# 2020生肖表

Position: Home > Technical Articles >

# Analysis of experimental results of 3D human scanning

Release time: 2015-06-04 10:33

Experimental results
1. Calibration In order to determine the accuracy of the device, a calibration target with a known three-dimensional coordinate value of a cylinder (radius 24 em) and a height of 180 cm) and the corners of each color block on the cylindrical surface is designed, as shown in Figure 6. . The height of each color block on the cylinder is the same in the vertical direction and the arc length in the horizontal direction is different. In this way, the coordinate values of each corner point on the cylinder have a non-linear relationship, thereby improving the accuracy of calibration. For real and virtual cameras, at least 30 calibration points are selected for calibration calculations.

An error analysis is performed on the calculated values of the same calibration point in space according to every 2 cameras. For example: two cameras capture the same point P in the space on the calibration target. , The three-dimensional coordinates obtained from the conversion of the images taken by the l camera and the other camera are respectively denoted as P ′, and P ″ .. The error is calculated using the Euclidean distance between P ′. And P ″ i.

The scanning system in this paper selected 31 identical calibration points, with an average error of 2.03 mm and a standard deviation of 0.86 mm.

2. Laser strip center extraction The speed and accuracy of laser strip center extraction will directly affect the performance of the scanning system. Therefore, several common laser strip center extraction algorithms (such as image thinning method and curve fitting method) are studied and compared. , Centroid method, Hessian matrix method, etc.), the centroid method with the advantages of high positioning accuracy and fast calculation speed is used to extract the center of the laser light band '1'. First, the filter method is used to remove the noise interference on the light band, and then use The law obtains the threshold value of the light band image. Based on this threshold value, find the point that meets the threshold value in the unit of column, and initially determine that the extreme point is the center point of the light band and calculate the light intensity threshold value t of the column. . Suppose that the position of the maximum point is the Yth ... row of the column, and the threshold is t. Then select n pixels above and below the maximum point, usually 10-15. Then you can use the formula (14 ) Perform the calculation of the center of gravity of the curve.

, (Z ,, Y.) is the gray value of the corresponding pixel point, and (x., Y.) is the coordinate of the center point of the optical band.
Because the light intensity of the image collected by the CCD camera conforms to the Gaussian distribution, the Hessian matrix method can be used. The normal direction of the light band in the image is calculated, and the extreme point in the normal direction is obtained, so the sub-pixel position of the center line of the light band is obtained. The coordinates of the center point of the light band obtained in this way are more accurate.

3. Human body scanning Select a female with a height of 165 cm for scanning experiments. The scanned person stands facing the real camera. The frontal surface data of the human body was taken directly by a real camera. A total of 3 286 data points were collected with a standard error of 0.73 mm. The maximum distance between two adjacent points in the horizontal direction does not exceed 3.1 mm. The point cloud diagram of human frontal data is shown in Figure 7.

The left and right sides of the back of the human body are not directly taken by the real camera, which is equivalent to the camera taking two virtual cameras in the flat mirror to shoot the image obtained from the left and right sides of the target. 16 784 to 18,372 points on the surface of the human body were obtained, with a standard error of 0.92 to 1.08 mm, and the distance between two adjacent points in the horizontal direction should not exceed 3.9 mm at the maximum. FIG. 8 is a point cloud diagram of the back of the human body after the backs of the left and right sides of the human body are combined. Table 1 shows the time and accuracy of the current major commercial 3D body scanners. Compared with current commercial 3D scanning systems, this article

Although the scanning system has lower accuracy, the scanning time is significantly shorter than most systems. Therefore, the accuracy of the scanning system in this paper can meet the requirements of clothing design and platemaking, and can be used for 3D clothing design and virtual three-dimensional cutting.

More about 3D body scanners : http://airulm17.com/productlist/list-5-1.html