Generate point-cloud diagram

The wide-angle LiDAR we designed is based on the ToF principle and uses the flight time of the beam to calculate the distance between the object and the light source. In addition to adding a MEMS mirror and a wide-angle lens, in addition to extending the one-dimensional single-point laser range measurement into a three-dimensional space detection, it also increases the field-of-view of the system through the wide-angle lens to achieve a small size and high-performance LiDAR scan system. Through the scanning of our wide-angle LiDAR system, the depth relationship of each position in space can be captured on the computer. But if only the distance between the position of each object in the space and it can be obtained, only a large number of digits will be displayed. Therefore, we developed an image processing program by ourselves, which was written in python.

The scan result of the LiDAR system is to use the display window in the Arduino control program to read the spatial information, and export it into a file in the form of txt. Then these files with only digital information are imported into the image processing program and combined into one a picture showing depth information using color. Because the ranging unit in our scanning system, namely the laser light source and photodetector, uses a modular laser rangefinder, the hardware architecture cannot be changed too much, so we independently design the ranging unit and the imaging unit. Because the equipment manufacturer of the laser rangefinder only provides the option of using Arduino control, we developed the image processing unit with python separately, and imported the data measured by LiDAR into it to form a point cloud image. Our system limits the number of points detected due to the limitation of the performance of the ranging module. In Fig.

4.7, there are 1000 distance data points in the vertical direction, and one data point every

degree in the horizontal direction. As shown in Fig. 4.6, we place cardboards at different distances in the corridor, and use our LiDAR to scan its depth relationship, and in Fig.

4.7, the detected data is processed into a depth image showing the distance in different colors.

Fig. 4.6 Place cardboards at different distances on the floor, and use our LiDAR to scan its depth relationship.

Fig. 4.7 The depth image calculated using the detected data.

In Fig. 4.7, the scanning system has not added a wide-angle lens, so the range that can be scanned is small. We then tested the effect of the wide-angle lens installed in the scanner. As mentioned in section 4.2, because the lens at the receiving end of the ranging

module severely limits the maximum measurement angle, the lens and lens housing in front of it are removed to increase the angle of the maximum receiving optical signal. But doing so will affect the farthest distance that can be detected, because after removing the lens at the receiving end, the photodetector cannot detect the weak light signal bounced back from the distant target. When we measure LiDAR with a wide-angle lens, we place the target closer to ensure that we can receive the optical signal at every angle. In Fig. 4.8, we use LiDAR with a wide-angle lens and LiDAR without a wide-angle lens to scan the target object. According to the results of the scan, the field-of-view of the LiDAR with the wide-angle lens has been increased to 100 degrees, which allows the scanning range to be larger. Finally, we adjusted the fast axis of the MEMS mirror to be parallel to the ground, so that the horizontal field-of-view of the scanning range reached 100 degrees.

As shown in Fig. 4.11, we made the abbreviation "NTU" of National Taiwan University and scanned it with a wide-angle LiDAR to compare the difference before and after adding a wide-angle lens.

Fig. 4.8 The scanned object. The results of the scan are shown in Fig. 4.9 and Fig.

4.10.

Fig. 4.9 The scan result of the wide-angle lens has not been added.

Fig. 4.10 After adding the wide-angle lens, the maximum scanning angle reaches 100 degrees.

Fig. 4.11 We made a "NTU" card and scanned it with LiDAR to compare the difference before and after the wide-angle lens was added. The results of the scan are shown in

Fig. 4.12 and Fig. 4.13.

Fig. 4.12 The scan result of the wide-angle lens has not been added.

Fig. 4.13 Add the scan result of the wide-angle lens. In this scan, we rotated the scanning direction of the entire system by 90 degrees, so that the maximum horizontal

angle reached 100 degrees.

We summarize the error of the scanning result of each angle as shown in Fig. 4.14, and the target when measuring the error is set at 2.5 meters away. Because the wide-angle lens is axisymmetric, we measured the scan error from 0 degrees to 50 degrees. Here we define scanning error ∆ℎ ≡ ℎ^′− ℎ, where ℎ^′ is the actual measurement result value, and

ℎ is the actual distance. After the calculation, the ∆ℎ generated is divided by the actual distance of the target to obtain the scanning error expressed as a percentage. From the measurement results, the measurement error of our LiDAR system is very small, so we can get very good quality images.

Fig. 4.14 The measurement error of each angle, the target is set at 2.5 meters away. At large angles, because the beam diameter is large and the incident angle of the light beam

received by the photodetector is large, the intensity of the light signal per unit area is low, so the error is relatively large. However, the maximum error value is within 2.5%,

which has almost no effect on the imaging quality of the point cloud diagram.