VEYE CS Camera for Jetson TX2

From wiki_veye
Jump to navigation Jump to search

查看中文

How to use VEYE and CS series camera module on NVIDIA Jetson Nano,TX2,AGX Xavier and Xavier NX

1 Overview

This guide shows how to use VEYE and CS series camera modules on Jetson Nano,Jetson Nano 2GB,TX2,TX2 NX,AGX Xavier and Xavier NX. Jetson SDK Version is

  • Jetpack4.2.2,L4T r32.2.1
  • Jetpack4.3,L4T r32.3.1
  • Jetpack4.4,L4T r32.4.3
  • Jetpack4.4.1,L4T r32.4.4
  • Jetpack4.5,L4T r32.5
  • Jetpack4.5.1,L4T r32.5.1
  • Jetpack4.6,L4T r32.6.1

We provide two ways to do so: Prebuilt Binaries or Source Code. Yes, It's Open Source!

VEYE and CS series camera modules are camera module with ISP functions build in. It output UYVY data using MIPI-CSI2. We provide V4L2 interface for video streaming apps , and Video Control Toolkits (which is Shell Script) to control the camera module directly, which is called DRA(Directly Register Access).

1.1 Camera module list

Camera module dirver status list
Series Model Status
VEYE series VEYE-MIPI-IMX327S Done
VEYE series VEYE-MIPI-IMX385 Done
VEYE series VEYE-MIPI-IMX462 Done
VEYE series VEYE-MIPI-IMX335 Done
CS series CS-MIPI-IMX307 Done
CS series CS-MIPI-SC132 Done
TX2-XAVIER-nCAM Series CS-TX2-XAVIER-nCAM Done

1.2 How to check the current L4T version

On Jetson board

cat /etc/nv_tegra_release

If it shows:

# R32 (release), REVISION: 4.3......

It means L4t Version is 32.4.3

2 Hardware Setup

We have designed a 6cam interposer board for TX2 devkit and Xavier, It support up to 6 cameras at the same time.

6cam interposer board connected with 6 cameras
6cam interposer board connected with 6 cameras


In particular, the VEYE-MIPI-IMX335 must use a 12V DC terminal to provide auxiliary power due to its high power consumption. Refer to J11 of ADP-N1.

2.1 NVIDIA TX2 Developer Kit

It is connected as shown in figure:

TX2 Devkit 6 cameras connection
TX2 Devkit 6 cameras connection


Be sure to screw the ADP adapter board to the bottom board.

2.2 NVIDIA TX2 AGX Xavier


Connect VEYE cameras to AGX Xavier
Connect VEYE cameras to AGX Xavier


Be sure to screw the ADP adapter board to the bottom board.

2.3 NVIDIA Nano ,TX2 NX and Xavier NX Devkit

The Xavier NX interface is compatible with raspberry pi and Jetson Nano. The connection mode is shown in figure:

Xavier NX connection with VEYE-MIPI-327
Xavier NX connection with VEYE-MIPI-327


Xavier NX connection with CS-MIPI-IMX307
Xavier NX connection with CS-MIPI-IMX307
Jetson Nano A02 与VEYE模组连接
VEYE327 connection with Jetson Nano A02

3 Upgrade Jetson Nano, Jetson TX2, TX2 NX,AGX Xavier and Xavier NX system

3.1 Overview

This section describes how to upgrade the Jetson system to support our camera module. To support our camera module, we need to update the two parts of the L4T (Linux for Tegra) of the Jetson system, Image and DTB.

In the Image, we added the camera driver, while the DTB indicates the camera model used. In general, you only need to use the Image and DTB, that we have prebuilt, and you don't need to build from source when it is not necessary.

3.2 BSP package introduction

3.2.1 Path

https://github.com/veyeimaging/nvidia_jetson_veye_bsp

3.2.2 bsp package includes:
  • prebuild linux kernel:Image
  • prebuild dtb for different boards
  • driver source code
  • dts source code
  • i2c toolkits

3.3 DTB file name and Jetson Board

3.3.1 Nano A02

tegra210-p3448-0000-p3449-0000-a02.dtb

3.3.2 Nano B01

tegra210-p3448-0000-p3449-0000-b00.dtb

3.3.3 Nano 2GB

tegra210-p3448-0003-p3542-0000.dtb

3.3.4 TX2 Devkit

tegra186-quill-p3310-1000-c03-00-base.dtb

3.3.5 TX2 NX(p3509-0000 carrier board)

use the same carrier board with XAVIER NX Devkit

tegra186-p3636-0001-p3509-0000-a01.dtb

3.3.6 AGX XAVIER

tegra194-p2888-0001-p2822-0000.dtb

3.3.7 XAVIER NX Devkit

tegra194-p3668-all-p3509-0000.dtb

3.4 DTB file name and camera module correspondence

Under the dtb directory of each motherboard, the directory corresponding to the camera model is stored, and you can usually know the corresponding camera model directly by the directory name.

