Identification of traffic object's motion behavior based on track

To extract the track parameters of a traffic object in traffic video and identify its motion behavior,a new method is proposed based on CamShift(Continuously Adaptive Mean Shift)and HMM(Hidden Markov Model).First,an object entering the video scene is located and tracked by the CamShift based algorithm,then its track parameters are obtained.Next,the track parameters are processed to form the observation sequence of HMM,and the motion behavior modeling and probability evaluation are implemented based on HMM.At last,the behavior identification and behavior statistics of the tracked traffic object in video are achieved.Experiments show that this method can be used to sort and recognize the motion behavior of the traffic object by its corresponding behavior track,and to do some statistics or corresponding process schemes.

Abnormal driving behavior identification based on direction and position offsets

Abnormal driving behavior identification( ADBI) has become a research hotspot because of its significance in driver assistance systems. However,current methods still have some limitations in terms of accuracy and reliability under severe traffic scenes. This paper proposes a new ADBI method based on direction and position offsets,where a two-factor identification strategy is proposed to improve the accuracy and reliability of ADBI. Self-adaptive edge detection based on Sobel operator is used to extract edge information of lanes. In order to enhance the efficiency and reliability of lane detection,an improved lane detection algorithm is proposed,where a Hough transform based on local search scope is employed to quickly detect the lane,and a validation scheme based on priori information is proposed to further verify the detected lane. Experimental results under various complex road conditions demonstrate the validity of the proposed ADBI.

Transactional Network Analysis and Money Laundering Behavior Identification of Central Bank Digital Currency of China

With the gradual application of central bank digital currency(CBDC)in China,it brings new payment methods,but also potentially derives new money laundering paths.Two typical application scenarios of CBDC are considered,namely the anonymous transaction scenario and real-name transaction scenario.First,starting from the interaction network of transactional groups,the degree distribution,density,and modularity of normal and money laundering transactions in two transaction scenarios are compared and analyzed,so as to clarify the characteristics and paths of money laundering transactions.Then,according to the two typical application scenarios,different transaction datasets are selected,and different models are used to train the models on the recognition of money laundering behaviors in the two datasets.Among them,in the anonymous transaction scenario,the graph convolutional neural network is used to identify the spatial structure,the recurrent neural network is fused to obtain the dynamic pattern,and the model ChebNet-GRU is constructed.The constructed ChebNet-GRU model has the best effect in the recognition of money laundering behavior,with a precision of 94.3%,a recall of 59.5%,an F1 score of 72.9%,and a microaverage F1 score of 97.1%.While in the real-name transaction scenario,the traditional machine learning method is far better than the deep learning method,and the micro-average F1 score of the random forest and XGBoost models both reach 99.9%,which can effectively identify money laundering in currency transactions.

System Identification Modeling and Unstable Behavior of the Dynamics of Flows within the Tip Region of an Axial Compressor Blade Passage

In recent years, the correlation coefficient of pressure data from the same blade passage in an axial compressor unit has been used to characterize the state of flow in the blade passage. In addition, the correlation coefficient has been successfully used as an indicator for active control action using air injection. In this work, the correlation coefficient approach is extended to incorporate system identification algorithms in order to extract a mathematical model of the dynamics of the flows within a blade passage. The dynamics analyzed in this research focus on the flow streams and pressure along the rotor blades as well as on the unsteady tip leakage flow from the rotor tip gaps. The system identification results are used to construct a root locus plot for different flow coefficients, starting far away from stall to near stall conditions. As the compressor moves closer to stall, the poles of the identified models move towards the imaginary axis of the complex plane, indicating an impending instability. System frequency data is captured using the proposed correlation based system identification approach. Additionally, an oscillatory tip leakage flow is observed at a flow coefficient away from stall and how this oscillation changes as the compressor approaches stall is an interesting result of this research. Comparative research is analyzed to determine why the oscillatory flow behavior occurs at a specific sensor location within the tip region of the rotor blade.