386 APPENDIX 5.1 UX/UI design 5.1.2 User Authentication and Information Protection In this application, a user authentication function has been implemented in order to ensure data security and personal information protection. Upon entering their email address, the user is prompted to supply an authentication code, which is subsequently sent to their pre-authenticated email address. This process enables them to access the application. This authentication feature is pivotal in ensuring that the shooting data, training data, and discrimination models assigned to each user are managed individually, thereby limiting access to data from disparate sites. This design is a significant factor in preventing data leakage and ensuring user privacy. Furthermore, the administrator of the organization utilizing this application is capable of meticulously configuring the operating privileges for each individual user within the application itself. The subsequent Figure 5 illustrates the interface through which the administrator can modify the user's operating privileges. The administrator's authority encompasses the execution of all operations, including the uploading of training images, the generation of discrimination models, and the application of judgments to unknown images. Conversely, child users are granted specific authorizations, such as the ability to create training images. This privilege management system has been developed to minimize the risk of accidental deletion or overwriting of the discrimination model, thereby increasing the safety of important data. Figure 5 – Administrator interface for managing user privileges 5.1.3 Creation of dataset and generation of classification model The application provides an efficient and intuitive workflow for the entire process, from the creation of training images to the generation of discriminant models and their application to unknown images. The detailed procedure is outlined below. (1) Preparation of Training Dataset The initial stage of model building requires the creation of a training dataset. As illustrated in the Figure 6(a), the user uploads a .dat file of hyperspectral images from a PC to the application. Following this, the user is required to label the image. The application permits the user to establish up to three distinct labels; for instance, "Component A," "Component B," and "Other" are anticipated to be classified. The user is then able to select any area in the image and assign a corresponding label using a mouse drag operation, as illustrated in the Figure 6(b). This operation is intuitive and greatly improves the work efficiency of labeling compared to the conventional
RkJQdWJsaXNoZXIy MTM0Mzk2