,and the experimental verification is completed,which shows excellent results.4.Due to the poor recognition of dynamic water level lines in complex environments by traditional methods,a dynamic water line recognition algorithm based on YOLO v2 is designed.And experiments has been conducted,which shows the higher accuracy,the better real-time performance.