1. Introduction Reverse engineering is a rapidly developing emerging technology. Physical reverse engineering is used for product redesign based on existing physical products, or for the digital conversion of product models (such as clay models) that cannot be directly represented or designed digitally but can only be expressed in physical form, to implement digital design and manufacturing of these products. For example, when designing the aerodynamic shape of aircraft, missiles, and other flying vehicles, the physical model obtained entirely based on experimental optimization cannot be directly modeled and represented using existing CAD/CAM systems. Therefore, to achieve the digital design and manufacturing of these flying vehicles, it is necessary to use point cloud data acquisition technology in physical reverse engineering to convert the physical model into a CAD model. Currently, physical reverse engineering technology has been widely used in new product design, product repair, and online product testing. In physical reverse engineering, to achieve object digitization, it is necessary to use appropriate measurement or scanning equipment to measure or scan the three-dimensional physical model of the product to obtain the spatial topological discrete point cloud data of the physical model. Therefore, point cloud data acquisition is the first task that must be completed in physical reverse engineering. Among various physical measurement technologies, the optical scanning point cloud data acquisition technology that has emerged in recent years has many advantages such as high measurement efficiency, good data integrity, wide applicability, and wide data acquisition range (from a few millimeters to tens of meters). In recent years, many universities and research institutes in China have conducted a lot of research and development on physical reverse engineering technology and have made breakthroughs in some key technologies. Through independent development and technical cooperation, China has become increasingly mature in the development and application of contact measurement technology and equipment. However, for non-contact measurement optical scanning point cloud data acquisition technology, there are very few domestically developed mature products and their successful applications in the automotive, mold and other industries. In order to promote the research and development of optical scanning point cloud data acquisition and processing technology and provide external technical services, our school has introduced a mobile optical scanning device produced by a foreign company. During the use of the equipment, the problem of excessive measurement error often occurs due to operational errors, improper adjustment and other reasons. In order to thoroughly digest the foreign-introduced technology, the author has conducted research and analysis on this issue based on work practice and proposed corresponding solutions. [b]2 Main forms of measurement error[/b] (1) Missing data or insufficient data density. Using such incomplete data for point cloud fitting results in a large error and makes it difficult to achieve the required measurement accuracy. (2) Data acquisition results for the same surface are multi-layered point clouds. This often occurs when the measured object is a large workpiece or a transparent object. (3) Inaccurate single-frame data acquisition affects the overall measurement accuracy. (4) Excessive cumulative error causes significant deviation in measurement results. (5) Incorrect point cloud splicing leads to a large measurement error. (6) Too many coarse points (noise) in the measurement results. 3 Error Cause Analysis and Countermeasures to Improve Accuracy Based on practical work experience, the author summarizes the main causes of large measurement errors through testing and analysis as follows: improper calibration, improper use of the scale; improper selection of probe lens combination; improper measurement sequence; improper selection of measurement strategy; improper placement of workpiece surface markers; improper operation during measurement; improper pre-processing of the workpiece surface; improper post-processing; improper selection of measurement environment, etc. The following are the analysis results: (1) Improper calibration and improper use of the scale The scanning measuring head (probe) consists of a light source, a CCD camera and a corresponding lens group. Before acquiring point cloud data, the probe must first be initialized. This mainly includes: ① Selecting different lens combinations based on the size, number, and complexity of the surface features of the object being measured; ② Determining the light intensity of the main light source based on the measurement site conditions, the surface morphology of the object, and its surface treatment; ③ Calibrating the selected lens combination according to the system's standard workflow to ensure a calibration accuracy ≤ 0.020. If the above steps are not performed before measurement, and a previously calibrated probe is used directly, the measurement accuracy may be compromised due to incompatibility between the lens combination, light source intensity, and calibration accuracy, leading to significant errors. During measurement, if the probe is subjected to impact or collision due to operational errors, it should be inspected immediately. If damaged, it should be repaired; if not damaged, the probe must be recalibrated. Even if no operational errors occur during measurement, if the measurement time is long, the probe should be quickly calibrated periodically to check its accuracy. The scale is an essential tool for locating the marker points on the entire workpiece when collecting data from large workpieces using a digital camera. The standard dimensions on the scale should be consistent with the dimensions displayed when processing the actual photograph. (2) Improper selection of probe lens combination When collecting point cloud data of the surface of large workpieces, a lens combination with a larger measurement range should be selected to achieve rapid acquisition of overall data. For some areas with more and smaller features, it is best to select a lens combination with a smaller measurement range to perform local small feature highlighting measurement in order to obtain better measurement results. For large workpieces, if a lens combination with a smaller measurement range is selected for measurement, there will be more marker points on the workpiece surface for point cloud stitching. This will prolong the workpiece preprocessing time, increase the measurement time span, and the error caused by the change of ambient temperature over time will be reflected in the measurement