Research On The Key Algorithms Of Visible Foreign Matter Detection In Chinese Herbal Oral Liquid

Abstract:Due to the shortcomings of traditional manufacturing processes and environment in the production of Chinese herbal oral liquid, different kinds of foreign substances, such as aluminum, glass and rubber scraps, may be found in the product. This can lead to the serious consequences when taken by the consumers. Currently, there are two different approaches to detect foreign particles in the oral liquid. One is based on the traditional artificial glare detection and the other is based on the machine intelligent detection. The traditional method has many drawbacks, such as high production costs, high-level labor intensity, and strong subjective detection. Moreover, the machine intelligent detection outweighs the traditional method by not only improving the speed and accuracy of the oral liquid foreign matter detection, but also reducing the costs of production. Thus, the machine intelligent detection method is the future of the oral liquid detection. For this reason, it is important and meaningful to do research on the oral liquid intelligent detection based on machine vision.This thesis mainly focuses on the algorithms of visible foreign matter detection in the Chinese herbal oral liquid images in order to achieve real-time online detection. The work presented focuses on image pre-processing, the detection and determination of visible foreign matter, and the light inspection system design and experiment. In the image pre-processing section, the secondary differential energy accumulated method has been used to restrict the disturbance from the static background image which is caused by scratches, dust and texture on the surface of the oral liquid bottle. The median filter method is adopted to remove the random noise in the image, and mathematical morphology operators have been selected to enhance the image contrast. In the detection and determination of visible foreign matter section, adaptive threshold segmentation technology has been applied to obtain the image segmentation threshold value which follows the changing of the environmental factors. Adaptive threshold visible foreign matter detection algorithm based on Otsu multi-frame sequence images is proposed to extract the foreign matter target. To identify the presence of impurities in the oral liquid, visible foreign determining algorithm is advanced to check whether the impurity exists or not. In the design of light inspection system and experiment, we build the Chinese herbal oral liquid light inspection system to check the accuracy, speed and stability of the algorithm proposed in the thesis. A large number of experiments and corresponding results verify the effectiveness and feasibility of the proposed methods are. It has figure18, table3, and reference67.