Data security technology is of great significance to the application of high resolution remote sensing image (HRRS) images. As an important data security technology, perceptual hash overcomes the shortcomings of cryptographic hashing that is not robust and can achieve integrity authentication of HRRS images based on perceptual content. However, the existing perceptual hash does not take into account whether the user focuses on certain types of information of the HRRS image. In this paper, we introduce the concept of subject-sensitive perceptual hash, which can be seen as a special case of conventional perceptual hash, for the integrity authentication of HRRS image. To achieve subject-sensitive perceptual hash, we propose a new deep convolutional neural network architecture, named MUM-Net, for extracting robust features of HRRS images. MUM-Net is the core of perceptual hash algorithm, and it uses focal loss as the loss function to overcome the imbalance between the positive and negative samples in the training samples. The robust features extracted by MUM-Net are further compressed and encoded to obtain the perceptual hash sequence of HRRS image. Experiments show that our algorithm has higher tamper sensitivity to subject-related malicious tampering, and the robustness is improved by about 10% compared to the existing U-net-based algorithm; compared to other deep learning-based algorithms, this algorithm achieves a better balance between robustness and tampering sensitivity, and has better overall performance.

Integrating machine learning (ML) with Statistical Process Control (SPC) is important for Industry 4.0 environments. Contemporary manufacturing data exhibit high-dimensionality, autocorrelation, non-stationarity, and class imbalance, which challenge classical SPC assumptions. This systematic review, conducted following the PRISMA 2020 guidelines, provides a problem-driven synthesis that links these data challenges to corresponding methodological families in ML-based SPC. Specifically, we review approaches for (1) high-dimensional and redundant data (dimensionality reduction and feature selection), (2) autocorrelated and dynamic processes (time-series and state-space models), and (3) data scarcity and imbalance (cost-sensitive learning, generative modeling, and transfer learning). Nonlinearity is treated as a cross-cutting property within each category. For each, we outline the mathematical rationale of representative algorithms and illustrate their use with industrial examples. We also summarize open issues in interpretability, thresholding, and real-time deployment. This review offers structured guidance for selecting ML techniques suited to complex manufacturing data and for designing reliable online monitoring pipelines.

High-quality processes, characterized by low defect rates, typically exhibit an exponential distribution for time-between-events (TBE) data. To effectively monitor TBE data, one-sided exponential Adaptive Exponentially Weighted Moving Average (AEWMA) schemes are introduced. However, the run length (RL) distribution varies with the magnitude of the process mean shift, rendering the median run length (MRL) a more pertinent performance metric. This paper investigates the RL properties of one-sided exponential AEWMA schemes using a Markov chain method. An optimal design procedure based on MRL is developed to enhance the one-sided exponential AEWMA scheme. Comparative analyses reveal that the one-sided exponential AEWMA scheme provides better balanced protection against both minor and major shifts in the process mean compared to EWMA-type and Shewhart schemes. Finally, two practical case studies illustrate the application of AEWMA schemes in manufacturing.

Gamma models are natural for positive, right-skewed quality characteristics and grouped time-between-event observations. Fixed Gamma cumulative sum (CUSUM) charts can detect persistent scale changes, but their deployed performance depends on the in-control (IC) scale estimated from a finite Phase-I sample. This study considers one-sided fixed Gamma CUSUM charts with known shape and a scale estimated once in Phase I. Conditional on this estimate, Phase-II observations follow a Gamma distribution with an effective scale ratio, allowing a one-dimensional Markov approximation to the complete run length (RL) distribution. Integration over the exact estimator law gives the marginal RL distribution while retaining the same estimate throughout each Phase-II path. Although every conditional average run length (ARL) is finite, the marginal ARL can be infinite; the marginal median run length (MRL) remains finite. Explicit integrability conditions characterize this distinction. The marginal MRL, the mean conditional MRL, and the distribution of conditional MRL are treated separately. Pathwise ordering yields the conditional-MRL distribution through scalar estimator crossings and reduces its probability constraint to one evaluation at a known Gamma quantile. Design selection profiles the reference value over a prespecified finite candidate grid, calibrates the corresponding control limits, and enforces finite marginal in-control ARL. The Python implementation reproduces published exponential and fixed-Gamma benchmarks. Across the eligible paired scenarios, conditional-MRL protection increased the probability of meeting the in-control conditional MRL target. The protected profiles had larger target-shift marginal out-of-control (OC) MRLs and larger marginal IC MRLs, so these comparisons do not hold in-control performance fixed. Independent continuous-state simulation assessed the finite-state probabilities at prespecified validation points; no exact-binomial comparison was rejected after Holm adjustment. A protected design whose integer MRL crossing remained sensitive to state resolution was excluded from the paired comparison. Supplementary sensitivity analysis shows reduced protection under downward shape misspecification, while matched-IC comparisons show shift-dependent detection tradeoffs.