camera_stream

Raw video streaming for the Arducam 1200° wide-angle camera (5120×800) running on a Jetson Nano B01, streamed over a direct Ethernet cable to a receiver.

---

Overview

The system has two scripts:

Script	Runs on	Role
stream_raw.py	Sender	Captures, processes, and sends frames
receive_raw.py	Receiver	Receives, displays, and optionally saves frames

Frames are sent as raw BGR pixel data over a TCP socket — no video encoding.

---

Hardware Setup

• Arducam wide-angle camera CamArray HAT connected to the Jetson Nano via CSI
• Sender and Receiver connected directly via an Ethernet cable (no router/switch needed)

---

Network Configuration

The Sender connects outbound to a fixed IP. Set a static IP on the receiver to match.

Receiver static IP (required):

Field	Value
IP Address	192.168.1.100
Subnet Mask	255.255.255.0

Windows: Settings → Network & Internet → Ethernet → Edit IP assignment → Manual

Linux:

sudo ip addr add 192.168.1.100/24 dev eth0
sudo ip link set eth0 up

The Sender does not need a specific IP — only the receiver does.

---

Installation

Sender

pip3 install opencv-python numpy

The ArducamUtils dependency comes from Arducam's SDK. Follow their setup guide for your specific camera module and ensure utils.py is in the same directory as stream_raw.py.

Receiver

pip install opencv-python numpy

---

Usage

Can be launched in any order. Both sides will wait and automatically reconnect if the connection drops.

1. Start the receiver

python3 receive_raw.py

The receiver will listen on port 5000 and wait for the Sender to connect.

2. Start the Sender

python3 stream_raw.py

The Sender will wait for a frame from the camera, then connect to the receiver and begin sending.

Receiver controls (while stream window is open)

Key	Action
s	Save current frame as uncompressed PNG to captures_raw/
q	Quit

---

Wire Protocol

Each frame is sent as a binary packet:

Bytes 0–7   : Capture timestamp  (float64, big-endian)
Bytes 8–11  : Frame data size    (uint32, big-endian)
Bytes 12+   : Raw BGR pixel data (uint8, row-major)

Frame resolution after downscale: 2560 × 400 px, 3 channels = 3,072,000 bytes per frame.

---

Configuration

All tuneable constants are at the top of each script.

stream_raw.py

Constant	Default	Description
HOST	192.168.1.100	Receiver IP address
PORT	5000	TCP port
TARGET_FPS	30	Sending rate cap

receive_raw.py

Constant	Default	Description
PORT	5000	TCP port to listen on
CAM_W	2560	Expected frame width
CAM_H	400	Expected frame height
SAVE_DIR	captures_raw	Directory for saved PNGs

---

Reconnection Behaviour

Both sides handle disconnects without needing a restart:

• Sender: If the connection is refused or drops mid-stream, it closes the socket and retries every 2 seconds until the receiver is available again.
• Receiver: If the Sender disconnects, it closes the connection and immediately starts listening for a new one. The display window stays open while waiting.

The camera capture thread on the Jetson runs independently and is never interrupted by network events.

---

Image Processing (Sender-side)

Before sending, each frame is processed on the Jetson Nano:

1. Captured at native 5120 × 800
2. Converted from raw sensor format to BGR via ArducamUtils
3. Brightness/contrast boost — alpha=1.3, beta=30
4. Per-channel colour correction:
   • Blue ×1.2
   • Green ×0.95
   • Red ×0.90
5. Downscaled to 2560 × 400 via cv2.resize

---

Troubleshooting

Sender says "Receiver not ready, retrying..." indefinitely

• Check the static IP is set correctly on the Receiver (192.168.1.100)
• Make sure receive_raw.py is running before or shortly after stream_raw.py
• Check no firewall is blocking port 5000 on the Receiver

Receiver shows no window / black frame

• Confirm CAM_W and CAM_H in receive_raw.py match the resolution being sent by the Sender (default: 2560×400)
• mismatch will cause reshape() to throw an error

High latency

• Use ethtool to confirm the link is running at 1 Gbps: ethtool eth0 | grep Speed

Camera not found on Jetson

• Verify the camera is detected: ls /dev/video*
• Check Arducam driver is installed and utils.py is present
• Check Cable connection (Usually the issue)