results, and will affect the overall measurement efficiency. If a digital camera is used to locate the markers on the entire workpiece, and the measurement is automatically stitched together, the stitching error is likely to occur because there are many markers and the probability of the markers having the same relationship is increased. If a digital camera is not used to locate the markers on the entire workpiece, and adjacent single point clouds are stitched together using common markers, the stitching error will be too large due to the large number of stitching times. Conversely, for small workpieces, if a lens combination with a large measurement range is selected for measurement, it will not be able to accurately reflect the small features on the workpiece, and the measurement results will not meet the required accuracy. It is necessary to replace the lens combination and recalibrate and measure. (3) Improper measurement sequence The measurement sequence refers to the superposition order between adjacent single measurement results during measurement. Taking the measurement of a slender workpiece as an example, the rectangular areas shown in Figures 1, 2, 3, 4, and 5 are the measurement range of the current calibration probe. When measuring a workpiece using the radial arrangement shown in the figure, the first measurement is taken at the center of the workpiece. After measuring the first image, the second image is measured. Then, the second image is combined with the first image using three common marker points. This process introduces a splicing error. Similarly, a splicing error occurs when the third image is combined with the second image. Assuming all splicing errors are of equal magnitude, the cumulative error between images 1, 2, and 3 is 28, and the cumulative error between images 1, 4, and 5 is also 28. The measurement results show that the cumulative errors between images 1, 2, and 3 do not overlap with those between images 1, 4, and 5; therefore, the total cumulative error remains 28. If the workpiece is measured using the arrangement shown in Figure 2, the maximum cumulative error is 48. Therefore, the measurement sequence of "radial arrangement with the center as the reference" should be used whenever possible to reduce cumulative errors. (4) Inappropriate selection of measurement strategy. When measuring, the workpieces to be measured should be classified into large workpieces, medium workpieces, small-sized multi-feature workpieces, internal cavity workpieces, etc., and different measurement strategies should be adopted for each type of workpiece. When measuring large workpieces, the overall positioning of the marker points can be performed first with a digital camera, and then a lens combination with a large single-frame measurement range can be selected for measurement. If the workpiece size is too large, it can be measured twice, and then the common reference points can be used to combine them. If there are many small local features in the large workpiece, a lens combination with a smaller measurement range can be selected for local measurement after the basic measurement is completed. In order to facilitate the automatic combination of small-scale measurements, the density of reference points should be increased when preprocessing the local area. When measuring medium and small workpieces, attention should be paid to adopting the correct measurement sequence to reduce cumulative errors. In fact, the measurement strategy for large workpieces can also be used for medium and small workpieces, but a digital camera and related software for overall positioning of reference points must be provided. When measuring the inner cavity surface of a workpiece, in order to overcome the depth-of-field limitation of optical scanning measurement equipment, some technical means can be adopted to transform the inner cavity measurement into the outer shape measurement. For example, silicone can be injected into the inner cavity of the workpiece, and after it solidifies, it can be taken out and its shape measured. (5) Improper placement of workpiece surface markers Whether it is a large workpiece or a medium or small workpiece, the problem of placing workpiece surface markers will be encountered. Data acquisition of large workpieces generally requires the use of a ruler and a digital camera. The measurement can be carried out in two steps: First, the markers used for single-frame measurement point cloud stitching are constructed as a whole using large digital points. In order to ensure that the single-frame measurement point cloud is obtained through normal calculation, the arrangement rules shown in Figure 1 must be followed; Second, the marker point cloud is used as a reference system. The system will compare the markers in each single-frame point cloud with the markers in the existing reference point cloud. If the two match, they will be automatically stitched together. There will be no need for overlapping parts between two adjacent single-frame point clouds. Measurements can also be performed, but each single point cloud must contain at least three marker points. In this case, marker points need to be appropriately affixed to the measured area; otherwise, measurement difficulties or decreased accuracy will occur. The marker point affixing for medium-sized workpieces differs from that for large workpieces. Since the automatic (or manual) stitching of two adjacent point clouds requires the common marker points of adjacent single point clouds, the marker point affixing density for medium-sized workpieces should be greater than that for large workpieces; otherwise, it will be difficult to stitch adjacent point clouds together. Because the placement of marker points on small workpieces will obscure the features of the workpiece to varying degrees, marker points should be affixed to the workpiece surface as few as possible or not at all to obtain more complete scanning data. In addition, marker points should generally be affixed to relatively flat positions on the workpiece to reduce the difficulty of filling in point cloud gaps at the marker points and the corresponding measurement errors. (6) Improper operation during measurement. During the measurement process, the following key points should be noted: ① Adjust the position of the probe so that the part to be measured is within the measurement range of both probes at the same time; ② Adjust the light intensity of the main light source and adjust the clarity of the marker points and the workpiece surface respectively so that the marker points and the workpiece surface of the measurement part are as clear as possible; ③ Avoid impact or collision with the probe as much as possible during the measurement