Detecting System For Component Content Of Praseodymium/Neodymium In Extraction Process

China is one of the few countries which contain rich rare earth resources, besidesproduction scale and product quality of rare earth has been at the forefront of theworld. However, due to the extraction and separation of rare earth production processcontrol is still hovering in the stages of offline analysis, experience adjustments andmanual control, which leads the presence of rare earth enterprises to have many issues,such as huge consumption of resources, low productivity, unstable product quality.Meanwhile these issues has restricted the development of rare earth industry.Therefore achieving in-line monitoring for component content of rare earth elementsis a key factor to implement automatic control in extraction process. For theseproblems of in-line detecting, this paper proposed a Pr/Nd component content rapiddetection system based on the characteristics of the mixed solution and the practicalrequirements.(1) According to the actual measurement requirements, we design a detectionhardware system for Pr/Nd component content based on machine vision. First, wechoose the hardware device, based on the basic principles and selection methods ofcameras, lenses, frame grabbers, etc.. Secondly according to the actual effects ofdifferent lighting schemes, we determine the best lighting method. Then we researchthe images under three light sources (LED, D65, and CWF), and compare thecircumstances of data processing and modeling, light-emitting properties, power andlifetime considerations, LED ring light is chosen as a system source. Finally,depending on the selected hardware and lighting conditions, we set up the detectionhardware system for Pr/Nd component content based on machine vision.(2) Aiming at mixed solution for video features, an automatic segmentationalgorithm of strong adaptive is proposed in the HSI color space, based on Otsuadaptive threshold algorithm and median filtering algorithm. Moreover, through theextraction of the color characteristics and value, the system can get the colorcharacteristics of the component,which provides the conditions for the subsequenttarget pattern recognition(3) Based on LabVIEW2012and Matlab7.0hybrid programming softwareplatform, we design and develope the application software. First operation of displayinterface is designed using LabVIEW software system. Second realizing image processing function of complex computations call MATLAB. At last least squaressupport vector machine (SVM) algorithm is used to establish Pr/Nd colorcharacteristic value and components in the relational model, using the method ofmixed programming method based on MATLAB Script node, implementing thecomponent content recognition module.(4) The final analysis demonstrates that the accuracy, repeatability, andspeediness of the system can meet the requirements of the fast rapid detection site,which also laid a good foundation for applying this system to the actual Pr/Ndextraction production line real-time online detection system.