[Article] Comparison and selection of RealSense D415 and D435 and D435i and T265

The popular 3D camera Intel RealSense D series now has three models: D415, D435 and D435i. In March 3, RealSense tracking camera T2019 was released.
This time I would like to introduce the differences between these models. We would be happy if you could refer to it when you are uncertain which model to choose.

* This article refers to the following WEB pages
Which Intel RealSense device is right for you
Choosing an Intel RealSense Depth Camera

First, the main differences in the specifications of the D415 and D435 as a 3D camera are as follows.
* D435i is exactly the same as D435 except that it has an inertial measurement unit (IMU).

Part 1: Viewing angle (FoV)

D415: approx. 65 °
D435: approx. 85 °

From a camera performance perspective, the wide viewing angle of the D435 means that the granularity value is small (that is, the resolution is low) when looking at a particular part of the image. The stereo camera uses two sensors to acquire two images from slightly different viewpoints and compare them to determine the depth.

This image is a magnified and exaggerated image captured by the sensor of each camera. Looking at the same size area in the image (boxed area), D415 has higher pixel density. This means that at the same point, D435 is more accurate than D415 for a particular area.

This does not make a big difference for applications such as obstacle avoidance and human detection, but for applications where accuracy is the highest priority, such as 3D scanning, the D415 can provide better and more accurate scanning. That's true (so to speak, the D415 is more than twice as accurate as the D435).

Part 2: Shutter type

D415: Rolling shutter
D435: Global shutter

The rolling shutter used in the D415 is a system that records all pixels in the scene by quickly scanning the target horizontally or vertically. These are typically done over a few frames, but each data is stored as a single frame.

* Image of rolling shutter

On the other hand, the global shutter used in the D435 is a method that snaps the entire scene into one frame, and differs from the rolling shutter in that each pixel is captured at the same time.

* Image of global shutter

Part 3: Differences regarding modules

D415: Uses Intel RealSense D415 module
D435: Uses Intel RealSense D430 module

The D415 has the advantage that the Imagers, Emitters and RGB cameras are on one computer board or stiffener, making calibration much easier.

On the other hand, in D435, the RGB camera is not on the module (the RGB camera is separately connected to the module). Since the depth algorithm is based on the exact placement of the sensor, it can be difficult to calibrate between depth and RGB without fixing it to the same stiffener.

However, the flexibility of being able to select an RGB camera is also a feature that users have long been asking for, and this is a merit depending on how the user uses it.

Part 4: Imager differences

D415: Uses a 2-megapixel Imager
D435: Uses a 1-megapixel Imager

Part 5: Difference of minimum Z

カメラとキャプチャ対象との距離(最小Z)について、D435はD415と比べて任意の範囲でより多くの深度ノイズ (>2倍) があり、同じ解像度での最小操作距離が小さくなる(約0.5倍)ため、カメラにより近づことができます。たとえば、848×480の解像度において、D415の最小Zはおよそ29cm、一方D435の最小Zはおよそ17cmとなります。

■ Comparison summary as a 3D camera

Each of the D415 and D435 (D435i) has its own characteristics, and it can be said that the best selection method is to use the D415 and DXNUMX (DXNUMXi) according to the user's application.

Recommended for such applications

D435:
Due to the wide viewing angle,Navigation and tracking applications for robots and drones Suitable for
The global shutter also provides excellent performance when shooting fast moving objects.

D415:
Accuracy is required for the acquired image dataApplications such as 3D scanning Suitable for

Next is an introduction to the D435i and T265.

The specifications of the D435i as a 3D camera are the same as the D435, but an inertial measurement unit (IMU) is additionally installed.
* The IMU makes it possible to measure the movement of the device itself in synchronization with the depth data.

The T265 also has this IMU, but it also uses two fish-eye cameras and a unique algorithm to track its position in space with great accuracy. Note that the T2 is not a 265D camera, so depth data cannot be obtained. However, you can combine the T3 with the D265 series 400D cameras to combine high precision tracking with high quality depth data.

It is also possible to use the D435i to run a similar algorithm for SLAM (in which case the D435i runs on the attached host device and the T265 algorithm runs on the camera itself). Depending on the application, this may provide sufficient tracking accuracy. However, if tracking accuracy is very important, or if the camera is expected to make large movements, such as spinning quickly, the T265 will perform better.

■ Comparison summary as a tracking camera

Each has such characteristics

D435i:
・ In addition to tracking information, depth information can be acquired simultaneously
・ Raw IMU dataConnected host deviceDeliver directly to

T265:
・ Suitable when more accurate tracking information is required
・ Depth information cannot be acquired because it is not a depth camera (*)
・ Wide field of view of 163 ± 5 ° is covered by the fisheye lens sensor.
・ IMU readingOwn Intel Movidius Myriad 2.0 VPURedirect to

* However, depth information can also be acquired by using it in combination with the D series (combination example: T265 + D435 or T265 + D415)

Note that both devices use the same open source Intel RealSense SDK 2.0, making it easy to write compatible code when switching to another device later. This is true for future devices, so the data you build will work in the future.

In Tegara's R & D overseas product procurement service "Unipos"RealSense Depth Camera D Series andRealSense Tracking Camera T265 Fool you.