This paper proposed a detection method with improved ViBe algorithm is proposed for the workpiece object aiming at the problem of low efficiency of manual inspection of workpieces in industrial manufacturing. Firstly, the bilateral filter method is used to remove the highlight area of the workpiece, and then a set of pixel values for storing the current position or adjacent positions is established for each pixel. Finally, the current pixel is compared with the pixel value in the set to determine whether the pixel belongs to the background, and the model is updated by randomly selecting the pixel value in the background model. This method requires only a single video frame to complete the initialization, greatly improving the detection efficiency. Compared with the traditional ViBe algorithm, it can effectively remove the influence of workpiece detection for highlights. Meanwhile, the detection accuracy and visual effect are superior to other three comparison algorithms,which can lay a good foundation for the identification and tracking of workpiece in industrial production.