Research On Multi-dimensional Data Fusion In Operator Training Evaluation

Using multi-dimensional data fusion technology for operator training evaluation,it covers many emerging fields such as Physiology,Neural Network and Ensemble Learning.In this paper,the operator's fatigue value is used as the direct reference for training evaluation,and the realtime and non-perceived fatigue monitoring and evaluation of the operator is taken as the direct requirement.The external features during fatigue is used as input information,and multidimensional data fusion technology is used to establish an unperceived monitoring model that evaluates operator fatigue in real time.The details are as follows:1)Determine the fusion objects required for data fusion.First,according to the external features of the operator in a fatigue state,the coordinate change amount of the sight line landing point,the head posture change amount,the blink frequency,and the mouth frequency are defined and quantified,and these four variables are used as input information of the fatigue monitoring model;Then,the fatigue value of the operator is defined and quantified as the output information of the fatigue monitoring model.2)A kinematics mapping model was established from the change in the line of sight coordinates to the change in head pose,which solved the practical problem of difficult to obtain the change in head pose.In the past,only intrusive devices could be used for the acquisition of head posture,which caused this type of fatigue monitoring model to be infeasible,and based on the proposed relationship model,it was possible to realize the non-perceptual acquisition of head posture information.The superiority of this relationship model is that it is suitable for scenarios where the center point of the operator's head is not fixed.3)A multiple linear regression model is established to solve the problem of data-layer fusion.Based on the three traditional mathematical methods of correlation analysis,analytic hierarchy process and entropy method,this paper proposes a new multiple linear regression model.4)A neural network model is established to solve the problem of feature-layer fusion.Based on the expansion of the original data set using Lagrange interpolation,a neural network model for feature layer fusion is established.Experiments show that this method has the best fusion effect.5)An Xgboost ensemble learning model is established to solve the problem of decisionlayer fusion.This paper uses Xgboost to perform ensemble learning processing on multiple decision trees based on the fusion of the expanded dataset using a decision tree algorithm.The fusion results of feature layers show that the fatigue monitoring model established in this paper has high accuracy.Compared with the previous methods of fatigue monitoring by invasive equipment,the model established can monitor and evaluate the fatigue of the operator without perception.After the model is established,it only needs to monitor the external features of fatigue to complete the assessment,which provides a technical guarantee for the real-time nature of information acquisition.In summary,the fatigue monitoring model established in this paper can take into account both real-time performance,non-contact and high accuracy.