Research About The Application Of Image Retrieval Technology In The Welding Workshop Production Safety

Content—based Image Retrieval(CBIR) technology is by extracting Image features to present the Content of the Image of the bottom of society, with its to identify and retrieve the corresponding Image, along with the computer technology, the development of different subjects, such as pattern recognition, Image Retrieval technology has made great achievements, applied to a variety of important areas, such as medical diagnosis, the public security system, and so on, the development prospects are very broad.Safety helmet is very extensive application in industrial production, in order to prevent accidents, ensure the safety of production, of the helmet to establish automatic detection and alarm system is becoming more and more urgent, at present the research of automatic identification of safety helmet or less. This article mainly aims at the recognition algorithm of the helmet, mainly divides into three parts, namely, image preprocessing, feature selection study, recognition model of choice. This article main research work is as follows:(1) image preprocessing part, the paper put forward based on the wavelet YCbCr homomorphic filtering illumination compensation, according to the characteristics of the YCbCr color space to separate brightness and Cb, Cr two component, the brightness for homomorphic filtering algorithm of wavelet, and this method is compared with other methods. Based on this, advances the face detection algorithm based on YCbCr space, eventually determine the image area of a helmet, the experimental simulation results show that the method is effective.(2) the part feature selection, this paper combined with the characteristics of the safety helmet, its shape characteristics are presented in the Hu moment and gradient direction histogram two characteristics, and the simulation statistics feature vector, to combine the two features together as characteristic vector of the image, used for training and identification of safety helmet.(3) identification model selection, this paper in-depth study of the more popular recently two identification model of neural network and support vector machine (SVM), for the selection of support vector machine parameters to choose the smallest optimization method, and finally compares the recognition rate of the two kinds of model, choose to use support vector machine (SVM) was finally determined. This article will Hu moment feature and HOG features the combination of vector as characteristic vector of the image, used for training and identification of support vector machine (SVM), be able to identify workers whether to wear a helmet when in and out of the door, the experimental results show that the method is effective and higher recognition rate, to realize intelligent monitoring system will provide strong support and practical guiding significance.