Design And Implementation Of Tube Detection System Based On Image Processing Technologies

With the improvement of living standard,people had higher requirements for the intelligence and automation of medical laboratory equipment.Tubes are the most common carriers for the samples of our medical testing equipment,how to recognize all kinds of tubes which are widely used in medical testing,is the key to improve the testing speed,significantly reduce the operation errors,and lift up the level of device intelligence.The recognition technology based on image processing can replace the eyes of medical personnel and accurately identify the various attributes of the test tube.The specific work includes: selection of the most suitable lighting system;using the CMOS image sensor to collect tube image and transmitted to the host computer system;the image processing algorithms,including color to grayscale image,image de-noise,barrel distortion processing,edge detection,template matching,color recognition,white balance algorithm,image stitching,barcode search and recognition algorithm.Using these algorithm,we could identify the tube type,judge the status of the tube cap;identify the status of liquid level,and recognize the tube's barcodes.The designed system achieves the desired functionalities.The prototype testing shows that the system can recognize the various status of different tubes efficiently,and provide a solid technical foundation for realizing the intelligence of medical equipment.The advantages of the implemented system include:1)One image sensor can fulfill multiple functions,it can be used to replace barcode scanners and color sensors,etc.,so the costs can be reduced.The barcode recognition function module could decode the 1D barcode on the tube correctly,the correct rate is >99.9%.2)With the using of reliable image processing algorithm,intelligent recognition at the end of tube injector is achieved.3)The influence of ambient light is considered,the robustness of the system is improved.4)For liquid level detection,the use of test tube rotating mechanism,combined with the image stitching algorithm,even if the barcode partially covered tube,system can still accurately identify the liquid level,plus the bubble monitoring,reliability of the system was improved.5)The accuracy of the liquid level detection for the whole blood in tube is within +/-2mm.System could correctly recognize the color of tube cap,the correct rate is > 98%,including at least red,orange,yellow,green,cyan,blue,purple seven colors.