Specially:

1, VEYE-MIPI-CAM2M is the new version of VEYE series dtb, which supports all VEYE 2M series products, including VEYE-MIPI-327E, VEYE-MIPI-IMX327S, VEYE-MIPI-IMX462, VEYE-MIPI-IMX385.

2, The ones with fpdlink as suffix correspond to fpdlink coaxial model module.

3.5 Upgrade Jetson Nano and Xavier NX 、AGX Xavier without HOST PC

Note: AGX Xavie Jetpack version >= 4.4 , support this upgrade mode. Earlier versions have not been verified and are not recommended.

3.5.1 Check the current L4T version

First check the current L4T version and try to replace it with the same version.

cat /etc/nv_tegra_release

If it shows:

# R32 (release), REVISION: 4.3......

It means L4t Verion is 32.4.3

3.5.2 Download the bsp package to the Jetson board

On Jetson Board:

wget https://github.com/veyeimaging/nvidia_jetson_veye_bsp/releases/latest/download/nvidia_jetson_veye_bsp.tgz

tar -xzvf nvidia_jetson_veye_bsp.tgz

3.5.3 Upgrade Image

Backup old version Image

cp /boot/Image /boot/Image.backup

cp /boot/Image.sig /boot/Image.sig.backup

Under the bsp package directory, find the corresponding version of the Image package, decompress it, and execute:

sudo cp <path to your Image dir>/Image /boot/Image -f

For Xavier NX Devkit

Need to use the signed file(For example,file inside Image_l4t_r32.*.*_veyecam_signed_4_xavier),copy Image and Image.sig to /boot/ dir on the Jetson Xavier board:

sudo cp <path to your Image dir>/Image /boot/Image -f

sudo cp <path to your Image dir>/Image.sig /boot/Image.sig -f

Effective after restart.

3.5.4 Upgrade dtb

Under the bsp package directory, find the dtb file corresponding to your camera module model and L4T version number.

sudo cp <path to your dtb dir>/<DTB file name> /boot/ -f

Backup extlinux.conf file first

cp /boot/extlinux/extlinux.conf / boot/extlinux/extlinux.conf .back

Then, edit the /boot/extlinux/extlinux.conf file and add the following line.

FDT /boot/<DTB file name>

Note: <DTB file name> should be replaced with the corresponding dtb name of the mainboard, ensure that the file specified in the FDT line actually exists correctly.

Effective after restart.

3.6 Upgrade Jetson Board using Host PC

3.6.1 Burn the standard Image

Note: if your Jetson board has correctly burned the standard version Image, you can skip this step.

You can burn the standard Image using SDK Manager or following the steps below.

  • Set PATH

Assume SDK install directory is <TOPDIR>.

export TOP_DIR=<absolute path to top dir accroding to your jetpack version and board name;for example /home/xumm/nvidia/nvidia_sdk/JetPack_4.5_Linux_JETSON_XAVIER_NX_DEVKIT/>

export L4T_DIR=$TOP_DIR/Linux_for_Tegra

  • Build default image, flashing to Jetson Board, prepare L4T environment

cd $L4T_DIR

sudo ./apply_binaries.sh  

- Nano

sudo ./flash.sh jetson-qspi-sd mmcblk0p1

sudo ./flash.sh jetson-nano-devkit mmcblk0p1

sudo ./flash.sh jetson-nano-2gb-devkit mmcblk0p1

- TX2

sudo ./flash.sh jetson-tx2 mmcblk0p1

- TX2 NX(p3509-0000 carrier board)

sudo ./flash.sh jetson-xavier-nx-devkit-tx2-nx mmcblk0p1

- AGX Xavier

sudo ./flash.sh jetson-xavier mmcblk0p1

- Xavier NX Devkit

sudo ./flash.sh jetson-xavier-nx-devkit mmcblk0p1

3.6.2 Download our bsp package for Jetson

cd $L4T_DIR

wget https://github.com/veyeimaging/nvidia_jetson_veye_bsp/releases/latest/download/nvidia_jetson_veye_bsp.tgz

tar -xzvf nvidia_jetson_veye_bsp.tgz

export RELEASE_PACK_DIR=$L4T_DIR/nvidia_jetson_veye_bsp

3.6.3 Upgrade Image and DTB

prebuilt Image on github:https://github.com/veyeimaging/nvidia_jetson_veye_bsp/tree/master/kernel_image

