Research On Pill Box Location And Recognition Technology Based On Binocular Vision

With the rapid growth of the national economy and the rapid increase in people's demand for medical and health care,the variety and quantity of medicines have increased rapidly.How to effectively manage medicines,improve the efficiency and safety of pharmacy storage/collection,and reduce management costs are major A serious problem faced by the pharmacy of the Chinese Medicine Hospital.For this reason,the automated transformation of traditional pharmacies has become the development trend of modern hospitals.In this paper,the development of domestic and foreign automatic pharmacy and the typical automatic pharmacy mode are analyzed and compared,and the overall design scheme of robot automatic drug taking robot based on binocular vision is given.The theoretical methods involved in the identification and positioning of the target drug box are analyzed,studied and tested.The main contents are as follows:First of all,according to the structure of pharmacy and the characteristics of drug taking work,the overall structure and motion scheme of drug taking robot are designed,and the design requirements and technical parameters of drug taking robot are determined.On this basis,the design and selection of binocular vision system are completed,and the principle of parallel binocular vision measurement is analyzed,and the three-dimensional reconstruction formula is deduced.Secondly,using the pinhole camera imaging model,the object image coordinate system of the camera imaging is established,and the transformation relationship between the coordinate systems in the imaging process is analyzed.The binocular camera system is calibrated by Zhang Zhengyou's calibration method,and the vision system is calibrated by image acquisition of chessboard calibration boards with different positions.Then,we extract and segment the target region,match the target cartridge and reconstruct the three-dimensional image.Based on the projection of accumulated pixel value,the extraction method of target cartridge region is proposed.Canny edge detection technology is used to establish the edge point set of cartridge,and Turkey weight least square method is used to fit the straight line to complete the contour fitting of cartridge.The text on the surface of the medicine box is removed by using the connected domain technology,and then the stacked medicine box is segmented based on the horizontal projection of the accumulated pixel value,and the medicine box can be grabbed.The target medicine box and the pre stored image are matched based on the SURF feature point detection method,so as to realize the identification of the medicine box.Finally,on the basis of theoretical analysis,the software design of binocular measurement system is completed based on Open CV function library and Visual Studio 2015 as the development platform,and the binocular vision system is constructed for drug box positioning and recognition experiments.The results verify the feasibility and accuracy of binocular location and recognition.