【作者】 刘英剑；

【导师】 张起贵；

【作者基本信息】 太原理工大学 ， 电子与通信工程， 2018， 硕士

【摘要】 随着计算机教育的普及,职业教育教学内容更新与课程体系改革取得了一定成果。其中Photoshop是职业教育中的一门主干课程,对此软件的熟练掌握已经成为学生的一项重要的技能。在学校教学过程中,存在最大的问题就是,该课程考试严重依赖人工评阅。存在评测效率低、主观评卷不合理等弊端。目前对Photoshop自动评测系统的研究较少,实现的判分点相对简单,在题目难度设定上无法达到职业教育要求。所以本文对此课题展开研究,研发Photoshop自动评测系统,目的在于提高Photoshop考试的评测速度和判分准确度。本文具体的研究工作包括以下几个方面:(1)研究了目前现有的Photoshop评测系统的弊端,根据现有系统存在的问题,本文提出一种新型评测算法。在评阅过程中从文件结果、操作过程和图像结果三个角度进行试题分析。从多个角度对试卷进行评阅,实现了图像处理方法和评测系统的结合统一,提高了评卷的准确率。(2)研究了Photoshop文档信息存储结构以及操作步骤信息获取。设计了解析PSD文档的方法,并且在此基础之上实现了解析PSD文件类库,为自动评测系统提供了底层判分依据。(3)针对图像结果,对不同类型的图像结果设计了不同的图像处理算法,来提高评测系统的严谨性。尤其是针对答案结果不唯一的复杂操作题,将深度学习技术和评测系统相结合。文中提出了适合本项目的深度学习网络架构,构建二分类数学模型。然后利用训练好的卷积神经网络对答题结果进行图像特征提取,最后根据提取出的特征对答题结果进行评阅,解决了目前主观评阅不唯一的缺陷,提高了Photoshop评阅的公平性。(4)设计整体评测流程,实现了能够批量化运行的Photoshop自动评测系统系统,采用多线程技术来加快评阅速度。最后通过实例分析说明Photoshop自动评测系统的评测效果。实践表明,基于本文中提出的新型判分策略,实现Photoshop自动评测系统是完全可行的。通过解析PSD文档和分析图像结果相结合的评测方式,提高了判分算法的准确性和市场应用性,对提高Photoshop教学质量也有一定的推动作用。更多还原

【Abstract】 With the popularization of computer education,some achievements have been made in the updating of content and reform of curriculum in vocational education.Among them,Photoshop is a main course in vocational education.The mastery of this software has become an important skill for students.In the school teaching process,the biggest problem is that the course examination and scoring relying on manual marking which has the disadvantages of low efficiency and subjective evaluation.So far,there are few researches on the Photoshop automatic evaluation system.The judgment point of the current system is relatively simple,and it can not reach the vocational education standard in terms of the difficulty setting of the subject.Therefore,this paper focuses on the automatic evaluation of Photoshop,and develops a Photoshop automatic evaluation system.The purpose is to improve the speed and accuracy of the Photoshop course examination.The specific research work of this paper includes the following aspects.(1)The disadvantages of current Photoshop scoring systems were analyzed.According to the existing problems of the existing system,this paper proposed a new marking algorithm.In the judging grades process,the document results,operation process and image results are analyzed from three angles.The examination papers were reviewed from various angles,and the combination of the image processing method and the marking system was realized,so the accuracy of marking was improved.(2)The Photoshop document information storage structure and operation procedure information acquisition were studied.The analysis method of the PSD document and its corresponding implementation scheme were specifically studied.Based on it,the analysis of PSD file class library was achieved,providing the underlying judgment criteria for the automatic marking system.(3)Different image processing algorithms were designed for different types of image results,to improve the accuracy of the scoring.In particular,for complex operation questions where the results are not unique,combine deep learning technology with marking system.I combines deep learning technology and marking system.The paper proposes a deep learning network architecture suitable for this project,constructs a binary classification mathematical model.and then uses the trained convolutional neural network to extract image features of the answer result.Finally,the answer is reviewed according to the extracted features.It has solved the defects of the present subjective evaluation and improved the fairness of the marking.(4)The overall marking process is designed,an overall batch Photoshop evaluation system was achieved,and the multi thread technology is used to speed up the marking.The operation flow and automatic marking effect of Photoshop automatic evaluation system were illustrated by checking examples.It is demonstrated that the achievement of the Photoshop automatic evaluation system based on the new marking strategy proposed in this paper is completely feasible.Through the marking method of the combination of analyzing PSD documents and processing image results,the accuracy of the verdict algorithm results and its market application are improved.The study and the achievement of the Photoshop automatic evaluation system plays an important role in promoting the quality of Photoshop teaching.更多还原