process. If this happens accidentally, the probe should be checked and recalibrated in time to maintain the accuracy of subsequent measurements. Otherwise, the measurement results will show missing data of the measurement parts, or even make the measurement impossible to continue. (7) Improper pretreatment of the workpiece surface. Before starting the measurement, the workpiece surface needs to be properly pretreated. If the workpiece shape is very simple and the workpiece size is small, data acquisition can be completed by single-frame measurement. Then, it is only necessary to make the workpiece surface diffusely reflective under the illumination of the main light source. However, under normal circumstances, it is difficult to complete the data acquisition of a complete workpiece through single-frame scanning measurement alone. Furthermore, the surface of a typical workpiece is unlikely to form diffuse reflection that meets measurement requirements under main light source illumination. Therefore, it is necessary to pre-set some reference points on the workpiece surface, use these common reference points to stitch together the measurement results, and uniformly spray a colorant onto the workpiece surface to create a more ideal diffuse reflection. Improper pre-treatment of the workpiece surface mainly refers to: ① Excessive reflection or absorption in certain areas of the workpiece surface, preventing the formation of diffuse reflection suitable for scanning requirements, resulting in the inability to form an effective point cloud, and measurement results showing missing data for that area; ② Lack of sufficient reference points, making stitching impossible, and even if a point cloud can be formed, it is only a scattered point cloud rather than a complete point cloud; ③ Overly consistent adhesion of reference points on the workpiece surface, lacking distinctive features, making it difficult for the system to effectively identify the stitching position of single-frame point clouds, thus easily causing stitching errors and making it difficult to form a complete point cloud of the workpiece. Improper pre-processing of the workpiece surface to be measured also includes non-measurement factors such as failure to correct parts of the workpiece surface that do not correctly reflect the design intent, failure to repair workpiece surface damaged during measurement, and improper workpiece placement (such as workpiece stress). In addition, when silicone is injected into the inner cavity of the workpiece (such as engine air passage) to form a model, insufficient injection volume or too many air bubbles in the silicone will also cause the formed model to not correctly reflect the actual shape of the inner cavity of the workpiece. (8) Improper post-processing In optical scanning measurement, the measured data is not necessarily the point cloud data. The measurement process is actually the process of forming a workpiece image. To obtain point cloud data, it is necessary to use the ATOS system to post-process the formed image data. For the result generated by stitching together single point clouds, it is first necessary to use several common marker points to align all the single data to reduce the cumulative error; then, the aligned point cloud is recalculated to convert the image data into point cloud data. At this time, the point cloud data may still have problems such as uneven density and many coarse error points. It can be further processed by triangular meshing (Polygonize) to finally obtain point cloud data with better quality. Of course, post-processing includes more than the above. In actual measurements, the data points obtained through scanning are not necessarily limited to the physical model being measured. Some random points in the measurement environment that do not belong to the model are also measured simultaneously. Therefore, these unnecessary points must be removed during post-processing to reduce the possibility of errors when constructing a 3D CAD model based on the point cloud data. Furthermore, post-processing also includes simplification of the point cloud. In a reverse engineering point cloud of a physical object, the accuracy requirements for different parts of the workpiece are not entirely the same. Therefore, for some less important parts, the point cloud density can be reduced; for some more important parts, the point cloud density can be increased. This not only ensures the accuracy requirements of the 3D model construction but also greatly improves modeling efficiency. Of course, post-processing includes more than just the above. In actual measurements, the data points obtained through scanning are not necessarily limited to the physical model being measured. Some random points in the measurement environment that do not belong to the model are also measured simultaneously. Therefore, these unnecessary points must be removed during post-processing to reduce the possibility of errors when constructing a 3D CAD model based on the point cloud data. Furthermore, post-processing also includes simplification of the point cloud. In reverse engineering point cloud acquisition of a physical object, the accuracy requirements for different parts of the workpiece are not entirely the same. Therefore, for some less important parts, the point cloud density can be reduced for simplification, while for some more important parts, the point cloud density can be increased. This not only ensures the accuracy requirements of the 3D model construction but also greatly improves modeling efficiency. 4. Conclusion In recent years, reverse engineering technology has played an increasingly important role in new product design, product modification design, and mold manufacturing, and its applications in automobile manufacturing, aerospace, machine tools, national defense, electronics, and mold making are becoming increasingly widespread. However, domestic research on related technologies is still relatively lagging, and related technical equipment still mainly relies on imports. Therefore, researching and developing reverse engineering technologies and equipment with independent intellectual property rights in China and realizing their commercial application as soon as possible is an urgent task in this field. This paper analyzes the accuracy influencing factors in the practical application of optical scanning point cloud data acquisition systems and proposes corresponding countermeasures, hoping to help improve the application level of related measurement equipment and provide some practical experience for the research of reverse engineering technology, thus promoting the continuous improvement of China's reverse engineering technology development and application level.