On-line coarse particle size analyzers can be used in crushing circuits to monitor ore size distribution of both feed and product streams. Fast and accurate information concerning these particle size distributions can improve the efficiency and lower crushing costs. There are several coarse particle on-line size analyzers commercially available, but most of them suffer from failing to provide volumetric information fast and accurately.
PC-Based Image-Analysis System
There are three major components that comprise the imaging analysis system. The first is a high- resolution video system to capture the image; the second is a personal computer fast enough to perform the modification/analysis of the image, and the final component is the necessary software modules. Consideration of the system hardware design and details of the software implementation are discussed as follows.
The hardware components of the on-line image-analysis system consist of a video camera, a video frame grabber and the host-computer system. The video camera, combined with the frame grabber, are often referred to as the image-acquisition device. The important function of this image-acquisition device is to put the image into the image memory. This involves the use of the video camera to digitize the continuous image into an array of pixels (picture elements) with different intensity values and to transfer to the image memory of the frame grabber. This image memory can be read or written to the host computer for subsequent image processing and analysis.
The software developed for this image-analysis system should have the operation functions such as image capture, image analysis, and finally the ability to transform the measured linear data from the 2-D image into volumetric data. In consideration of these requirements, three major classes of software modules were developed.
The major steps in image processing consist of: (1) Image Capture, (2) Segmentation, (3) Object Detection, (4) Measuring and (5) Analysis. Figure 3 shows these main operations of image processing. Routines for image capture have already been discussed in the previous section on interface modules. Usually, a multiple grey-level image will be obtained from the operation of the image capture step.
Experimental Testing
Since the statistically exact stereological solution from the measurement of regularly shaped particles such as spheres is known, filled circle and steel balls of different diameter were used to test the accuracy and resolution of the installed hardware and the implemented software.
The sieve-sizing technique is widely accepted as the standard method for coarse particle size analysis. However, for the image-analysis system that has been developed, the size of the ore sample is measured as the chord length distribution obtained from the projected particle image. It should be noted that the measured chord-length distribution presented in this paper is based on length. As mentioned previously, these data must be transformed in order to estimate the sieve size distribution from the measured chord-length distribution.
Results and Discussion
The shape of natural material such as rock or ore particles is irregular. In contrast to the regularly shaped particles, no exact stereological solution can be derived for the chord-length distribution. However, the distribution of the chord length of the rocks or ore particles can be measured experimentally. In practice, narrow sieve sizes of rocks or ore particles are taken and the chord-length distribution for each size class is measured from the projections of these particles using the image-analysis technique.
To characterize the chord-length distribution of the rocks or ore particles, onyx stone, copper ore and coal were used. Using out image-analysis system, the cumulative chord length distributions obtained for onyx stone, copper ore and coal. These curves indicate that the image-analysis system can distinguish between these narrow size fractions of ore particles. As mentioned previously, a transformation step is required to obtain the estimated sieve size distributions from these chord-length distributions.
A common problem associated with the inversion of the transformation equation is the difficulty in obtaining a good approximate solution for a steep sieve size distribution. Several narrow sieve sizes of the onyx stone were selected to evaluate our transformation solution.
On the other hand, the thresholding value is set too tight with Otsu’s technique. Based on these observations, it seems that the value set between these two techniques is just right for this specific example as shown in (d) of Figure 12. Table 2 shows the actual processing time required to finish all the image processing routines for these three techniques. All processing times are within the designed target of one minute.
A PC image-based on-line coarse particle size analyzer for a conveyor belt is currently under development. The system includes both hardware selection and software development. Regularly shaped and irregularly shaped particles are being used to evaluate the system. It has been found that the system being developed is accurate and fast enough for on-line particle size analysis.