Application Research Of Image Processing Method In Photoshop Marking

The traditional manual marking method is inefficient and teachers have heavy burdens,and the reading results are greatly influenced by subjective factors of teachers.In the photoshop exam,the disadvantages of the traditional marking methods become more prominent.The current photoshop auto-scoring technique mainly uses the psd document information acquisition method for processing.This method can obtain key points such as image size and layer information,but it is difficult to obtain the specific content of the image,thus making a judgment on the incapable image.In order to solve the above problems,this paper proposes an image-based photoshop marking method.This method uses the improved convolutional neural network to achieve the review of the candidate’s operation result images.It has a wide applicability and a high evaluation accuracy.The main work of this paper is as follows:(1)Photoshop marking program design.Due to various types of operation problems,the operation result image involves a variety of feature changes,and a photoshop checking program based on convolutional neural network is designed.According to the characteristics of marking,the candidate’s answer sample and the convolutional neural network scoring model were designed.The sample training network was used to implement the review of the candidate’s answer image.(2)The design of the answer sample of the operation question.Because it is unrealistic to collect candidate answer images,and the existing data set is not suitable for training this paper,a single-material and dual-material answer question sample of two types of operating questions is designed.Single-material operation,using contour scaling,selection and affine transformations to generate samples.Dual material operation problems,generate samples by processing the background and the selected area of the superimposed image.The samples generated through these methods cover the possible situations of the candidates and are in accordance with the actual results of the answers.This is conducive to the training and practical application of the photoshop marking model.(3)Convolutional neural network model design for photoshop marking method.The VGG-16 deep convolutional neural network is used to solve the problem of small differences in candidates’ answer images.The model is complex,the training time is long,the speed is slow,and the network structure is optimized according to the marking characteristics.The network parameters are adjusted,and the randomness is adopted.Inactivation technology and Adam optimization algorithm training network.The results show that the network can improve training accuracy while reducing training time.Based on image processing,photoshop automatic marking technology solves the problem that the analysis of PSD document information method can not judge the image content.It has great significance for solving the problem of automatic marking technology.