In a generic object tracking, depth (D) information provides informative cues for foreground-background separation and target bounding box regression. However, so far, few trackers have used depth information to play the important role aforementioned due to the lack of a suitable model. In this paper, a RGB-D tracker named TSDM is proposed, which is composed of a Mask-generator (M-g), SiamRPN++ and a Depth-refiner (D-r). The M-g generates the background masks, and updates them as the target 3D position changes. The D-r optimizes the target bounding box estimated by SiamRPN++, based on the spatial depth distribution difference between the target and the surrounding background. Extensive evaluation on the Princeton Tracking Benchmark and the Visual Object Tracking challenge shows that our tracker outperforms the state-of-the-art by a large margin while achieving 23 FPS. In addition, a light-weight variant can run at 31 FPS and thus it is practical for real world applications. Code and models of TSDM are available at https://github.com/lql-team/TSDM.

Continuous action recognition in video is more challenging compared with traditional isolated action recognition. In this paper, we proposed a hybrid framework combining Convolutional Neural Network (CNN) and Latent-Dynamic Conditional Random Field (LDCRF) to segment and recognize continuous actions simultaneously. Most existing action recognition works construct complex handcrafted features, which are highly problem dependent. We utilize CNN model, a type of deep models, to automatically learn high level action features directly from raw inputs. The LDCRF model is used to model the intrinsic and extrinsic dynamics of actions. The CNN is embedded in the bottom layer of LDCRF, which converts the structure of LDCRF from shallow to deep. This framework incorporates action feature learning and continuous action recognition procedures in a unified way. The training of our model is in end-to-end fashion. The parameters of CNN and LDCRF are jointly optimized by gradient descent algorithm. We test our method on two public dataset: KTH and HumanEva. Experiment shows our method achieves improved recognition accuracy compared with several other methods. We also demonstrate the superiority of features learnt by CNN compared with handcrafted features.

Anomaly detection and root cause analysis of energy consumption not only optimize energy use and improve equipment reliability but also contribute to green and low-carbon development. This paper proposes a comprehensive diagnostic framework for detecting anomalies, conducting causal analysis, and tracing root causes of energy consumption in medium and heavy plate manufacturing, integrating process mechanisms, expert knowledge, and industrial big data. First, a two-stage anomaly detection method based on box plot analysis is developed to identify energy consumption irregularities. Next, a weighted Granger causality analysis method based on LSTM is introduced, which effectively captures the nonlinear and temporal relationships of process variables, enabling the identification of abnormal causal pathways. Finally, a root cause tracing algorithm using an Adam-based variational inference Bayesian neural network is proposed to pinpoint the underlying factors responsible for the anomalies. Experimental results validate the effectiveness of the proposed methods.

Permanent magnet synchronous motor (PMSM) drive controllers are widely used in industrial systems due to their high reliability, and low operational and maintenance costs. However, PMSM produces unwanted harmonics while operating, which can cause issues including motor vibration, and the degradation of internal components. Excessive harmonics lead to issues directly affecting motor performance and efficiency, therefore, it is crucial to mitigate or eliminate unwanted harmonics. Herein, a model predictive current control (MPCC) method with an adaptive harmonic elimination strategy is utilized to reduce unwanted harmonics. By adjusting the weights of the adaptive filter, this method effectively removes unwanted frequency components from the main input signal, as determined by a cost function, resulting in an ideal reference voltage for the converter input. Experiments on a PMSM show that the proposed method effectively eliminates harmonics generated by the converter’s dead-time effect while maintaining system response speed, thereby enhancing the stability of the electric propulsion system.

To solve the problem of unsatisfactory control and poor real-time performance of nonlinear time-varying multi-input systems, this article proposes an intelligent model predictive control (MPC) algorithm inspired by heuristic dynamic programming (HDP), biological control theory, and operations research. Considering that the internal feedback information from a neural network (NN) is low, a multilevel feedback NN is proposed. Combining an NN with a biofeedback mechanism increases the internal feedback information and improves the convergence accuracy of the NN. The multilevel feedback network is used in three internal networks of the intelligent MPC algorithm. In order to improve the convergence speed of the proposed algorithm, a biologically inspired central coordination module and operations research theory inspired priority factor module is incorporated within the HDP algorithm. The prediction accuracy and control speed of the algorithm for nonlinear time-varying systems is greatly improved without affecting the control accuracy. The stability and convergence of the intelligent MPC algorithm is demonstrated on test data. Finally, the effectiveness and superiority of the proposed MPC algorithm is verified and compared against several traditional algorithms.