1.1 辅助缺陷识别（AssistDR）描述了一类计算机算法，用于帮助操作员确定无损检测数据。本指南使用术语AssistDR来描述这些计算机辅助评估算法和相关软件。就本指南而言，“缺陷”、“评估”、“评估”等词的使用绝不意味着算法正在处理或以其他方式进行独立的最终处理。根据应用情况，AssistDR计算机算法检测并选择性地对采集图像中的缺陷、缺陷、不连续或其他异常信号的指示进行分类。进行独立最终处置的软件被归类为自动缺陷识别（AutoDR）。 虽然本指南中讨论的概念与AutoDR应用程序相关，但在实施AutoDR时，可能需要额外的验证测试或控制。 1.2 本指南规定了使用AssistDR对非胶片射线照相试验数据进行部件射线照相检查的最低考虑因素。为了简单起见，本指南中的大多数示例和讨论都是围绕二维测试数据构建的。这些原理可以应用于三维（例如，体积计算机断层扫描）或更高维度的测试数据。 1.3 本指南中描述的方法和实践旨在应用AssistDR，其中图像分析将帮助操作员检测和评估适应症。 AssistDR融入测试和评估过程的程度将帮助用户确定所需的适当工艺鉴定和控制水平。本指南不适用于希望使用AutoDR的应用程序，其中没有对结果进行人工审查。 1.4 本指南适用于使用X射线源的射线照相检查。经AssistDR系统买方批准后，提出的一些概念可能适用于其他无损检测方法。 1.5 单位- 以国际单位制或英寸-磅单位表示的数值应单独视为标准值。每个AssistDR系统中规定的值可能不是精确等效值；因此，每个AssistDR系统应相互独立使用。 1.6 本标准并非旨在解决与其使用相关的所有安全问题（如有）。本标准的用户有责任在使用前制定适当的安全、健康和环境实践，并确定监管限制的适用性。 1.7 本国际标准是根据世界贸易组织技术性贸易壁垒（TBT）委员会发布的《关于制定国际标准、指南和建议的原则的决定》中确立的国际公认标准化原则制定的。 ====意义和用途====== 5.1 本指南描述了使用软件协助识别数字射线照相图像中显示的推荐程序。 提出的一些概念可能适用于其他无损检测方法。 5.2 当正确应用时，本指南中概述的方法和技术为射线检测从业人员提供了提高检验可靠性、缩短检验周期时间以及利用检验统计数据改进制造过程的潜力。 5.3 无损检测的典型目标是识别超过验收标准的缺陷。由于任何检查过程中都存在可变性和不确定性，因此确定了验收阈值，以便丢弃一些可接受的部件，以防止不连续性超过验收标准的零件投入使用。这种类型的错误称为误报，被认为不如允许不合格零件投入使用的误报错误严重。 成功应用AssistDR可以最大限度地降低假阳性率，同时将假阴性率降低到适合预期应用的水平。本指南中描述的方法和技术有助于实现这一预期结果。 5.4 随着深度学习、卷积神经网络和其他形式的人工智能的出现，AssistDR系统在获得生产使用资格后继续进化或学习的场景成为可能。本指南不涉及基于学习的AssistDR系统。本指南仅涉及具有软件代码和参数的确定性系统，这些软件代码和参数在鉴定后固定。请注意，此限制并不禁止使用本指南为使用深度学习技术的软件制定鉴定和使用策略。 深度学习系统的培训或学习过程需要在鉴定之前完成，并且深度学习系统的所有参数在鉴定之后和使用期间保持固定（与基于传统图像处理的确定性软件方法一样）。

1.1 Assisted defect recognition (AssistDR) describes a class of computer algorithms that assist a human operator in making a determination about nondestructive test data. This guide uses the term AssistDR to describe those computer assisted evaluation algorithms and associated software. For the purposes of this guide, the usage of the words “defect,” “evaluate,” “evaluation,” etc., in no way implies that the algorithms are dispositioning or otherwise making an unaided final disposition. Depending on the application, AssistDR computer algorithms detect and optionally classify indications of defects, flaws, discontinuities, or other anomalous signals in the acquired images. Software that does make an unaided final disposition is classified as automated defect recognition (AutoDR). While the concepts discussed in this guide are pertinent to AutoDR applications, additional validation tests or controls may be necessary when implementing AutoDR. 1.2 This guide establishes the minimum considerations for the radiographical examination of components using AssistDR for non-film radiographic test data. Most of the examples and discussion in this guide are built around two-dimensional test data for simplicity. The principles can be applied to three (volumetric computed tomography, for example) or higher dimensional test data. 1.3 The methods and practices described in this guide are intended for the application of AssistDR where image analysis will aid a human operator in the detection and evaluation of indications. The degree to which AssistDR is integrated into the testing and evaluation process will help the user determine the appropriate levels of process qualification and control required. This guide is not intended for applications wishing to employ AutoDR in which there is no human review of the results. 1.4 This guide applies to radiographic examination using an X-ray source. Some of the concepts presented may be appropriate for other nondestructive test methods when approved by the AssistDR system purchaser. 1.5 Units— The values stated in either SI units or inch-pound units are to be regarded separately as standard. The values stated in each AssistDR system may not be exact equivalents; therefore, each AssistDR system should be used independently of the other. 1.6 This standard does not purport to address all of the safety concerns, if any, associated with its use. It is the responsibility of the user of this standard to establish appropriate safety, health, and environmental practices and determine the applicability of regulatory limitations prior to use. 1.7 This international standard was developed in accordance with internationally recognized principles on standardization established in the Decision on Principles for the Development of International Standards, Guides and Recommendations issued by the World Trade Organization Technical Barriers to Trade (TBT) Committee. ====== Significance And Use ====== 5.1 This guide describes the recommended procedure for using software to assist with the identification of indications in digital radiographic images. Some of the concepts presented may be appropriate for other nondestructive test methods. 5.2 When properly applied, the methods and techniques outlined in this guide offer radiographic testing practitioners the potential to improve inspection reliability, reduce inspection cycle time, and harness inspection statistics for improving manufacturing processes. 5.3 The typical goal of a nondestructive test is to identify flaws that exceed the acceptance criteria. Due to the variability and uncertainty present in any inspection process, acceptance thresholds are established so that some acceptable components are discarded in an effort to prevent parts with discontinuities that exceed the acceptance criteria from entering service. This type of error, called a false positive, is considered less critical than a false negative error which would allow a nonconforming part into service. A successful application of AssistDR minimizes the false positive rate while reducing the false negative rate to levels appropriate for the intended application. The methods and techniques described in this guide facilitate achieving this desired outcome. 5.4 With the advent of deep learning, convolutional neural networks, and other forms of artificial intelligence, scenarios become possible where an AssistDR system continues to evolve or learn after qualification for production use. This guide does not address learning-based AssistDR systems. This guide addresses only deterministic systems that have software code and parameters that are fixed after qualification. Note that this limitation does not prohibit the use of this guide for developing a qualification and usage strategy for software using deep learning technology. The training or learning process for the deep learning system would need to be completed before qualification and all parameters of the deep learning system held fixed (as with deterministic software approaches based on traditional image processing) after qualification and during use.