As an emerging biomedical image processing technology, medical image segmentation has made great contributions to sustainable medical care. Now it has become an important research direction in the field of computer vision. With the rapid development of deep learning, medical image processing based on deep convolutional neural networks has become a research hotspot. This paper focuses on the research of medical image segmentation based on deep learning. First, the basic ideas and characteristics of medical image segmentation based on deep learning are introduced. By explaining its research status and summarizing the three main methods of medical image segmentation and their own limitations, the future development direction is expanded. Based on the discussion of different pathological tissues and organs, the specificity between them and their classic segmentation algorithms are summarized. Despite the great achievements of medical image segmentation in recent years, medical image segmentation based on deep learning has still encountered difficulties in research. For example, the segmentation accuracy is not high, the number of medical images in the data set is small and the resolution is low. The inaccurate segmentation results are unable to meet the actual clinical requirements. Aiming at the above problems, a comprehensive review of current medical image segmentation methods based on deep learning is provided to help researchers solve existing problems.

No abstract

No abstract

In this paper, the discontinuous projection-based adaptive robust control (ARC) approach is extended to a class of nonlinear systems subjected to parametric uncertainties as well as all three types of nonlinear uncertainties-uncertainties could be state-dependent, time-dependent, and/or dynamic. Departing from the existing robust adaptive control approach, the proposed approach differentiates between dynamic uncertainties with and without known structural information. Specifically, adaptive robust observers are constructed to eliminate the effect of dynamic uncertainties with known structural information for an improved steady-state output tracking performance-asymptotic output tracking is achieved when the system is subjected to parametric uncertainties and dynamic uncertainties with known structural information only. In addition, dynamic normalization signals are introduced to construct ARC laws to deal with other uncertainties including dynamic uncertainties without known structural information not only for global stability but also for a guaranteed robust performance in general.

No abstract

This paper proposes a novel robust adaptive algorithm for train tracking control with guaranteed prescribed transient and steady-state performance. As speed increases, the inherent time-varying uncertainties and unmodeled dynamics in the longitudinal dynamics of a high-speed train seriously impacts the tracking performance of automatic train operation. To improve train operation performance, an estimator based on immersion and invariance technology is developed to recover the unknown and time-varying plant parameters, and it renders the estimation error converging to a bounded residual set exponentially while providing more freedom for the controller. After certain error transformation, the prescribed tracking performance is introduced into the controller design. Then, an input-to-stable stable controller is developed through the backstepping technique, and it is proven that stabilization of the transformed system is sufficient to guarantee the prescribed performance. Rigorous theoretical analysis for the presented algorithm is provided, and a series of simulation studies also are given to verify the effectiveness of it.

A new framework to design parameter estimators for nonlinearly parameterized systems is proposed in this paper. The key step is the construction of a monotone function, which explicitly depends on some of the estimator tuning parameters. Monotonicity-or the related property of convexity-have already been explored by several authors with monotonicity (or convexity) being a priori assumptions that are, usually, valid only on some region of state space. In our approach monotonicity is enforced by the designer, effectively becoming a synthesis tool. In order to dispose of degrees of freedom to render the function monotone we depart from standard (gradient or least-squares) estimators and adopt instead the recently introduced immersion and invariance approach for adaptation.

No abstract