Datasets play a crucial role in the training of deep learning models. For industrial datasets, the collection and annotation of images and videos is time-consuming, labor-intensive, and error-prone. In the past decade, With the development of rendering technology and hardware capability, more and more researches tend to use virtual datasets to overcome the shortcomings of real datasets. We studied the method of expanding the small sample data set of sprayed workpieces to solve the positioning problem of sprayed workpieces. We build the 3D model of sprayed workpieces and the factory environment in the virtual environment. We use blender software to render workpieces in different environments, and automatically generate the ground-truth label. In order to verify the effectiveness of this expansion method, We use real dataset, virtual dataset, and mixed dataset for model training. In our study, enhancements were made to the SiamFC++ model. Specifically, the backbone network was replaced with the ConvNeXt model, which boasts superior feature extraction capability. Additionally, we innovated the loss function by transitioning from IoU loss to CIoU loss, thereby introducing penalty terms for central point distance and shape consistency. Within the experimental section, we compared the performances of the SiamFC++ model using the AlexNet backbone network and the ConvNeXt backbone network. When trained solely on real datasets, the accuracy rates of the two model versions were 80.1% and 80.5% respectively. With virtual dataset training, the accuracy rates of the two versions improved by 6% and 7.4% respectively. When trained on mixed datasets, the accuracy rates of the two model versions saw respective enhancements of 8% and 8.6%. In all three training conditions, the ConvNeXt-based version of the model consistently outperformed the AlexNet-based version. Our improved model was further compared to mainstream object tracking models to validate its tracking efficacy. To substantiate the effectiveness of our model enhancements, we performed comprehensive ablation studies. [ABSTRACT FROM AUTHOR]