  • Installing the Kernel

Whether you use a prebuild kernel or you build it yourself

- Nano and TX2

copy Image to /boot/ dir on the Jetson Nano or TX2 board:

sudo cp <path to your Image dir>/Image /boot/Image -f

- AGX Xavier and Xavier NX Devkit

Need to use the signed file(For example,file inside Image_l4t_r32.*.*_veyecam_signed_4_xavier),copy Image and Image.sig to /boot/ dir on the Jetson Xavier board:

sudo cp <path to your Image dir>/Image /boot/Image -f

sudo cp <path to your Image dir>/Image.sig /boot/Image -f

on the Host PC:

cd $L4T_DIR

sudo ./flash.sh -k kernel jetson-xavier mmcblk0p1

- Xavier NX

on the Host PC:

cd $L4T_DIR

sudo ./flash.sh -k kernel jetson-xavier-nx-devkit mmcblk0p1

  • Flashing dtb file

Enter the recovery mode,and on the Host PC:

cp <path to your dtb dir>/dtbfilename $L4T_DIR/kernel/dtb/ -f

cd $L4T_DIR

- Nano

sudo ./flash.sh -r -k DTB jetson-nano-qspi-sd mmcblk0p1

sudo ./flash.sh -r -k DTB jetson-nano-devkit mmcblk0p1

sudo ./flash.sh -r -k DTB jetson-nano-2gb-devkit mmcblk0p1

- TX2

sudo ./flash.sh -r -k kernel-dtb jetson-tx2 mmcblk0p1

- TX2 NX(p3509-0000 carrier board)

sudo ./flash.sh -r -k kernel-dtb jetson-xavier-nx-devkit-tx2-nx mmcblk0p1

- AGX Xavier

sudo ./flash.sh -r -k kernel-dtb jetson-xavier mmcblk0p1

- Xavier NX

sudo ./flash.sh -r -k kernel-dtb jetson-xavier-nx-devkit mmcblk0p1

Reset Jetson Board after flashing.

3.7 Are Image and DTB upgrade successful ?

3.7.1 For Image

ls /sys/bus/i2c/drivers/

There should have veye327 veyecam csx307 cssc132 directory.

3.7.2 For DTB

DTB varies from platform and from camera module.

- Nano A02/ Nano 2G

ls /proc/device-tree/host1x/i2c@546c0000/

- Nano B01

ls /proc/device-tree/cam_i2cmux/i2c@*

- TX2 Devkit

ls /proc/device-tree/i2c@3180000/tca9548@70/i2c@*

or

ls /proc/device-tree/i2c@3180000/tca9548@77/i2c@*

- AGX Xavier

ls /proc/device-tree/i2c@3180000/tca9548@70/i2c@*

or

ls /proc/device-tree/i2c@3180000/tca9548@77/i2c@*

- Xavier NX and TX2 NX(p3509-0000 carrier board)

ls /proc/device-tree/cam_i2cmux/i2c@*

You should be able to find the camera model you used.

3.8 Protection against being overwritten by upgrades

After making sure Image and dtb are successfully updated, please execute the following command to prevent kernel and dtb from being overwritten by update.

sudo apt-mark hold nvidia-l4t-kernel nvidia-l4t-kernel-dtbs

3.9 Build the drivers from source

Please refer to: Nvidia Jetson platform VEYE and CS series camera driver source code compilation guide.

Could be used for Flashing or Upgrading.

4 Applications and Test

4.1 Check system status

Run the following command to confirm whether the camera is probed.

  • For VEYE-MIPI-327(using VEYE-MIPI-327 dtb)

dmesg | grep veye327  

The output message appears as shown below:

Detected VEYE327 sensor

subdev veye327 X-003b bound

  • VEYE-MIPI-IMX327S/VEYE-MIPI-IMX462/VEYE-MIPI-IMX385(using VEYE-MIPI-CAM2M dtb)

dmesg | grep veye  

The output message appears as shown below:

camera id is veyecam

sensor is IMX327/IMX462/IMX385

subdev veyecam X-003b bound

  • For CS-MIPI-IMX307

dmesg | grep csx307  

The output message appears as shown below.:

Detected CS307 sensor

subdev csx307 X-003b bound

  • For CS-MIPI-SC132

dmesg | grep cssc132  

The output message appears as shown below.:

subdev cssc132 X-003b bound

  • For VEYE-MIPI-IMX335

dmesg | grep imx335

The output message appears as shown below.:

camera id is VEYE-MIPI-IMX335

subdev veye_imx335 10-003b bound

  • Run the following command to check the presence of video node.

ls /dev/video*

The output message appears as shown below.

videoX  

4.2 Video Stream Toolkits Manual

Our camera appears as /dev/videoX device node,where X is between [0-5].

