The earliest definition of lidar is LIDAR, English is Light DeteaTIon and Ranging, Chinese meaning "light detection and ranging".
In fact, a more accurate definition is LADAR: LAser DetecTIon and Ranging, that is, "laser detection and ranging". This is a definition proposed in 2004, which is more in line with the concept of lidar.
Lidar is actually a kind of radar working in the optical band (special band), and its advantages are very obvious:
1. Extremely high resolution: Lidar works in the optical band, and the frequency is 2 to 3 orders of magnitude higher than microwave. Therefore, compared with microwave radar, lidar has extremely high range resolution, angular resolution and speed Resolution
2. Strong anti-interference ability: the laser wavelength is short, it can emit a laser beam with a very small divergence angle (on the order of μrad), the multi-path effect is small (will not form directional emission, and generate multi-path effects with microwave or millimeter wave), detectable Low-altitude/ultra low-altitude targets;
3. The amount of information obtained is rich: the distance, angle, reflection intensity, speed and other information of the target can be directly obtained to generate the multi-dimensional image of the target;
4. Can work all day: laser active detection, does not depend on external lighting conditions or the radiation characteristics of the target itself. It only needs to emit its own laser beam and obtain target information by detecting the echo signal of the emitted laser beam.
But the biggest disadvantage of lidar is that it is easily affected by atmospheric conditions and smoke from the working environment. It is very difficult to achieve an all-weather working environment.
Lidar classification
If the classification of lidar is divided into systems, there are mainly direct detection lidar and coherent detection lidar. In fact, what we have mentioned at present, including automatic driving, robots, and lidar used for surveying and mapping, basically belong to this type of direct detection lidar. Some special radars, such as wind measurement and speed measurement, generally use a coherent system. If the classification of lidar is divided into systems, there are mainly direct detection lidar and coherent detection lidar. In fact, what we have mentioned at present, including automatic driving, robots, and lidar used for surveying and mapping, basically belong to this type of direct detection lidar. Some special radars, such as wind measurement and speed measurement, generally use a coherent system.
According to the application classification, we can divide more, such as: laser rangefinder, laser three-dimensional imaging radar, laser speed measurement radar, laser atmospheric detection radar, and so on.
Whether it is single-line lidar, multi-line lidar or surveying lidar, we can basically divide it into the category of laser three-dimensional imaging radar.
A laser three-dimensional imaging radar, in fact, it needs to obtain two core information: target distance information and target angle information.
If we determine its three-dimensional sitting standard, we need to get its distance, azimuth, and pitch angle information. Then we calculate the three-dimensional coordinate point of the target based on the three information of distance, azimuth angle and pitch angle.
Generally speaking, the technique of obtaining angle information by measuring the encoder is very mature. We are more concerned about how to obtain the distance information of the lidar.
The laser three-dimensional imaging radar can obtain the three-dimensional point cloud data of the target through direct ranging and direct angle measurement technology, and the obtained data is itself three-dimensional data. It does not require a large amount of calculation and processing to generate the target three-dimensional image, and the laser ranging has Very high precision.
Therefore, the laser three-dimensional imaging radar is currently the most efficient sensor that can obtain images of a large range of three-dimensional scenes, and is also the sensor that can currently obtain the highest accuracy of three-dimensional scenes.
A
Laser ranging method
At present, the distance measurement methods we can usually see can be divided into broad categories: laser time-of-flight (TIme of Fly, TOF) method and triangulation method.
The laser time-of-flight method can be divided into two categories, one is pulse modulation (pulse ranging technology), and the other is the phase modulation of the laser continuous wave intensity modulation, which measures the distance information through the phase difference.
The rangefinders we can see on the market, or single-line and multi-line lidars, basically use these three types of ranging methods.
A
Laser pulse ranging technology
The principle of laser pulse ranging technology is very simple: obtain the distance information of the target by measuring the time of flight of the laser pulse between the radar and the target. A benchmark is used here, which is the speed of light. All measurements must have a datum. For a laser, there are two datums: speed and frequency (the two most accurate datums), because the datum used for TOF is the flight speed of the laser.
Among the three ranging methods mentioned above, I think the most difficult technical problem is the pulse ranging method. But the advantages it brings are obvious: the measurement speed is very fast. Since the measurement is performed by a laser with a high peak value, its anti-jamming ability is very strong.
The disadvantage is that it is difficult to improve the ranging resolution and the detection circuit is difficult. For example, if we want to achieve a resolution of 1.5 millimeters for phase ranging, we need to achieve a timing clock resolution of 10 picoseconds, which is equivalent to 100G bandwidth. This is a very difficult technique.
A
Laser phase ranging
Laser phase ranging, such as common handheld laser range finder, uses phase ranging. It mainly obtains distance information by measuring the phase difference generated by the intensity-modulated continuous wave laser signal flying back and forth between the radar and the target.
The biggest advantage of this technology: the range resolution is very high. At present, the phase range finder on the general market can achieve millimeter-level resolution.
