| As a kind of gravity beneficiation,shaker beneficiation is widely used in the sorting of rare ores.As the initial link of the beneficiation process industry,the quality of the ore sorting directly determines the overall ore output.However,traditional shakers are often used in the beneficiation field,which requires manual adjustment of the deflector in real time.The ore grade is greatly affected by the technical proficiency of the workers,and the overall beneficiation accuracy cannot be improved.Aiming at the above problems,this paper studies the automatic ore receiving system of the shaking table based on vision,focuses on solving the problems existing in the identification process of the boundary line of the ore belt,and proposes a feasible design scheme of the shaking table deflector control system.(1)Due to the influence of ore quality,there is also a large difference in the distribution of the ore belt.The color gray value of the tail is similar to the gray value of the middle ore,and how to identify the overall ore belt demarcation position is the main difficulty of the current segmentation algorithm.A recognition algorithm of shaker ore belt boundary based on mixed image segmentation is proposed in this paper.First of all,the collected mineral belt image is preprocessed,and then the OTSU algorithm is used to obtain the left dividing line of the concentrate,and finally combined with the adaptive segmentation algorithm to calculate the threshold.The integral projection method is used to obtain the approximate position of the demarcation line of the middle ore belt,and the coarse positioning position is used as the growth point of the regional growth algorithm to obtain the location of the middle ore dividing line.(2)At the same time,it is found that the shaker ore belt in the process of beneficiation is affected by light adjustment in practical application because of the relatively complex conditions of the beneficiation site.An image enhancement algorithm under low light conditions is introduced to improve the recognition accuracy under low light conditions.A multi-scale Retinex algorithm based on HSL color space(MUTILScale Retinex,MSR)is used to compress the dynamic range of images while maintaining image color consistency.By transferring the RGB color space in the MSR algorithm to the HSL color space,and the new image is obtained by adaptively adjusting the brightness of the L channel and adjusting the S and H channels according to the changing proportions.Compared with the traditional MSR algorithm,the algorithm greatly solves the overall darkness of the mineral belt image due to insufficient lighting,and the phenomenon of overexposure of the mineral belt image and loss of shadow details.(3)In order to predict the movement of the dividing line of the mineral belt and determine the change rate of the mineral belt,the image speed characteristics of the rocked ore belt are extracted by multi-task unsupervised deep learning network algorithm.A simplified CNN network is designed for training point-of-interest detectors and descriptors in unsupervised models,respectively.In addition,a new loss function has been redesigned for the overfitting phenomenon showed in the Super Point algorithm.The dynamic characteristics of the mineral belt are obtained by combining the Kalman filter and the feature point matching algorithm of the RANSAC algorithm.(4)Due to the relatively harsh conditions of the mine and other influencing factors,it is particularly important to choose a reasonable control algorithm.For such problems,a position tracking control scheme of permanent magnet synchronous motor based on sliding mode control and adaptive reverse footwork is designed.The adaptive reverse step control of the position ring and the speed ring of the permanent magnet synchronous motor is proposed to attenuate the load disturbance,and the current controller is given.Considering the current control situation of the permanent magnet synchronous motor,the sliding mode controller using a novel sliding mode index approach law can effectively suppress the flutter.Finally,the stability of the entire system is verified.The experimental shows that the proposed method has a great performance in position tracking and load disturbance attenuation. |