Our camera module, mapped as /dev/videoX device node in the Jetson system.

During the OS boot process, the cameras are detected in the order of i2c bus from smallest to largest. For example, if only one camera is connected, X is 0 regardless of the hardware connection to the location. if 5 cameras are connected, X is reflected as [0-4] according to i2c bus from smallest to largest.

in gstreamer command ,v4l2src has a param device=/dev/videoX to identify which camera to use.

4.2.1 Gstreamer Usage
  • Preview FHD

gst-launch-1.0 nvv4l2camerasrc device=/dev/video0 ! "video/x-raw(memory:NVMM),format=(string)UYVY, width=(int)1920, height=(int)1080" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420" ! nvoverlaysink sync=false

gst-launch-1.0 nvv4l2camerasrc device=/dev/video0 ! "video/x-raw(memory:NVMM),format=(string)UYVY, width=(int)1920, height=(int)1080" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)NV12" ! nvoverlaysink sync=false

  • Preview FHD(using xvimagesink sink)

export DISPLAY=:0

gst-launch-1.0 -e v4l2src io-mode=4 device=/dev/video0 do-timestamp=true ! 'video/x-raw, width=1920, height=1080, framerate=30/1, format=UYVY' ! xvimagesink sync=false

  • Preview 720p@60 (CS-MIPI-IMX307)

gst-launch-1.0 nvv4l2camerasrc ! "video/x-raw(memory:NVMM),format=(string)UYVY, width=(int)1280, height=(int)720, framerate=(fraction)60/1" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420" ! nvoverlaysink sync=false

  • Preview 1280*1080@45 (CS-MIPI-SC132)

gst-launch-1.0 nvv4l2camerasrc ! "video/x-raw(memory:NVMM),format=(string)UYVY, width=(int)1280, height=(int)1080, framerate=(fraction)45/1" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420" ! nvoverlaysink sync=false

  • Preview 640*480p@130 (CS-MIPI-IMX307)

gst-launch-1.0 nvv4l2camerasrc ! "video/x-raw(memory:NVMM),format=(string)UYVY, width=(int)640, height=(int)480, framerate=(fraction)130/1" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420" ! nvoverlaysink sync=false

  • Preview 2 cameras 1080p HD

WIDTH=960

HEIGHT=540

CAPS="video/x-raw(memory:NVMM),format=(string)UYVY, width=1920, height=1080"

gst-launch-1.0 nvcompositor name=comp sink_0::xpos=0 sink_0::ypos=0 sink_0::width=$WIDTH sink_0::height=$HEIGHT sink_1::xpos=$WIDTH sink_1::ypos=0 sink_1::width=$WIDTH sink_1::height=$HEIGHT ! nvoverlaysink nvv4l2camerasrc device=/dev/video0 ! $CAPS ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420"! comp. nvv4l2camerasrc device=/dev/video1 ! $CAPS ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420"! comp.

  • Gstreamer cmd line for OpenCV

I think maybe OpenCV do not support I420 data format input,so you should convert it to BGR format.

gst-launch-1.0 nvv4l2camerasrc ! video/x-raw(memory:NVMM), format=(string)UYVY, width=(int)1920, height=(int)1080 ! nvvidconv ! video/x-raw(memory:NVMM), format=(string)I420 ! nvvidconv ! video/x-raw, format=(string)BGRx ! videoconvert ! video/x-raw, format=(string)BGR ! appsink

  • Record FHD in H.264 format to a video file

gst-launch-1.0 nvv4l2camerasrc num-buffers=300 ! "video/x-raw(memory:NVMM),format=(string)UYVY, width=(int)1920, height=(int)1080" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)NV12" ! nvv4l2h264enc control-rate=1 bitrate=10000000 ! h264parse ! qtmux ! filesink location=filename.mp4 -e

  • Playback of saved video file

gst-launch-1.0 filesrc location=filename.mp4 ! qtdemux ! queue ! h264parse ! nvv4l2decoder ! nvoverlaysink -e

  • Capturing FHD still image  

gst-launch-1.0 v4l2src num-buffers=1 ! "video/x-raw,format=(string)UYVY, width=(int)1920, height=(int)1080" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420" ! nvjpegenc ! filesink location=jpgname.jpg

4.2.2 VEYE-MIPI-IMX335 gstreamer example

Since VEYE-MIPI-IMX335 supports only a few specific resolution modes, its driver we use use_sensor_mode_id mode.