The disadvantage is that the measurement speed is slower than pulse ranging. After all, we need to calibrate a phase difference at least tens or even hundreds of cycles. In fact, it is equivalent to lengthening its measurement time in phase, then its measurement speed Relatively low. In addition, its measurement accuracy is relatively susceptible to target shape movement. If in the measured light spot, the two targets are in tandem, the specific information it actually measured is an average of the distance between the two targets, not the previous target information or the next target information.
But in pulse ranging, it is easy to separate such information. For example, for a laser pulse, if we can achieve a pulse width of 10 nanoseconds, then we can distinguish a target that is 30 centimeters from front to back by multiple echoes.
It is difficult to distinguish this method in phase ranging. Because in the measurement process, its time will be longer, and the distance information brought in by the target motion is introduced into the measured value. In fact, it measures an average distance information, not real-time information. But laser pulse ranging is actually real-time information about the current position.
This is why lidar for vehicles or robots often uses laser pulse ranging technology instead of phase ranging technology.
A
Triangulation
Triangulation distance measurement is to obtain distance information by measuring the imaging position of the laser irradiation point in the camera. The biggest advantage of the triangulation method is that the technical difficulty is low, the cost is also very low, and the accuracy of ranging at close range is also very high. For example, industrial use can achieve 100-micron ranging accuracy.
But the disadvantage is that its accuracy will gradually deteriorate with the increase of distance, and basically can not be compared with pulse ranging and phase ranging.
Another point, because the CMOS camera must use a continuous laser to illuminate synchronously, its average power is relatively low, and its anti-interference ability will be very strong. This method of ranging is generally suitable for indoor close-up work, but not suitable for Work under outdoor glare background or indoor glare background.
Triangulation distance measurement is more suitable for scenes with low performance requirements such as robots. In addition to the relatively high cost and technical difficulty, pulse ranging has excellent performance in other aspects. Of course, its ranging accuracy will be slightly lower than that of phase ranging. But this kind of accuracy, according to the current technology, we can basically reach the distance measurement accuracy in the order of centimeters, or even a few millimeters, which can basically meet the requirements of our use in many occasions.
Our main direction is to use pulse ranging to do single-line radar, including multi-line radar.
A
What is single-line lidar
The single-line lidar is actually a high-frequency pulse laser rangefinder, plus a one-dimensional rotation scan. Features of single-line lidar:
1. There is only one way to transmit and one way to receive, the structure is relatively simple and easy to use;
2. High scanning speed and high angular resolution;
3. Low volume, weight and power consumption;
4. Higher reliability;
5. Low cost;
A
What can single-line lidar do?
In the field of autonomous driving, we basically see multi-line lidar, single-line laser
What can radar do?
As shown in the figure above, the first car to participate in the DARPA Autopilot Challenge in the United States is the 2005 Stanford car named Stanly. This is the car that won the championship that year. The other is the car from Carnegie Mellon University.
At that time they were basically using single-line lidar. Especially for the Stanford University competition car, there are five lidars installed just above, we can think of it as the originator of multi-line lidar, but it uses five single-line lidars to achieve multi-line lidar Function.
After Velodyne launched 64-line lidar in 2007, many self-driving vehicles basically used Velodyne products. But does this mean that single-line lidar has no market in assisted or autonomous driving? I don't think so. Because single-line lidar has its characteristics, for example, it is difficult for multi-line lidar to achieve the same technical indicators at high repetition rate and high angular resolution. In terms of pedestrian detection, obstacle detection (small target detection), and front obstacle detection, single-line lasers have many advantages over multi-line lidars, because single-line laser radars can have a higher angular resolution than multi-line lidars. This is very useful in detecting small objects or pedestrians. This technology is very useful in intelligent robots and service robots, and this one is also a hot field.
Many people may ask a question, why use lidar for lane detection instead of a camera. Isn't the ADAS algorithm very mature? Why must I use lidar?
This is because the camera is particularly susceptible to interference from background light or strong light. For example, when we walk on a tree-lined avenue, if the shade of the trees falls in spots and then combines with the white lane lines, it is very difficult to recognize the lane lines, and the recognition probability is under complex lighting or strong light conditions. Its recognition probability is very, very low, and the algorithm is very complicated.
So, what are the benefits of using lidar for lane detection? First, we are using infrared lasers, which have much lower radiation in the infrared band than visible light. Second, we will add a very narrow filter to filter out the strong background light directly. Then we use infrared light to detect it. In this way, we can obtain a very high-quality image of the lane line, and through the grayscale of the image, it is very easy to detect the lane line. In other words, using lidar for lane line detection, its performance will be higher than the camera.
The application of single-line lidar in assisted driving is pedestrian detection. In fact, this is also a forward collision prevention application, which is basically similar to automobile collision prevention. Because the angular resolution of the single-line lidar can be higher than that of the multi-line lidar, pedestrians can be detected in advance at a greater distance, leaving more warning time for the control system or the driver.
