| In automobile driving,driver fatigue is one of important causes of traffic accidents,and has become a major problem that can not be ignored in the current society.Therefore,the research on recognition of driving fatigue state is of great significance to the prevention of traffic accidents.In this paper,by studying the current status and development trend of fatigue state recognition at home and abroad,the non-contact fatigue state recognition method based on computer vision is studied deeply.Further the driver’s fatigue face is taken as the research object,then the key techniques of fatigue state recognition is studied;including fatigue face detection,location of significant deformation zones of fatigue,fatigue feature extraction and fatigue state classification.In this paper,the main work is illustrated as follows:1.For the complexity of the actual driving environment,firstly,the preprocessing of the collected images is carried out,including adaptive illumination compensation and median filtering to get rid of impatience.Secondly,considering the particularity of the driver’s driving position,the human face in the image is detected by the stable skin color information.On this basis,the deformable model is adopted to locate the deformed part of the face,and the posterior probability is added to avoid the dependence of the model on the initial position.Finally,the "three chambers and five eyes" is used to further locate the significant deformation area.2.In the fatigue state recognition system,whether the fatigue feature extraction is accurate or not directly affects the recognition result of fatigue state.Considering the driver fatigue characteristics mainly concentrated in several local deformation regions,under the pyramid structure,a new feature extraction method based on the shape feature PHOG operator and the texture feature SVTP is proposed.Experimental results confirmed that,the proposed method not only can extract more effective fatigue features than single features,but also can lay a good foundation for subsequent recognition process.3.Some problems such as high feature dimension,low recognition efficiency and large amount of computation are drawbacks in the existing driving fatigue state recognition algorithms.To tackle these problems,a novel fatigue state recognition method based on sparse deformation model is proposed in this paper.Firstly,the idea of sparse representation is introduced into the deformation model.Then the sparse deformation model of the fatigue face is established,in which the online learning algorithm is used to get over-complete basis function matrix of the training sample.,and sparseness of linear combination coefficients is used to achieve fatigue image classification and recognition.Last,take into account the fatigue state has a certain transfer law,the time window is added to optimize the recognition algorithm The experimental results show that the proposed method can improve the robustness and recognition rate of the system,can reduce the computational complexity of the system.The proposed method can obtain good recognition results in actual driving environment. |
