| With the advancement of society,security issues have received more and more attention.Content-based image retrieval technology is one of the important applications of video surveillance systems.Image retrieval technology has made great progress with the development of multi-disciplines such as computer technology and pattern recognition,and has been applied to many important fields.Among them,public security,medical care and other fields are involved,and the development prospects are very broad.Especially in the petrochemical industry,the demand for video surveillance systems is growing rapidly.The urgent need for intelligent video surveillance technology in various fields has made the related topics of image retrieval technology a research hotspot of intelligent monitoring systems in recent years.Safety helmet can protect the head of the operators when they are injured by falling objects and other specific factors.Wearing the helmet correctly is a safety standard that has been emphasized in the chemical production.In order to ensure production safety arid reduce operation risk,the demand for automatic detection of safety helmet and alarm system of wearing status recognition in chemical industry places is becoming more and more urgent.For example,there are external construction and maintenance personnel in our subordinate chemical plant all the year round,and the safety awareness of individual personnel is weak,and personal injury accidents are easy to occur in the process of operation.As the supervisor,the intelligent helmet detection system can not only improve the work efficiency,but also achieve the actual effect of improving the level of enterprise safety management.At present,there is not much research on automatic identification technology for wearing safety helmet status.This paper mainly studies the wearing state recognition algorithm of safety helmets,mainly from image preprocessing,pattern recognition theory and feature expression.The main work is to use video object segmentation method based on interframe difference method to judge video images.The moving target uses the Histogram of Oriented Gradients(HOG)feature of the image and the Support Vector Machine(S VM)classifier to determine whether the moving target in the image is a pedestrian,and then uses an algorithm based on a Deformable Parts Model(DPM)to detect the wearer wearing state of the worker.The algorithm models the safety helmet wearing state from multi-scale and multi-view.Then,the Latent SVM is used to train the helmet to correctly wear the state recognition model.Finally,the target matching is performed to determine whether the field operator complies with the regulations and wears the helmet.On this basis,the safety protection needs of key areas such as large equipment and power distribution rooms in the chemical plant,as well as the purpose of retaining operation records.This paper also studies the anti-foreign object intrusion detection based on background subtraction method and Gaussian Mixed Model(GMM)background modeling,and recognizes pedestrians and mobile machinery entering the defined range.In order to prevent foreign body intrusion detection,this paper uses a Mixture of Gaussian model to update the background in real time,and extracts the target of invading foreign objects by threshold segmentation.The experimental results show that the method adopted in this paper can meet the requirements of safety helmet wearing status recognition and monitoring of workers in chemical production plant.The foreign body intrusion detection results are reliable and achieve the expected accuracy and reliability.The construction and test of intelligent video surveillance system in chemical plant prove that the system runs stably and achieves the expected goal,which lays a foundation for the application of intelligent video surveillance system in chemical enterprises.It has practical guiding significance for the intellectualization of chemical safety production. |
PDF Full Text Request