Sensor Fusion (LIDAR Camera and Radar )

Images of Camera , Lidar and Radar used in Automotive.

The Lidar shown in the above image is Velodyne lidar with 360⁰ field of view placed at the top of the automotive. We can also place the lidar in front of the car.

Lidar placed in-front of car.

LIDAR:

Lidar system senses thousands of laser pulses every second. Pulses collide with the various objects around the vehicle and reflect back the signals.

These light reflections are used to create the 3-D point cloud. On board computer records the laser reflections , translates rapidly updating point cloud into animated 3-D representation.

In case of lidar point a small infra red laser beam at surface and measure the time it takes for the laser to return to the source.

By having lidar with 360⁰ viewing angle (Using rotating mirror as example) it is possible to obtain point cloud of environment. Then specific software makes 3D image that reproduces shape around lidar with precise position in space.

Lidar system generates a 3D point cloud of the surrounding environment by scanning the beam across the region arround the vehicle.

• A 3D point is created by the scanning the module to a specific angle and firing off a pulse.
• The reflected signal at each point n the cloud is received and converted to distance measurement with time-to-digital converter which gives depth measurement.
• 3D depth map is generated in the process can provide the high resolution image when the reflection occurs very close to the vehicle.
• Lidar system can readily detect the objects located in the range of 30 meters to 200 meters . When it comes to objects identification in the vicinity system is big let down.
• Lidar works well in all light conditions but the performance starts to dwindle in the snow,fog,rain and dusty weather conditions.

What is point cloud ??

Lidar is remote sensing method that uses the light in the form of “pulsed laser” to measure the ranges does kind of 3-D scanning. Point Cloud is created by this scanning technology.

Point cloud comparision between Lidar and Radar.

Sonar uses sound waves , Radar uses radio waves, Lidar uses light waves.

There are basically 2 types of lidar :-

1. 2-D lidar → Uses only single laser beam. Pulse based spin movement to collect the horizontal distance of targets.
2. 3-D lidar → Several beams spread out on vertical axis to get on X, Y and Z.

Lidar generally operates at 50 HZ meaning to say 50 frames per second. Where as the camera operate at 30 HZ meaning to say 30 frames per second.

Where the lidar is placed in car ??

Generally it is mounted on top of the vehicle. Velodyne lidar is 360 degree lidar placed on the top. Some cars it can be placed on the bonnet as well.

Early Fusion of Camera and Lidar

Late Fusion of Camera and Lidar

Performance

RADAR :

• Radar stands for radio detection and ranging.
• Uses radio waves to measure distances to objects as well as the velocity and angle information.
• Unlike lidar, radar is not affected by weather conditions like Fog ,Snow ,Rain and Dust are not obstacles.
• Radar does not have the moving parts and requires less power than the lidar sensor.
• Tesla believes that radar combined with the passive optical sensors can do exactly same job as lidar.
• Finally we have to detect the objects and classify whether the given object is car or pedestrian.
• Point Cloud images from lidar have less accuracy and resolution.
• That is the reason why the radar sensor won’t allow the autonomous vehicle to distinguish all the objects on the road.
• In current vehicles , short range (24 GHZ) and long range (77 GHZ) radar modules are being used for adas systems including the object tracking, hazard detection and adaptive cruise control.

Comparision Chart Between Camera Lidar and Radar

• Lidar can provide good resolution about the position but can suffer from accuracy in poor weather.
• Unlike Lidar Sensor, radar can provide information about the velocity and bearing of the object.
• There is doppler lidar from which we can also obtain the velocity information.
• Laser data is computationally intensive because laser sends lot of data about each individual laser point.

What is the bearing angle ??

• It is nothing but the horizontal angle between direction of one object and another object (or) between them.
• Bearing to target is determined by direction in which the reflected echo returns.
• Bearing is general because it is defined as the angle separation between points.

• Radar can perform in bad environment conditions where Lidar cannot.
• Lidar cannot deliver accurate images of surroundings in fog, snow or dust.
• Lidar is not good in detection of the speed of the other vehicles. To make matter worse spinning lidar has lot of moving parts.
• Now Google’s lidar costs $7500 per unit. Comparing with Radar, Lidar is more expensive.
• A spinning lidar is better at detecting the objects. It provides the good range and angular resolution.
• But the Lidar’s range is not great as Radar which can see more than 200 meters.
• Lidar can’t also measure the radial velocity which radar sensor can.
• The wavelength of radar is 30 cm and 3 mm while lidar has micro-meter range wavelength (903 nm and 905 nm).
• Radar is less angularly accurate than lidar. It may get confused if multiple objects are placed very close to each other.
• Radar may consider two small cars in the vicinity as a large vehicle.
• Radar can determine relative speed or velocity using the doppler frequency shift.

Radar gives range but not good range resolution compared to the Lidar.

What is difference between the range and range resolution ??

• Range is about the calculation of what distance object is located.
• Range Resolution is about the ability to distinguish two or more objects on the same bearing angle at different ranges.

• In the above image the range resolution for a pulse with width 1 micro second. If the spacing between the aircrafts is small. Radar these two objects as a single object.

• In the above image the spacing between the two objects is more. In this case radar sees it as the 2 different objects.