How do lidar "eyes" work?
Before talking about why dust affects the recognition effect of lidar, we need to first clarify how lidar works.
LiDAR (LiDAR, full name Light Detection and Ranging) is an active sensor that emits a laser beam by itself, and the laser beam reflects back after hitting surrounding objects. By measuring the time it takes for each laser pulse to return from emission, the distance and direction of the target object can be calculated, thereby constructing a three-dimensional point cloud of the surrounding environment.
This design can obtain very accurate environmental information under ideal conditions, but it will be greatly affected if it encounters objects such as raindrops, smoke, dust, etc. These obstacles will affect the laser beam, thus affecting the quality of the returned signal.
How does dust interfere with laser signals?
When humans drive cars, if there is dust in the environment, it actually has little impact. But for lidar, dust is actually a very troublesome source of interference.
When the laser beam encounters dust particles in the air, scattering occurs, and the light that should originally travel in a straight line is deflected by the dust particles. Such scattering will make the return signal weaker and more blurry, and some light may not even return to the receiving end. The more dust there is, the more serious the light spot scattering will be, and the weaker the detected effective signal will be. This will eventually manifest itself as an increase in noise in the point cloud data, unclear object outlines, and even a misjudgment by the system that there is no obstacle.
In addition to deflecting light, dust also causes the beam to lose energy during propagation, causing the signal strength received by the radar receiver to decrease. Once the signal strength drops to around the sensor's noise level, it becomes difficult to accurately distinguish between real reflections and background noise, which directly affects ranging accuracy and the ability to identify distant objects.
Dust can also cause contamination of LiDAR viewing windows. LiDAR transmitting and receiving beams need to pass through a transparent protective glass or window. If there is dust attached to the surface of this window, and it gradually accumulates and becomes thicker over time, the laser will produce diffuse reflection and absorption when passing through this layer of pollution, and the signal of the beam going out and coming back will be weakened or even change its direction. This kind of physical occlusion has a great impact on the overall quality of the point cloud. Not only will the distance measurement be inaccurate, but it may also make the system mistakenly believe that there is an obstacle ahead or not see the real object at all.
How to reduce the impact of dust on lidar
In fact, many countermeasures have been proposed and applied to dust interference.
One idea is to reduce the adhesion of dust to the window from the hardware. In the design of the radar's shell material and coating, materials with high light transmittance and strong anti-fouling ability can be used to reduce the accumulation of dust on the protective cover, thereby ensuring that the laser is as little blocked as possible. For example, in some application scenarios, protective covers with nano-antifouling coatings on the surface are used to prevent dust from adhering and extend the cleaning cycle of the equipment.
At the software level, the industry has also developed targeted filtering and recognition algorithms. These algorithms will combine the intensity and distance of the laser echo and the distribution of points around the point cloud to determine which points are more likely to be noise caused by dust scattering, and then remove them from the point cloud data. Such a "dust removal algorithm" can restore the point cloud information of the real environment to a certain extent and reduce the impact of false obstacles.
Another method is sensor fusion, which is to combine lidar with other types of sensors. For example, cameras can provide image information to help distinguish dust from real targets. Millimeter-wave radar has better penetrating capabilities for rain, fog, and dust. Combining them can form a more robust perception system, which is much more reliable than a single lidar in complex environments.
In some special extreme scenarios, active cleaning measures will be added, such as installing air blowing devices, brushes or other mechanical cleaning modules on the outside of the lidar to regularly clean away dust on the surface of the window. However, this type of solution has higher cost and maintenance requirements and is mainly used in industrial or special robot environments.
In conclusion,
dust affects LiDAR in many ways. It not only disrupts the laser propagation path but also reduces signal strength, contaminates the sensor window, and ultimately leads to increased noise in point cloud data, decreased recognition accuracy, shortened detection range, and even misjudgment of obstacles. For safety-critical applications like autonomous driving, these impacts cannot be ignored.
