Research And Implementation Of Specific Object Recognition Algorithm In Complex Environment

With the development of technology,the heat of object recognition tends to grow.Object recognition,as one of the most cutting-edge research topic in the field of computer vision,is extensively applied in security monitoring,industrial production,traffic supervision,national defense industry,etc.Since the founding of deep learning in 2006,convolutional neural network(CNN),as one significant deep learning method,has obtained fruitful achievement in object recognition.Hence,the researchers have attached great importance to the studies of CNN in the development of object recognition.This paper mainly investigates the object recognition problem based on the combination of object recognition algorithm and real-time monitoring system.Since the recognition accuracy requirement of back counter monitoring is high for the sake of preventing bankers from violating regulations or misoperations,the proposed method is applied to the currency recognition in bank counter.The currency recognition of this paper is based on the combination of the open source computer vision code base Open CV and the CNN based on deep learning frame Tensor Flow.First,in lab environment,frames of currency video shot via cellphone is collected and pre-processed to serve as data set,and the routes,names and categories of the training set as well as validation set are obtained,and saved as text files.Thereafter,a deep CNN model is established and trained via the prepared training set,and the parameter of CNN is modified according to the training result.Thirdly,the trained CNN is combined with video to recognize the currency presented in the video,and the recognition probability is given.