Research On Character Recognition Of Chips Under Tape And On-Line Detection Of Tape And Reel Packaging Products

The existing chip character printing and finished product packaging processes are inevitably flawed,and the outflow of defective products will affect the production of downstream enterprises and cause losses to the enterprise itself.Therefore,it is necessary to conduct defect detection after chip packaging was completed.Tape and reel packaging achieves the best degree of automation and has become the preferred method of chip packaging.However,most of its packaging defect detection uses traditional manual visual detection,which is difficult to meet the increasingly production needs.According to the actual needs of enterprises and the characteristics of the target features to be tested,a visual system for character recognition of chips under tape and online detection of tape and reel packaging was designed based on template and feature fitting.This system can solve the problem of low accuracy in manual visual detection of tape and reel packaged products,and has significant application and economic value.The main research contents of this thesis are as follows:(1)Image acquisition of chips in tape and reel packaging: To reduce the impact of film reflection on the acquisition of chip character images under the tape,the reflective characteristics were analyzed and the effects of different light source were experimentally compared.Ultimately,a red low-angle annular light source with the best overall performance was chosen,and an image acquisition system was constructed according to the requirements.(2)High quality chip character template image training: A high-quality chip character template image training method was proposed to address the issues of nonstandard image templates and difficulty in processing the tested image,which could suppress random noise on the chip image,improve character quality and details,make the characters complete and full without adhesion and breakage.Moreover,the chip character template image was superior than the test image,which could be used for subsequent chip character recognition to improve recognition accuracy.(3)Character recognition of chip under tape: The recognition accuracy of universal character classifiers was low,and it was necessary to segment the collected images to train the character classifier.A geometric affine character segmentation method based on template character binary images was proposed to overcome the difficulty of character segmentation,such as i and j in the two-connected field of one character,and character adhesion and breakage caused by printing errors.Experiments on collected images had verified the accuracy and completeness of the method in segmenting individual characters.Using segmented character images to train multiple character classifiers.Finally,multiple algorithms were used to identify the error correction strategies that could mutually confirm the results,further ensuring the accuracy of the warehousing results.(4)Tape and reel packaging detection: Tape and reel packaging detection was divided into chip under tape detection and pressing tape detection,an overall detection process was designed based on the importance of defects.Extracting the feature edges of the tape and reel in the image,and using the features to detect the defects of the pressing tape.According to the different characteristics of overprinting and missing characters,different detection strategies were used,and a method of binary image morphological comparison of characters was proposed to detect the printing quality of characters,which met the special requirements of enterprises for printing characters.(5)Construction of visual application interface: Using Qt and HALCON for joint programming,a human interaction interface was designed,and secondary development of camera parameter control interface was finished to meet the requirements of the visual system.Experimental verification and preliminary on-site testing showed that the image acquisition system could effectively reduce the interference of packaging film reflection,and the image quality of the chip part under the tape of the tested image was relatively stable.The character recognition accuracy rate of under tape chip could reach around99.9%,and combined with multiple algorithm recognition results correction strategies to ensure the accuracy of characters registered and stored.The design of the detection process was reasonable,and the accuracy of defect detection was high,without considering the pause time of the mechanical system during image acquisition,each roll of tape took 5.1-7.8 minutes,which could meet the project needs.