VEYE-MIPI-IMX335 mode
./veye5_mipi_i2c.sh video mode v4l2-ctl sensor_mode video format
1 0 2592x1944@20fps
2 1 2592x1944@12.5fps
3 2 2560x1440@25fps
4 3 2560x1440@30fps
  • prepare

sudo apt-get install v4l-utils

Refer to the following section and download the veye5_mipi_i2c.sh tool.

  • 2592x1944@20fps mode preview, framerate statistics

v4l2-ctl -d /dev/video0 --set-ctrl sensor_mode=0

./veye5_mipi_i2c.sh -w -f videomode -p1 1

v4l2-ctl -d /dev/video0 --set-ctrl preferred_stride=5376

Preview

gst-launch-1.0 nvv4l2camerasrc device=/dev/video0 ! "video/x-raw(memory:NVMM), width=(int)2592, height=(int)1944" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420" ! nvoverlaysink sync=false

Framerate statistics

gst-launch-1.0 nvv4l2camerasrc device=/dev/video0 ! "video/x-raw(memory:NVMM),format=(string)UYVY,width=(int)2592,height=(int)1944" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420" ! fpsdisplaysink video-sink=fakesink -v

  • 2560x1440@30fps mode video encode and capture

v4l2-ctl -d /dev/video0 --set-ctrl sensor_mode=3

./veye5_mipi_i2c.sh -w -f videomode -p1 4

v4l2-ctl -d /dev/video0 --set-ctrl preferred_stride=5120

Video encode

gst-launch-1.0 nvv4l2camerasrc num-buffers=300 ! "video/x-raw(memory:NVMM),format=(string)UYVY, width=(int)2560, height=(int)1440" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)NV12" ! nvv4l2h264enc control-rate=0 bitrate=10000000 ! h264parse ! qtmux ! filesink location=filename.mp4 -e

Video Replay

gst-launch-1.0 filesrc location=filename.mp4 ! qtdemux ! queue ! h264parse ! nvv4l2decoder ! nvoverlaysink -e

Capture

gst-launch-1.0 v4l2src num-buffers=1 ! "video/x-raw,format=(string)UYVY, width=(int)2560, height=(int)1440" ! nvvidconv ! "video/x-raw(memory:NVMM),format=(string)I420" ! nvjpegenc ! filesink location=jpgname.jpg

4.3 Video Control Toolkits Manual

  • VEYE-MIPI-CAM2M Series(VEYE-MIPI-327、VEYE-MIPI-IMX327S、VEYE-MIPI-IMX462、VEYE-MIPI-IMX385)

Video Control Toolkits Manual :VEYE-MIPI-327 I2C

  • CS-MIPI-IMX307和CS-MIPI-SC132

Video Control Toolkits Manual :CS-MIPI-X I2C

  • VEYE-MIPI-IMX335

Video Control Toolkits Manual :VEYE-MIPI-IMX335 I2C

5 Realtimes RTSO-9001

5.1 Hardware Setup

CS-MIPI-IMX307 and RTSO-9001(TX2)


Matching with Realtimes RTSO-9001, you need to use RTST-RPI adapter board.

Note: the power supply capacity of the MIPI CSI-2 interface of the current version of RTSO-9001 is not enough to support the power requirements of our camera, which requires an additional 5V power supply to the camera. It is not shown in the picture above.

5.2 Software bsp

  • The corresponding directory for Realtimes RTSO-9001, is:

nvidia_jetson_veye_bsp\TX2\JetPack_4.2.2_Linux_RTSO-9001

  • Subject to the agreement with Ruitai, we do not release the source code
  • Our code have been submitted to Realtimes, and customers are advised to contact Realtimes for a complete version of the supported image.

6 Document History

  • 20220110

Add VEYE-MIPI-IMX385 support.

  • 20220105

Use nvv4l2h264enc instead of omxh264enc in gstreamer command,because omxh264enc has been deprecated.

  • 20211025

Use nvcamerasrc instead of v4l2src in gstreamer command , and the data goes directly into DMA memory.

7 References

Jetson start up: https://developer.nvidia.com/embedded/learn/getting-started-jetson

Jetson Download Center: https://developer.nvidia.com/embedded/downloads

Xavier NX: https://developer.nvidia.com/embedded/learn/get-started-jetson-xavier-nx-devkit

TX2 development kit: https://developer.nvidia.com/embedded/jetson-tx2-developer-kit

SDK Manager: https://docs.nvidia.com/sdk-manager/index.html

L4T Doc: https://docs.nvidia.com/jetson/archives/l4t-archived/l4t-3242/index.htm

TX2 user guide: link

Realtimes website:http://www.realtimes.cn/