Recognition And Defect Detection Of Characters Print On Canned Food Bootle Cap Based On Machine Vision

With the development of economy and technology,people's living standards has been great improved,so the demands of food quality is increased.The production date is an important and intuitive criterion to measure the food safety or not.Therefore,to ensure that the production date and other related information is clearly and correctly marked is an important part of food production process.So using machine vision technology to detect the defection of production date has a very broad market demand.The paper research the defect detection method of character information print on the surface of the bottle cap.Considering the shape of the bottle,there have spatial distortion in the images which are taken by cameras.The distrtion will result in reading or defect inspecting difficulties.So,it's important to take a correction of the distortions before ocr and defect inspection.A correction method based on 3d calibration technology rebuild the 3d model of the bottle was proposed for this cylindrical object.By establishing the contact between 3d model and image pixels,the cylindrical region was unrolled into plane,meanwhile,the label and characters on the surface of the bottle was corrected into standard image which greatly reduced the difficulty of recognition and defect detection.The steps of character recognition and defect detection are:Firstly,the character image was segmented into a series of single character.Then extract the features of characters to train the multi layer perceptron recognition model.In the end,use standard character template to compare the quality of the character.In the end of the paper,a 360-degree panoramic image was stitched by four images which taken by four different cameras installed in four perspective.This greatly expanding the scope of application of the system include label defect detection and recognition of the barcode on the label and other tasks which is needed.