After finishing his doctorate at MIT, Leaf Jiang worked for the military designing laser range devices for more than a decade. When developing future autonomous vehicles, Leaf found that laser-based detection technologies were too expensive to use, which is where NoDar comes in.
The data collected by a Light Detection And Ranging (LiDAR) system is used to build 3D images by scanning the environment with laser beams. Companies always want to deliver a seamless driving experience for their customers, so they rely on LiDAR technologies for mapping roads, detenting pathways, and helping cars make key judgements like whether an object is a person or a tree branch. Several different technologies have been used to create the costly LiDAR system. The system is powerful and could cost tens of thousands of dollars, but it is not foolproof and must deliver 100% accuracy or its value is lost. However, the digital camera that Leaf’s business NoDar promises to give is a significantly cheaper alternative based on one of the most extensively used technological equipment.
LiDAR: Why It’s Better Than A Camera.
A camera is an indispensable device for seeing suspicious activity and other crucial details. Attempts to achieve 3D vision with cameras have repeatedly failed. A camera-based system, in contrast to LiDAR, requires external light in order to create images. These cutting-edge imaging methods yield varying outcomes at different times of day, with low and poor-quality photos typically being produced while it’s raining or foggy outside.
Camera technology, according to NoDAR, has improved throughout the years. It has also developed unique algorithms to improve camera performance and guarantee that it can outperform LiDAR devices. NoDAR uses a pair of spaced-apart cameras mounted on a vehicle to gather several viewpoints and alternate realities of the road ahead. Using the two points of view, the distance and location of an item can be triangulated with greater accuracy. In earlier iterations of this method, precise calibration was essential for accuracy.
The cameras in Leaf Jiang’s startup are automatically calibrated, and their frames are synchronized so that the user experience is optimized. The method used for this calibration is patented by the company. As a result, you won’t need any supplementary equipment. In addition, the algorithm can be executed in real time on the chips already present in vehicles.
How quickly does it operate?
The company wanted to test the efficacy and functionality of the device in a dark environment, so they chose a secluded airport in Maine. Two 5-megapixel cameras placed around 1.2 meters (4 feet) distant were compared with data from a cutting-edge LiDAR system. An IEEE Spectrum article states that NoDAR has found that, in comparison to 600,000 LiDAR, its system is capable of producing 40 million data points per second during daylight hours. The group has also made progress on a car simulation chamber that, if fixed, would allow them to test in situations like fog and rain.
The amount of data points for the LiDAR system used to be 60%, but drops to 30% when heavy rain falls on cars. When visibility was just about 145 feet (45 m), the camera-based system nevertheless managed to provide accurate readings for 70% of the distance. On the other hand, just 20% of measurements taken with a LiDAR were reliable.
The company in question intended to conduct the “real” test at night. When compared to LiDAR systems, Nodar was able to detect a piece of timber approximately five inches (12 cm) in diameter from a distance of over 400 feet (130 m). Only at a distance of 164 feet (50 m) was it detected by the sophisticated LiDAR. Leaf is positive that it can create a NoDAR system at the same price as a LiDAR system. Experts on the technique claim that unlike NoDAR, which only operates in one direction, LiDAR can see all around any vehicle. More cameras and processing power are needed if a comprehensive analysis from every aspect is required, which could drive up the price of the system.