Research on video image preprocessing for monitoring abnormal behavior of mechanical operators

Abstract

Working environment of mechanical operators is complex, and the amount of process monitoring video information is large, which brings serious interference for human body target recognition of abnormal behavior. The algorithm is more complex, and the reaction time is longer. Through the analysis and comparison of image RGB model, YUV model and gray processing, when segmenting video sequence image, transforming RGB color image into gray image is conducive to image recognition. When the scene background is complex, the image can be analyzed in the normalized RGB color space, which can effectively eliminate the influence of shadow. The filtering denoising of Gaussian low-pass filter and median filter are analyzed, which have good denoising effect. It is very suitable for image preprocessing and denoising after recognition. Through the analysis of morphology operation such as corrosion and expansion, the cavity problem can be effectively improved. Before the target detection, preprocessing the target video sequence by computer graphics method can highlight the useful image characteristics and remove the useless image information, which can help to improve the detection effect.