In the transformation of industrial automation to smart manufacturing, visual inspection systems as critical sensing technologies are hindered by their high costs and algorithmic complexity, impeding the intelligent upgrading of small and medium-sized enterprises. This study focuses on low-cost visual inspection systems, enhancing performance through the selection of domestic industrial cameras, optimization of OpenCV and lightweight deep learning model algorithms, and the use of a C++ parallel computing framework, thereby constructing a solution that balances accuracy and cost. Experiments demonstrate that the system achieves sub-millimeter-level positioning and highly reliable detection in scenarios such as assembly guidance and defect identification, significantly reducing hardware costs while maintaining millisecond-level response capabilities, providing a feasible path for the intelligent upgrading of small and medium-sized enterprises.