Research On Visual Inspection And Analysis Method Of Nickel Electrolysis State

The nickel smelter’s nickel electrolysis workshop uses the nickel sulfide soluble anode diaphragm electrolysis process to produce refined electrolytic nickel.During the electrolysis process,workers need to regularly lift the electrolytic nickel cathode plate to check the surface defects of the electrolytic nickel cathode plate,in the meantime,check the water flow condition of the fluid infusion port.The test takes time and effort,and the test standard is unstable.It is the main link that restricts the quality control and efficiency improvement of the nickel electrolysis production process.Usually,fully automated testing is the main method to solve the above problems.This article is aimed at the software detection part of the automatic inspection robot in the nickel electrolysis workshop.The main content is to study the common computer vision algorithms to achieve the detection function of the above problems.This article is an important part of the automatic inspection robot and the key to solving specific inspection tasks.This paper first solves the surface defect detection of the cathode plate.Its image processing includes a dynamic threshold binarization method to segment the cathode plate image area.The boundary tracking method tracks the segmented cathode plate area to obtain its contour information.The outline of the cathode plate is approximated to a quadrangle by polygonal approximation,and the shape is corrected by perspective transformation.Finally cropped to the whole cathode plate picture.Image detection includes extracting grayscale,texture and contour features of the image and using principal component analysis(PCA)to reduce the dimension to obtain a better feature vector than before.Support vector machine is used(SVM)classify and judge the characteristics,and detect the whole cathode plate image by sliding window detection,the accuracy rate can reach 82.3% under the existing samples.Then solve the electrolyte replenishment status detection,image processing includes suppressing background by video frame-to-frame differences on the electrolyte replenishment area,and median filtering to obtain the contour information of the water ripple area.Image target detection includes the detection of water ripples and water flow through the cascade classifier to obtain the detection results.Extract water ripple contour features and use support vector machine(SVM)to further classify the detection results and the results of the cascade classifier to obtain the final judgment results.In the case of existing samples,the detection accuracy can meet the system requirements.After summing up,under the existing sample conditions,the detection method proposed in this paper has certain detection capabilities for the above problems,and has good engineering application value and academic reference value.In the future,when the conditions are mature in all aspects,it can get further field practices and adjustments in the field,even methodological improvements.