Research On Methods Of Indirect Measuring Pouring Flow Based On Computer Vision

Existing automatically pouring system of precision casting primarily detects full state of the pouring cup through the images of liquid level. But as a result of liquid metal can not be stable down flow, "full state of fake" image is often forms at the gate of the pouring cup, which result in miscarriage of justice of computer to stop pouring, produce wasters.This paper puts forward a new solution idea that the flow situation of liquid column of liquid metal is observed by the video images in the process of pouring, and then to judge the pouring situation of the casting. The key innovation is that the flow of pouring is gotten indirectly through the image of the liquid column.Two methods of measuring flow indirectly:the method of modeling based on the width of liquid column and the method of classification and recognition based on image of liquid column. Measuring the flow through the method of modeling based on the width of liquid column needs to establish mathematical model between the width of liquid column and the pouring flow. But because of limited field device, the level of liquid metal of ladle is unknown. So the mathematical models are more difficult to build.In the procession of measuring flow with the method of classification and recognition based on image of liquid column, firstly, carrying on the pretreatment to the real-time gathering image extracts the main characteristics of the image of fluid column; secondly images of the liquid column are described and classified by quantitative descriptor, such as the average width of the liquid column, invariant moments, etc., and then different types of the liquid column are distinguished through decision-making function; thirdly when experiment is repeated many times for every type of the liquid column, pouring time are collected at each calibration volume, and then the average flow is calculated with the straight line fitted by least square method, and random error of the average flow is reached by the error transfer function.Finally, the average flow are verified by simulation test. After period of time of pouring, the relative error between molten metal volume calculated actually and expectation volume is only 0.82%. In order to improve the accuracy, when the actual volume achieves 90% of the expectation volume, casting condition is detected through matching the image of the pouring cup. This method makes further improve of the degree of automation of gating system, reduces defects, and costs savings, so it has a certain practical value.