Research On Target Recognition And Measurement Algorithm Based On Machine Vision

The so-called osmotic pump controlled-release (OPCR) tablet is a new kind of drugs where the drugs are released constantly from the Drug-release hole. The effect of the drugs is directly concerned with the speed of the drug-release, which depends much on the size of the drug-release hole. Therefore, the size of the drug-release hole must be kept in a predetermined range rigorously. Recently, the drug-release holes are generally drilled by using the laser drilling technology and the sizes of the drug-release holes are commonly on the scale of millimeter or even micrometer. In practice, the laser power varies over time and the tablets undergoing drilling vibrates randomly. As a result, the size of the drug-release hole can be far from the standard. Obviously, measures should be taken to guarantee the quality of OPCR tablets. The sizes of the drug-release holes are usually measured manually in domestic production line. Due to the low efficiency and unreliability of the manual detection, it is preferable to employ the promising machine vision technology as an alternative of the manual detection. Motivated by this consideration, we develop a detection system based on machine vision to recognize the drug-release hole and measure the size.The main contributions of this thesis can be summarized as follows:1. Due to the size of the drug-release hole is so tiny and thereby difficult to detect artificially, this paper proposes a detection system to recognize the drug-release hole and measure the size of it automatically, where a machine vision system consisting of industrial camera and lens is deployed. The hardware design scheme and the workflow of the system are proposed in this thesis. The functions of all modules and the implementations are also studied and introduced. In the hardware aspect, we focus on the image acquisition module. For the purpose of best acquisition of pictures, extensive experiments and researches have been carried out to determine the most suitable resolution and focal length.2. Based on the hardware testbed, efforts have been made to investigate the software algorithms for automatic recognition of the OPCR drug-release hole. Firstly, image filtering algorithms have been studied and compared. The median filter has been finally selected to remove the noise with satisfactory results. Secondly, we analyze and compare the performance of different the edge detection algorithms via experiments. We find that, though the edge of the drug-release hole can be assessed by using the Canny operator, this edge is usually far from the actual edge of the drug-release hole. In order to improve the edge detection performance, we propose a new algorithm which combines the Canny operator and the dilation operation of the mathematical morphology. It is verified via experiments that the proposed algorithm can be used to obtain the edge of the OPCR tablet and the drug-release hole with satisfactory accuracy.3. Based on the edges obtained by using the proposed edge detection algorithm, a size measurement algorithm has been derived and analyzed via experiments. Our key idea is that the number of the pixels covered by the tablet and/or the drug-release hole is proportional to the actual area. Therefore, we measure and calculate the size of the drug-release hole by calculating the number of pixels of the tablet and that of the drug-release hole. Extensive experiments have been carried out with a sufficient number of tablets to verify the performance of the measurement algorithm and compare it with the size measured by microscope. The experimental results show that the proposed size measurement algorithm together with the proposed edge detection algorithm provides satisfactory precision.