Research On The Key Technologies Of Detection Of Milk Somatic Cell Based On Computer Vision

Milk,one of the oldest natural beverages,known as "white blood",is important to human body.Milk somatic cell count(SCC)which is an important indicator of milk and udder health,not only re:flects the yield,quality of raw milk and health of the cow body,but also the management status of a region's dairy industry.The introduction of SCC detection to dairy industry in China is relatively late.Some scholars have already used image processing technology to detect the SCC,but there are still some disadvantages such as inaccuracy of the segmentation of cell image,low speed of operation.So,this thesis studies the key technologies of the detection of milk somatic cell based on computer vision.The main research works are summarized as follows:1.Before the system gets the accurate physical size of the somatic cell,the camera needs to be calibrated.After studying the calibration algorithms of computer vision,a new calibration algorithm based on Hough transform is proposed in this thesis.2.After obtaining images of somatic cells,every cell needs to be segmented out so as to identify the cells and count the number of the cells.Therefore,a cell segmentation method is presented.First the color images of the somatic cells are transformed into gray images.Then an image segmentation approach based on threshold is adopted.Finally the morphological processing is applied.3.Some appropriate features of image are selected to describe the milk somatic cells according to the characters of somatic cells.And one class support vector machine 1s adopted to train and recognize the cell images because only based on sample images of normal cells can one class support vector machine distinguish between the normal cells and abnormal cells.That avoids the impact of debris or other impurities on cell counting.4.In cooperation with a company in Hangzhou,a prototype of milk somatic cell counting apparatus based on computer vision is manufactured,which can automatically detect the somatic cell count in milk quickly.