基于遗传算法的最大熵双阈值分割的芯片缺陷图像分割
Chip Defect Image Segmentation Based on Maximum Entropy Double Threshold Segmentation Based on Genetic Algorithm

本文着重提出一种针对塑封芯片缺陷的基于遗传算法的最大熵双阈值分割算法，对芯片缺陷进行识别分割提取。塑封是指把支撑集成芯片的引线框架、集成芯片和键合引线等部分用塑封材料包裹密封起来，从而为集成芯片提供保护和隔热的功能。由于塑封为非气密性封装，故塑封工艺流程中容易产生各类缺陷，降低芯片的可靠性。对塑封完成的芯片进行质量检测可以避免缺陷芯片进入下一生产环节。传统的的图像缺陷阈值分割方法对其焊层图像缺陷分割的效果并不理想，本文结合缺陷超声图像本身的灰度高度图和自身特点进行分析后，提出基于遗传算法的最大熵双阈值分割法进行图像分割，最大熵结合了遗传算法的鲁棒性，并行性，效率高等优点，很好的对缺陷的细节进行分割，明显优于传统的算法。
This paper focuses on proposing a genetic algorithm-based maximum entropy double-threshold segmentation algorithm for plastic-encapsulated chip defects to identify, segment and extract chip defects. Plastic encapsulation refers to wrapping and sealing the lead frame supporting the integrated chip, the integrated chip, and the bonding wire with a plastic encapsulation material, so as to provide protection and heat insulation for the integrated chip. Since the plastic packaging is a non-hermetic package, various defects are prone to occur in the plastic packaging process, which reduces the reliability of the chip. Quality inspection of plastic-encapsulated chips can prevent defective chips from entering the next production process. The traditional image defect threshold segmentation method is not ideal for its welding layer image defect segmentation. After analyzing the gray height map of the defect ultrasonic image itself and its own characteristics, a maximum entropy double threshold segmentation method based on genetic algorithm is proposed. For image segmentation, the maximum entropy combines the robustness, parallelism, and high efficiency of the genetic algorithm, and it can segment the details of the defect very well, which is obviously better than the traditional algorithm.