Study On Machine-vision Based Control System For A Cleaning Robot Of Air Duct Of Central Conditioner

With the development of science and technology, the central air-conditioner has been used in recent years for its advantages, such as, energy saving, higher seasonal efficiency rating and so on. Although, the central conditioner brings comfortable to people, due to lake of cleaning, and the air-duct becomes a hiding place for dust, viruses and bacteria which threaten seriously to human's health. In recent years, the horizontal air-duct cleaning robots had been produced at home and abroad and the technologies are relatively mature. Because of the structural features of vertical air-duct in central air-conditioner, research on the vertical air-duct cleaning robots are still in an initial stage.A cleaning robot used for vertical type air duct is proposed in the thesis as one example. The thesis focuses on the detection and control system based on machine vision, after briefly introducing the components and mechanism of the robot. Some topics are studied as follows:Firstly, the components of machine vision system of the cleaning robot are illustrated in the thesis. Based on the performance of machine vision system, the basic functions including image pre-processing and image matching are analyzed. In image pre-processing, according to the features of the images and the processing effect in experiments, the median filtering method is chosen to filter noise and the Laplacian operator to sharpen images.Secondly, in image matching, three matching methods are introduced to analyze the images. Image matching based on color features is used to analyze color histogram information. There are some disadvantages in the algorithm, e.g. the location information of colors are missed, the bumps or other attachments can not be identified and cost much more computing time. Image matching based on histogram feature only considers the gray information of images. It had a fast operation and still existed the deficiencies which cannot accurately identify the bump, depression, and other attachments of the air-duct. An improved image matching algorithm which can make up the deficiencies of the above two algorithms is presented in this thesis. However, the calculation speed should be improved.Finally, the control subsystem of the cleaning robot which mainly includes the machine vision inspection and control of mechanical moving structures is analyzed. A controller designed in the project mainly contains a host computer and a client computer. They are connected by the bus which can ensure the accuracy of transmitting control signals and the stability of operating the robot.The applications of machine vision system of the cleaning robot were discussed in this thesis. The algorithms are verified in MATLAB, and the programs are integrated in Visual C++ 6.0. It analyzed the calculation accuracy and the real-time of images captured by CCD, proposed and verified some practical methods which can apply some references in relevant field.