Research Of Face Recognition Technology For Food Sensory Evaluation

In Chinese culture,food is the first necessity of the people,flavor is the soul of food,and people pay much attention to sensory quality of food.A survey indicates that the sensory quality of food will affect consumer's purchase intension.Therefore,food sensory evaluation possesses a crucial position in food industry.Aiming at the research of face recognition technology for food sensory evaluation,in virtue of the advantages of face recognition technology,such as intuitiveness,easy to distinguish,non-invasive,non-mandatory and non-contact,by observing facial expressions of consumers when they taste food,we can analyze what food and flavour they like or dislike.The research of this paper is fit for making an auxiliary and objective evaluation of food quality.Furthermore,the relevant information can be fed back to food merchants,to improve customer satisfaction level and also to provide decision support for enterprise producers and managers.This project involves fields spanning machine learning,image processing,psychology,food science and so on.The research method combines theoretical research and experimental testing.Traditional food sensory evaluation is carried out by evaluators.Due to long training cycle of evaluators,different hobbies and sensitivities of them.As a result,the evaluation result is unfair and is lack of generality and authenticity.While,food sensory evaluation based on face recognition technology is customer-centered,the evaluation result has a high stability and consistency.It overcomes the shortcomings of insufficient product information,short testing time and individual sensitivity difference,which lays a foundation for the development of food sensory evaluation technology in the future.As the key of the paper is to use facial expression recognition technology for making a sensory evaluation of food.Traditional facial expression feature extraction methods have certain limitations.In order to extract more effective information in facial expression image,this paper proposes a feature extraction algorithm based on feature operator and a deep learning method to extract facial expression features.The main research works of this paper are as follows:Firstly,this paper puts forward a facial expression extraction method using Gabor features and a novel Weber Local Descriptor.By combining the advantages of two feature extraction algorithms,the proposed method enhance the ability to extract facial expression images.The experimental results show that the proposed algorithm compared with the related algorithms has a better performance for facial expression recognition.Secondly,this paper proposes a deep small convolutional neural network for larger expression datasets,batch normalization and data enhancement strategies are used to alleviate over-fitting problems;for a small facial expression sample dataset with data limited,this paper proposes a transfer learning method for facial expression recognition.By learning the higher-order feature representation in image of Xception neural network model,image representation learned is applied in small facial expression dataset.The experimental results illustrate that the proposed deep small convolutional neural network and transfer learning method have achieved high recognition rates for facial expression.Thirdly,this paper develops a sensory evaluation system of food based on face recognition technology in real-time.Adopting webcams for real-time detecting facial expressions of customers when they are tasting food,the system can be applied for real-time sensory evaluation task of food.