Design And Key Technology Research On Appearance Defect Detection System For Smart Meter

The smart meter as the national energy metering instrument, is veryimportant for its accurate and reliable information display, so the appearance defectsdetection for smart meter is necessary. But the traditional manual testing method withless-automated and person-influenced is more and more difficult to meet the need ofmodern production. Based on this, a smart meter appearance defects detection systembased on machine vision was proposed, and the related key technologies are studied,to overcome the disadvantages of traditional manual detection method, improve thedegree of automation, reduce the labor intensity, and achieve a higher detectionaccuracy and efficiency of continuous on-line detection. The main research works asfollows:Analysis and research the features of smart meter defects, detection requirementsand related background, discuss the emphasis and difficulty in establishing system,and make the overall plan of smart meter appearance vision detection based onmachine vision.Design the hardware and software of system according to the overall plan,choose the illuminant, camera, lens, image acquisition card and so on, and design thelighting mode and image acquisition system based on the features of defects and theimaging principle. And adopt the double control system of IPC combined with PLC.Develop the software system based on the LabVIEW development platform, whichrealizes the functions of image processing and system control.Image processing algorithm is the core of a vision defects detection system. Thisdissertation mainly studies the smart meter image processing algorithm including thelabels and barcode image preprocessing algorithm, the image positioning algorithmand the character recognition algorithm, the LCD and LED indicator light defectsdetection algorithm, the LCD screen parallel detection algorithm and relatedalgorithm, etc.Aim at the barcodes easily smeary interference and low recognition rate, anadvanced barcode recognition algorithm was put forward. The proposed recognitionalgorithm is high recognition rate and robust while divides recognition area averagely and adaptively. In view of the recognition rate of traditional character recognitionmethod on tightening even touching character is not high, an algorithm for touchingcharacters segmentation reconstruction and recognition of labels based on centroidalgorithm was proposed. The proposed algorithm gets the geometric center of eachcharacter through the centroid algorithm, and locates each character frame adaptivelythrough the found characters centroid, then segments the character by the foundcentroid and frame, and reconstruction and recognition character. The proposedalgorithm puts forward a new way to segment touching character, and is goodadaptability, strong robustness and high recognition rate. For the LCD backlightinterfering the detection, a LCD screen background eliminate algorithm was studied,the proposed algorithm makes the LCD screen background gray value sum subtract itsmode, and then rebuilds the background, which reduces the interference by unevenbacklight.After debugging, the device was installed on the production line, a large numberof actual detections were conducted, and the data received was used to themeasurement system analysis. From the analysis result, the defects detection systemproposed is good reproducibility and repeatability, is able to complete the smart meterappearance defects detection, can meet the needs of the defects detection, and has avery good practical value.