Moving object detection of assembly components based on improved background subtraction algorithm

Abstract

In the automatic assembly of mechanical components, it is difficult to guarantee the quality of assembly image acquisition caused by the change of assembly environment. Firstly, the color space is compared and analysed, and histogram equalization is used to preprocess the image to reduce the impact of image brightness on subsequent processing. Then, based on the texture features of the image, an image reconstruction idea from coarse to fine is created, and an image representation method from macro to micro is established to suppress the pseudo texture and noise texture in image processing. On this basis, an adaptive threshold algorithm based on the target distance is proposed to improve the detection accuracy of the assembly target. Experimental results show that the proposed algorithm can significantly improve the detection accuracy of moving objects in assembly video frames.