Research on intelligent detection methods for tobacco packaging quality based on 3D point cloud

This paper proposes an intelligent inspection method for tobacco packaging quality based on 3D point cloud technology, addressing the issues of low efficiency, subjectivity, and high missing detection rates in manual visual inspection and traditional automated inspection. By utilizing the Hexagon ROMER83 seven-degree-of-freedom articulated arm measuring machine equipped with the RS5 laser scanner as the data acquisition device, and combining the iterative closest point algorithm, multi-view data fusion technique, principal component analysis (PCA), region growing algorithm, and least squares method, the precise identification and quantitative evaluation of micro-defects on cigarette packet surfaces are achieved. Taking the quality defect of uneven hard package seams as an example, the process of point cloud data acquisition, segmentation, plane fitting, and seam height difference measurement is elaborated. Experimental results demonstrate that this method is efficient, accurate, and stable, with a measurement sensitivity of up to 0.1mm for seam height differences, providing data-driven decision-making bases for tobacco packaging process optimization. Furthermore, future research directions in defect classification model construction, inspection feedback mechanisms, and inspection target expansion are proposed.