Design Of Table Automatic Ore Receiving System Based On Vision

Table concentrator is one of the main mineral processing equipment in the mineral industry,which is widely used to separate various rare minerals.However,the current table concentrator cannot achieve accurate mineral separation,real-time adjustment,considerable rate of mineral recovery,and it is also labor-consuming.Therefore,this paper studies the automatic ore receiving system based on vision,focusing on the difficult problems in the design of the automatic ore receiving system,and puts forward a feasible design scheme of the automatic ore receiving system.First of all,according to the demand analysist and the application scene of the automatic ore receiving system,this paper designs the overall architecture scheme of the automatic ore receiving system of table based on the vision,and divides the system architecture into the preparatory layer(training and evaluation detection algorithm),the control layer(interface and algorithm kernel)and the application layer(control element and actuator).In consideration of the development of various parts of the system,it defines the functions of each module to different levels.According to the architecture scheme of the automatic ore receiving system,the hardware selection and communication scheme design are carried out,in the same time,the test model is built.After all,this paper studies the requirements of the system for the detection algorithm,which aims at finding the separation point of ore belt.Finally,the error caused by the differential movement of the table is analyzed,and the location calibration scheme of the ore receiving plate is described,converting the coordinate of ore receiving plate to the image coordinate systemSecondly,this paper focuses on the difficult problems that observe the separation point of ore belt,and proposes the algorithm that detect separation point of ore belt based on convolution neural network.Because there is no open ore belt image dataset at present,this paper collects the ore belt images under various environments,and establishes the ore belt image dataset with label.This dataset is used for training detection algorithm in this paper,and also facilitates other researchers to carry out relevant research.In this paper,the detection network is built with Pytorch,Through the training and evaluation of dataset,the best network structure is obtained.Then improve the algorithm according to the characteristics of the ore belt image,and prove the superiority of the detection algorithm by comparing it with other detection algorithms.Thirdly,the automatic ore receiving system based on vision is realized.In this paper,the control layer and application layer in the system architecture are respectively implemented as: QT based software center platform and STM32 based motion control system,which communicate with each other through serial port.In the kernel of the software center platform,the detection algorithm is encapsulated as the object detection module,and a ladder control algorithm is proposed and implemented.Combined with the detection results of the object detection module,the algorithm calculates the control action needed to move the ore receiving plate to the separation point of the ore belt.By this way,this paper solves the problem of moving the ore board to the target position accurately.Finally,a test experiment is designed to verify the functional integrity of the motion control system and the software central platform.The performance experiment is carried out using the experimental model,and the experimental data are analyzed to verify the feasibility of the design scheme in this paper.this paper designs a scheme of automatic ore receiving system based on vision,focuses on the problem of which detecting separation point of ore belt and the accurately moving ore receiving plate to the target position.This paper also develop a software center platform based on QT,build a motion control system based on STM32,and finally realize the automatic ore receiving system,which can receive ore accutately.