Conventional LiDAR-based depth imaging systems

Compared to AM-CW and FM-CW LiDAR systems, direct time-of-flight ranging based on a pulsed laser is more straightforward, simplifying the configuration of LiDAR systems. In addition, direct time-of-flight LiDAR systems with discrete detectors combined with spatial scanning mechanisms (e.g. mirror-based raster scanning and

refractive-optics-based Risley prism scanning) or using focal panel arrayed (FPA) detectors are capable of depth imaging. To implement 2D mapping of targets, imaging using a discrete detector observes a series of spatial origin of instantaneous field of views (IFOVs) sequentially to cover the entire FOV of interest. In contrast, the FPA-based imaging is able to stare the entire FOV formed by pixel-wise IFOVs. LiDAR-based depth imaging systems can map the world in three dimensions which corresponds to an assembly of spatial-point-wise depths [2.123, 2.147]. As opposed to passive optical imaging techniques with stringent dependence on light conditions related to the diurnal cycle, LiDAR-based depth imaging systems that are active optical imaging systems have dedicated illumination sources to initiate photons and can retrieve information from the detected photons. This makes them powerful sensors for laser-based remote sensing [2.124, 2.148-2.149] and more recently computer vision communities [2.150].

Conventional hundred-metre-range LiDAR-based depth imaging systems employ laser sources with a pulse width in the order of a nanosecond FWHM and photon detectors with linear amplified gain. They can resolve depth resolutions of typically higher than a centimetre. For example, commercial scanning LiDAR sensors (e.g. HDL-64E S2 by Velodyne CA, USA [2.151]) can obtain high-definition 3D mapping of a target scene.

The high resolution 3D image of urban scenes as shown in Figure 2-58 depicts some hard targets (e.g. buildings and street curbs) and distributed targets (e.g. trees and overhead wires) clearly. The image was acquired using a high definition LiDAR sensor from Velodyne CA, USA. The wide-field-of-view LiDAR sensor using 64 lasers with 5 ns pulse duration time operating at λ=905 nm has range accuracy of 1.5 cm at a working distance of ~100 m. In addition, it can generate ~1.3 million data points per second.

Figure 2-58 High resolution 3D point-cloud mapping of urban scenes including hard targets (e.g. buildings and street curbs) and distributed targets (e.g. trees and overhead wires). Colours represent the depth information. The image was acquired using a high definition LiDAR sensor from Velodyne Inc., USA. The wide-field-of-view LiDAR sensor integrated with 5-ns-pulse-duration 64 lasers operating at λ=905 nm has range accuracy of 1.5 cm at a working distance of ~100 m and is capable of generation of ~1.3 million data points per second. (From [2.147]).

An example of an airborne LiDAR-based depth imaging system configuration using a detector array is shown in Figure 2-59. As for LiDAR-based 3D imaging systems using detector arrays and flood illumination, Advanced Scientific Concepts (ASC), Inc., USA produces 3D imaging cameras using FPA detectors (e.g. linear-mode InGaAs APD arrays with 128×128 pixels), which can be called staring 3D imagers. Some field trial results (for example, the one shown in Figure 2-60) at hundred metre distances using this type of depth imager with 128×128 FPA devices were reported using pulsed laser sources at 1540 nm and 1570 nm wavelengths with a few ns pulse duration and 10s mJ pulse energy [2.152-2.153].

Figure 2-59 Block diagram of an airborne LiDAR depth imaging system using arrayed detectors (i.e. focal plane array in the diagram). Ladar: laser radar;

GPS/IMU: global positioning system/inertial measurement unit. (From [2.154]).

Figure 2-60 Top: A photograph of a parking lot at a stand-off distance of ~300 metres.

Colours represent the depth information. Bottom: 3D images in two views of a parking lot at ~300 metres acquired by a 128×128 3D imaging camera (by Advanced Scientific Concepts, Inc., USA) which was put in an airplane platform. (From [2.153]).

In addition, another type of staring depth imager which uses burst illumination can be built based on a range-gated mechanism. This type of system requires prior knowledge

of the working distance in order to control the delayed gates appropriately. As shown in Figure 2-61 the system retrieves depth profiling of the target by integrating a series of gated-viewing intensity images, which were acquired by a CCD-based sensor with a precise gate control module. For example, French–German Research Institute of Saint-Louis (ISL) demonstrated a gated viewing active imaging system working up to a distance of 1500 m at λ=808 nm [2.155].

Figure 2-61 Depth imaging based on the range-gated mechanism. Left: Time-delayed slices of intensity images of target scene acquired by a CCD-based sensor with a precise gate control module. Middle: Post-processing of the time-related intensity information associated with the time-delayed slices; Right: The reconstructed depth profile of the target scene consists of a building and trees. The depth value presents the reflective intensity data. (From [2.156]).

This type of system is usually also called a burst-illumination LiDAR system. Due to its range-gating operation, this imaging technique is known to overcome strong backscattering from obscurants such as camouflage, fog or water. It can achieve time-delayed slices to represent the depth profiles of targets. For example, by configuring the CCD-based or other FPA-based devices, gated viewing depth imagers operating in kilometre-range free space at λ=1550 nm [2.157] or short-distance underwater at λ=532 nm [2.158] have been developed. In addition, Figure 2-62 shows 2D burst-illumination LiDAR image and a simultaneous range image of a vehicle and building at a stand-off distance of >10s metres. They were obtained by a burst-illumination LiDAR system based on a 320×256 HgCdTe APD array with 24 µm pitch. A laser source with 20 ns pulse duration.at a wavelength of 1550 nm was used to provide illuminating light [2.159]. Parameters of the mentioned state-of-the-art conventional LiDAR-based depth

imaging systems are summarised in Table 2-1.

Figure 2-62 2D burst-illumination LiDAR image (left) and a simultaneous range image (right) of a vehicle and building at a stand-off distance of >10s metres obtained by a burst-illumination LiDAR system based on a 320×256 HgCdTe APD array with 24 µm pitch at a wavelength of 1550 nm. The range image is encoded in colour.

Along the line-of-sight of the sensor, blue presents the nearest distance to the sensor whereas red corresponds to the furthest away range. (From [2.159])

[2.151] Table 2-1 Comparison of performance to some state-of-the-art conventional LiDAR-based depth imaging systems.

However, as the working distance is extended to kilometre range or longer, the limited pulse energy of the laser sources significantly impairs the detection performance of LiDAR-based depth imaging systems. In particular, the FPA-based systems are affected due to considerable atmospheric attenuation and turbulent effect. (refer to [2.123,

2.135, 2.160] and Appendix A of [2.161]).