Research And Implementation Of IC Chip Surface Character Recognition Algorithm

The surface characters of IC chips are used to identify the identity of chip devices in the manufacturing process of integrated circuit chips.Generally,they mark the key information such as the name of the manufacturer,the chip model,and the chip serial number,which have important practical value either for the subsequent production or the actual use.At present,in the process of chip packaging and testing,manual verification is still the main method to verify the accuracy of characters on the surface of IC chips.Because of the complexity of the industrial production environment,the production process prone to leaking character printing,character printing error,character defects and various printing quality problem,cause artificial check has great uncertainty,it is difficult to ensure the efficiency and accuracy of calibration,and prone to accidents,cause unnecessary economic and reputation losses to the enterprise.Therefore,it is of great scientific significance and practical technical value to realize the automatic verification of IC chip surface characters in the process of packaging and testing.Based on the industrial production line of the surface of the IC chip image as the research object,surrounding the IC chip surface character information automation check actual demand,character on surface of image preprocessing,chip positioning,character tilt correction,character segmentation and normalization,and character recognition algorithm as the main research content,has carried on the analysis and research of the related technical method respectively,On this basis,a set of IC chip surface character recognition system is developed and constructed.Finally,through the relevant experimental analysis and verification,it is shown that the relevant technology and method explored are correct and effective,and the effectiveness and feasibility of the chip surface character recognition system constructed.The main work of this paper can be summarized as follows:(1)The characteristics of IC chip images are analyzed,and the process and methods of image preprocessing are analyzed and studied.A multi-process parameter matching image preprocessing algorithm is proposed.Different character region localization methods are provided for different detection stages.For the printing stage,the detection area is selected in the form of hand animation rectangular frame.For the packaging stage,an automatic chip character region localization method based on the maximum character connected domain is proposed.In the case that some chips have a small Angle tilt,an algorithm based on the height difference of character profile is proposed to determine the tilt line.In order to solve the problem of character segmentation,a character segmentation algorithm based on the connected domain segmentation algorithm and the projection method is proposed.(2)Conduct in-depth research on three commonly used character classification and recognition algorithms: template matching,support vector machine and convolutional neural network.After experimental comparison,the convolutional neural network is selected as the character classifier,and based on the classical Le Net-5 model is improved.A network model suitable for IC chip surface character classification and recognition is designed.(3)Based on the analysis of the layout characteristics of the characters on the surface of different IC chips,a character recognition algorithm based on convolutional neural network and geometric features is proposed,which can effectively recognize the chip surface image information with complex distribution such as horizontal and vertical printing characters.The original edit distance algorithm is extended to add several functional modes,such as similar characters passing through,redundant characters passing through,correct/wrong character position identification,etc.,and it is applied to the comparison between the character recognition results on IC chip surface and the real results.(4)Through investigating the development status of IC chip packaging and testing in enterprises,as well as the actual requirements of IC chip surface character recognition in each stage,a set of IC chip surface character recognition system is designed and realized based on C++ and Open CV.The required hardware modules are selected,and a complete hardware system is designed and developed,and a userfriendly graphical user interface is provided in the software.(5)The researched IC chip surface character recognition algorithm and its realization technology have been successfully applied in microelectronics packaging and testing enterprises.It not only improves the level of intelligent manufacturing of enterprises,promotes the development and progress of enterprises' technology,but also brings considerable economic benefits to enterprises.