The Research On Visual Inspection For Visible Contaminating Particles In Solution Of Irregular Bottles

Along with the product quality issues drawing more and more attention, the artificial inspection for beverage and liquid pharmaceutical products can no longer satisfy the requirements of manufactures, consumers and more restrictive quality standards. Based on the research of automated inspectors for liquid products, an automated visual inspection system for visible contaminating particles in solution of irregular bottles is presented in this thesis. This system can inspect products more rapidly and accurately than inspecting workers. It has great economic and social benefits.Firstly, machine vision technology in industry inspection is represented, and its classification, construction, related products, applications and development trends are introduced. Then, apparatus and methods for on-line automated inspection for liquid products all over the world are introduced and analyzed.Secondly, some practical advice for designing industrial visual inspectors is presented. Then, the design scheme of the visual inspection system for solution in irregular bottles is proposed. The main innovations include following aspects: servo control manipulators, which can invert the irregular bottles to make a motion distinction between the contaminating particles and the static background interference, is designed to solve the problem caused by irregular bottle shape. Besides, special illumination schemes are proposed for two inspecting stations. Then, a structure for on-line inspection, which is consisted by servo control manipulators and a turret, is presented. At last, the principle and process of the visual system, the device configuration and the electrical control conformation are introduced.Thirdly, the algorithm design difficulties are analyzed. A method utilizing the continuity and direction features of the contaminating particles trajectory in image sequences is proposed. Then, an important motion law of the particles is obtained after analyzing the math model and data records of the particles trajectory in image sequences. Furthermore, the image pre-processing steps are presented. After the effective detection region calibration, an adaptive filter based of extreme and mean is used to remove the image noise.Finally, the algorithm for inspecting contaminating particles in solution is presented, which consists of three steps: pre-processing, motion detection and true particle recognition. Different methods are given in the thesis for different type of contaminating particles. In the step of motion detection, frame-difference is used to detect moving white particles; while an improved frame-difference method based on assimilation filling is used to detect moving black particles. In the step of true particle recognition, a method based on multi-frame combination technology is proposed to recognize the true black particles; while Kalman filter is used to predict the trajectory of the white particles such as glass scraps in solution. In the prediction process, the direction of the particle trajectory is judged to identify the true white particles. At last, an experiment result is given in the thesis, which shows that the algorithm proposed is effective and feasible.