Research On Detection Method And Its Application Based On Binocular And Heterogeneous Machine Vision

Copper has lots of excellent characteristics and has been widely used in manufacturer and daily life. Scrap copper plays an important role in copper resources. Scrap copper recycling reduces the consumption of resources and meets the concept of recycling. Sorting, cutting and smelting of scrap copper are the key steps in scrap copper recycling. Dection of physical properties of copper quality, cutting edge and temperature of copper refining furnace wall are needed when recycling scrap copper.These dections are usually realized by multiple dection systems and deeply relying on human experience, which leads to a low level of automatic dection.In order to solve the problem above, combining with the application background of sorting, cutting and smelting of scrap copper, we studied the binocular and heterogeneous machine vision and designed a corresponding machine vision detection system which can detect copper quality and cutting edge. Besides, we made full use of some existing software and used the existing heterogeneous data for temperature simulation of copper refining furnace wall. The main contents are as follows:1. After analyzing the basic principle of industrial CCD camera and hyperspectral imager, combined with the application background of this paper, we selected the hardware devices and built the hardware platform. Then we researched and developed a binocular and heterogeneous machine vision detection software.2. We used hyperspectral imager to detect the physical properties of the copper quality. According to copper’s light absorption characteristics at about800nm, we used hyperspectral imager to acquire the hyperspectral reflectance of copper samples. Also, we took advantage of CART to detect copper quality. Experiments showed that our method worked efficiently.3. We used industrial CCD camera to detect the cutting edge. After analyzing some typical filtering methods, we proposed an improved filtering method. An improved Sobel operator and sub-pixel edge detection method based on interpolation method were used to improve the accuracy of edge detection. Experiments showed that our detection method worked efficiently.4. In order to save the binocular and heterogeneous data, after learning the definition of BMP and BIL file formats, we proposed a custom file format which can save the data from CCD and hyperspectral in the same file. Combined with COMSOL software and the existing heterogeneous data, we made a simulation of temperature field of copper refining furnace wall.In the end, completed work is summarized and future research is proposed.