An air classifier is used in the recycling process of covered electric wire in the recycling factories, in which the covered electric wires are crushed, sieved, and classified by the air classifier, which generates wastes. In these factories, operators manually adjust the air flow rate while checking the wastes discharged from the separator outlet. However, the adjustments are basically done by trial and error, and it is difficult to do them appropriately. In this study, we tried to develop the image processing system that calculates the ratio of copper (Cu) product and polyvinyl chloride (PVC) in the wastes as a substitute for the operator’s eyes. Six colors of PVC (white, gray, green, blue, black, and red) were used in the present work. An image consists of foreground and background. An image’s regions of interest are objects (Cu particles) in its foreground. However, the particles having a color similar to the background color are buried in the background. Using the difference of two color backgrounds, we separated particles and background without dependent of background. The Otsu’ thresholding was employed to choose the threshold to maximize the degree of separation of the particles and background. The ratio of Cu to PVC pixels from mixed image was calculated by linear discriminant analysis. The error of PVC pixels resulted in zero, whereas the error of Cu pixels arose to 4.19%. Comparing the numbers of Cu and PVC pixels within the contour, the minority of the object were corrected to the majority of the object. The error of Cu pixels discriminated as PVC incorrectly became zero percent through this correction.
Over the past years, several studies have been made on recycling copper from wasted covered wire [
A schematic diagram of the experimental setup is illustrated in
The discrimination was carried out using linear discriminant analysis. In discrimination, the ratio of Cu and PVC in the wastes was calculated.
An image histogram plots the number of pixels in the image (vertical axis) with a particular value (horizontal axis). Using the histogram for a specific image, we can observe the entire tonal distribution. In image processing, it is important to select an adequate threshold of gray level for extracting the foreground from their background. Image histograms can be analyzed for peaks and valleys which can be used to determine a threshold value.
where
and red) were used in this study. It can be observed that the black PVC particle is buried in a black background. In this way, the particles having a color similar to the background color are buried in the background. Therefore, a new method to extract the particles from the background without dependent on the color of objects is proposed in this work. Hereinafter, the method is explained in the case of the black PVC particles. Usually, an image captured by a camera is repre- sented in the RGB color space. In image processing, RGB color space is converted to HSV color space because HSV color space is well suited to the human senses. HSV color space is composed of a hue (H), saturation (S) and value (V). H has a value from 0 to 360, S has a value from 0 to 255, and V has a value from 0 to 255.
It can be observed that there are two distribution peaks, the lower distribution (0 to 16) corresponds to the objects of the image, and the upper one (60 to 75) corresponds to the background. There is pronounced gap between the value distributions for the objects and the background, therefore, it is possible to extract objects from background. The Otsu’s method was adopted to find the threshold between two value distributions, and for the image presented in
The procedure for background removal is shown in
It can be observed that the foreground and the background are well separated into two value distributions. The value distribution of background was zero and the value distribution of the black PVC objects was distributed from 23 to 69.
Subsequently to the successful of background removal, the discrimination of Cu and PVC particles for the pre-processed images was executed. First, the pixels of Cu particles were set as group one and the pixels of six-color PVC particles were set as group two. Afterwards, the learning for LDA was executed. The learning derived constant terms for Equation (1) and determined the LDA function (Z). Discrimination using the obtained LDA function resulted in large errors, 83.85% for Cu particles and 2.23% for PVC particles. In order to reduce the error, instead of using one LDA learning, the pixels of single-color PVC particles
were set as group two and six LDA learnings were executed, respectively.
As an example of LDA learning,
where,
The color of PVC used on each LDA learning is described on the horizontal axis, and the learning error is presented on the vertical axis. The area enclosed with dashed lines corresponds to the results presented in
The discrimination between Cu and PVC pixels was carried out for the image
of mixed Cu and six-color PVC particles. The image of mixed Cu and six-color PVC particles is shown in
the initial stage was corrected as PVC as the stage progressed. For example, the black-white pixels enclosed with black dashed lines in
after discriminant analysis were then corrected using contours of objects to minimize the error. As shown in
In this study, we tried to develop the image processing system that calculates the ratio of Copper (Cu) and polyvinyl chloride (PVC) in the wastes as a substitute for the operator’s eyes. PVC consisted of six colors (white, gray, green, blue, black, and red) were used as particle model. By applying the difference of two color backgrounds, the particles and the background image could be separated without depending on the background. The Otsu’s thresholding was employed to select the threshold that maximizes the degree of particle-background separation. The ratio of Cu and PVC pixels was calculated from the mixed image by leaner discriminant analysis. The error of PVC pixels resulted in zero, whereas the error of Cu pixels increased to 4.19%. By using the contours of particles, the incorrect pixel discriminants were corrected. By comparing the numbers of Cu and PVC pixels within the particle contours, the PVC pixels, which was the minority of the object, were corrected into the Cu pixels, which was the majority of the object. The error of Cu pixels became zero through this object contours correction. Therefore, with this image processing system using linear discriminant analysis, it becomes possible to evaluate the discrimination result of air classification of copper and PVC which the operator has done so far, and it can be used for wind power control in air classification. Though discrimination between Cu and fixed six colors PVC was made in this experiment, when a new color sample appears, it can be dealt with by relearning with the sample. This image pro- cessing method can also be applied to the evaluation of discrimination results in samples of different colors, such as PET and labels, and is expected to be widely used in the recycling field.
This work was supported by a Grant-in-Aid for Science Research (JSPS KAKENHI Grant No. 16K14523) from the Japan Society for the Promotion of Science (JSPS).
Tayaoka, A., Ta- yaoka, E., Hirajima, T. and Sasaki, K. (2017) Image Processing System for Air Classification Using Linear Discriminant Analysis. Computational Water, Energy, and Environ- mental Engineering, 6, 192-204. https://doi.org/10.4236/cweee.2017.62014