Tech talks #3 What is the technology stack behind Google Street View? What's in the cars? How is power and cooling managed? What happens when the car stops? What cameras are used? How is data stored? When and how is it uploaded to a server?

Hello guys how are you and this Tech talk I will tell you about how Google maps street works?

What's on the cars?

The most visible part of the equipment is of course what’s mounted on the roof of the Street View vehicles, and that is the data recording equipment. This setup includes 15 cameras and three laser scanners securely fixed at a height of 8.2 feet.

The laser scanners are used to record the actual dimensions of the space being photographed. How quickly the three lasers reflect off surfaces tells the software how far a building or object is, and allows Google to construct 3D models.

The cameras themselves contain no mechanical parts—not even a shutter—and instead use CMOS sensors and an electronic rolling shutter.

The cameras are hooked up to a monitor inside the vehicle, and data is recorded to SSD drives. Vehicles are equipped with 4G Internet.

GoogleDriver419, who fielded questions on Reddit, said recording was simple. “I push record and the cameras do all of the work.”

How much storage? How is data stored? When and how is it uploaded to a server?

The vehicle is stocked with standard SSD drives. According to GoogleDriver419, there is typically around 50-100 TB in the car, good enough for at least a week’s worth of recording. Each drive that is recording is automatically backed up to a second SSD to allow for drive failures.

A typical city block requires about 2-3 GB on average. A day’s work might consume about 50-100GB of space for smaller areas, but 800GB or more in say New York City.

The data is collected by drivers who fill up hard drives and ship them back to Google. The drivers won't send them in until they have five disks completely full.

What happens when the car stops?

I’m not sure if this question refers to the car stopping for the evening, or stalling with a mechanical issue.

As for what happens at the end of the day: Drivers will pack the equipment from the roof and put it in the trunk. They are generally put up at hotels when on longer trips and will park the car in the lot. In some cities like New York, Google has a garage for the vehicle. Drivers aren’t allowed to take the car home without permission.

If there is a vehicle or equipment breakdown, according to GoogleDriver419, they are generally never too far from an office where they have spares. Otherwise someone is sent out to fix it and filming is stopped for the day.

How fast can the car go? What's the bottleneck for data throughput?

The current cameras can handle all speed limits in the US. However, if the drivers can go slower without holding up traffic, they will. The slower the car, the better the quality of the images.

Bad weather can affect whether the vehicle is driven out. The car is parked when there is rain or high winds as it can affect both the cameras as well as the quality of the images.

What cameras are used?

The cameras are built in-house by Google.

There have been four generations of cameras so far in the Street View program. The earliest version of the camera array (titled R2) used a ring of eight 11-megapixel CCD sensors with commercial photographic wide-angle lenses; a subsequent version (R5) had a ring of eight 5MP CMOS sensors with custom low-flare lenses, plus a fisheye lens on top to capture upper levels of buildings.

The current setup, the R7, uses 15 of the same sensors and lenses as the R5, but no fish-eye. The latest generation is now being used around the world, and takes photographs that are “near-HD” in quality.

How does the driver know where to go/not go?

Google decides which areas to map based on the frequency and volume of online searches.

According to GoogleDriver419, drivers are given the starting point and finishing point, as well as key points that need to be mapped. Plenty of info is provided to the drivers but it’s up to them to plan the route they will take.

Sometimes smaller streets are missed out for this reason, and may be added on a subsequent trip. When this happens, software helps to blend the old and new images and also tries to smooth out the differences in light levels.

What happens to the imagery before it's made live?

Street View drivers try to be aware at all times about their surroundings and whether the cameras might pick up something they shouldn’t, e.g. a traffic accident or people mooning them.

“If I see something that I know shouldn't be on camera, I know that the camera picked it up as well, so I will need to delete it and go back and try again later,” said GoogleDriver419.

Drivers are given special editing software and go over all the images they have taken. Each day of driving typically needs about two days of editing. GPS data is automatically embedded in each photo.

The disks get shipped to a data centre, the information is uploaded and then everything gets fed into a core database and goes through a few processing steps. This includes blurring of faces and license plates.

Sometimes the blurring is not quite accurate, though, as many pictures on StreetViewFun reveal.

There are 15 images taken for each finished shot and angle that you see on Street View today, and Google’s software takes all of these images and mashes them together, adjusts the exposure for sun, shadows, colour differences and brightness.

It usually takes six months after a Street View driver uploads the files before it goes live.

Here I full look at it.