Research On Hidden Camera Detection And Recognization Method Based On Machine Vision

With the increasing popularity of optoelectronic imaging technology,the camera has also developed towards miniaturization and intelligence,bringing great convenience to people's lives.However,there are also criminals who use cameras to steal state secrets and citizens' privacy in order to make huge profits.These improper uses of cameras have caused great losses to the country and the people.To this end,this paper proposes a new camera detection method based on the composition of the camera and imaging mechanism by integrating the infrared information and visible light information of the target,combining with the knowledge of image processing and pattern recognition,machine learning and deep learning to realize detection and recognition of the target camera.Further more,a fully automatic camera detection and recognition system is developed meanwhile.The main research contents are as follows:In the research of camera detection algorithm,a camera detection method based on the optical filter characteristics of Infrared-Cut Filter is proposed.The paper studied the reflection cut-off characteristics of Infrared-Cut Filter for infrared light and use the image difference-based suspicious target extraction algorithm to obtain the suspicious target set and design the camera feature descriptor to analysis the characteristics of elements in suspicious target set,then judge suspicious targets in sets.Finally,the DBSCAN-based detection clustering algorithm is used to solve the target rechecking problem to ensure the uniqueness of the same target in the detection results.Experiments show that the camera detection algorithm based on the optical filter characteristics of the infrared cut filter can better balance the recall rate and accuracy and has good running performance,which can better meet the application requirements in the task of camera detection with priority recall rate.In the research of camera recognition algorithm,a camera recognition algorithm based on model fusion is proposed.In this paper,the camera recognition method based on VGGNet16 deep learning model and random forest model is studied respectively.RGB color scene images are introduced on the basis of infrared images.The relevant data sets are respectively trained to train VGGNet16 and random forest models,and the respective disadvantages of the two models are analyzed according to the experimental results,then the two judgment results are integrated from the perspective of information complementation to achieve model fusion and improve the performance of the algorithm.The experiment proves that the camera recognition algorithm can suppress the false alarms which are in the result set of detection algorithm,and effectively improve the detection accuracy of the algorithm while ensuring the recall rate of the algorithm.In the research of camera detection system,a fully automatic camera detection and recognition system is developed.The detection and recognition system completes the automatic detection and recognition of the target camera in the scene by collecting the infrared image and the RGB color scene image.The system has an automatic detection mode and a manual detection mode,and can meet different detection requirements through the switching mode.