Study Of Mine Coal Dust Online Measuring Technique Based On Image Processing

Mine coal dust may cause explosion in case of high thickness, and shorten the service lifetime of underground machinery and equipment. Therefore, it is great important to precisely measure and classify the coal dust density. Current online measurement for coal dust has larger measurement error and usual zero drifts, and is easy to be influenced by environments. Although image processing technologies are often applied into measurement successfully, rare research has been carried out for coal dust online detection and identification, in our country, based on image processing technologies. So, it is of great importance to study the image processing based coal dust detection methods and design proper online measurement equipments, in both theoretical and practical aspects.The key factor of real time measurement of the mine coal dust is to acquire the images properly and process them effectively. Especially, the image processing algorithms are rather difficult but meanwhile important for our study. In order to acquire satisfactory dust images, proper image collection equipments are designed herein. Besides, by examining the characters of the collected dust images, various image processing algorithms are proposed for image mage preprocessing, image segmentation and target recognition. The main results of the thesis are as follows:(1) In order to collect ideal coal dust images, the real environments of the underground coal mine and image processing characters are took into account. After studying other similar image processing instruments, a set of image acquirement equipment is designed and the collected images are analyzed comprehensively. This lays solid foundation for the following image processing technologies and algorithms. The acquired images demonstrate that the equipment has good underground operation performances.(2) The dust images are collected when the dusts are still moving, so the target images may come to be unclear. To solve this problem, a restoration algorithm based on Z transform is proposed. A new parameter estimation of point spread function algorithm is introduced, which adopts Radon transform to estimate the fuzzy angles, and improved autocorrelation function method to estimate the fuzzy length. After that, Z transform is used to recover the dust image rapidly. Comparison results show that the proposed method can recover the moving dust images rapidly and accurately.(3) Considering the low SNR of dust images, improved minimum mean square-error estimation (IMMSE) denoising algorithm is proposed. Experiments show that the coal dust images are polluted by Gauss noise, impulse noise and other kinds of noises. The Minimum mean square-error estimation (MMSE) is effective to Gaussian noise, but less useful to pulse noises. However, adaptive median filter has good effect on salt and pepper noise denoising. Therefore, by combining IMMSE with adaptive median filter (IMMSE-AM) comprehensively, a novel denoising algorithm is adopted for coal dust images. Experiment results show that IMMSE-AM has better denoising performances and less computation.(4) Considering the characters of dust particles and their backgrounds, a segmentation algorithm based on improved particle swarm optimization and fuzzy entropy is proposed. Traditional particle swarm optimization (PSO) can hardly get good optimization performance because it easy to get stuck into local optimum. Herein, an improved PSO which combines proposed inertia adaptive PSO with partial particles Morlet mutation is proposed. By combining this algorithm with fuzzy entropy, the image segmentation problem is settled and the fuzzy parameters of maximum fuzzy entropy are explored. Experiment results show that the new algorithm has the capability of good segmentation performance and low time cost, which can be use for real time and precise measurement of coal dust images.(5) A new coal dust particle recognition algorithm based on concave points and ellipse fitting is proposed for irregular and overlapping particles. For overlapping particles, concave points are usually explored for separation. In this thesis, based on the area method, a simple but effective concave point extraction algorithm is proposed. Ellipse fitting recognition is often used for particles with multiplicity shapes. Direct least square method is one of the best methods, but it is instability for noise points. Therefore, we propose an improved ellipse fitting algorithm based on the direct least square method. The algorithm extracts six random dominant points for ellipse fitting, based on the direct least square method. Then, it measures the distance from the circle points to the fitting ellipse, and pick up those whose distance is smaller than the threshold. The number of these points is used as the PSO fit value. We use proposed differential mutation bare bones particle swarm optimization to hunt for the global optimization fitting parameters, and finally adopt the optimal one for ellipse fitting.(6) The overall scheme for the hardware platform of underground dust online measurement system is designed. The software flow chart and key solutions are presented. Experiments are carried out to validate the accuracy, measurement speed and stability of the proposed image processing algorithms. Results demonstrate that the designed coal dust measurement system meet the desired targets, and can be applied for online detection of underground coal dust.Finally, the main work of this thesis is summarized, and the future research directions are proposed.