Implemented using python version 3.7.2
This is a fairly simple image sending server. Images are captured using an OpenCV cv2.VideoCapture() object. The default selected device has the id 0. The UDP server simply jpeg-encodes the image and splits it in image_memory_size / max_packet_size chunks. The size of each chunk is determined by the set maximum packet size + 5. Each transmitted chunk contains 5bytes of meta info preceeding the actual image byte array.
| Chunk Type | Image Id | Chunk Id | Imagedata |
|---|---|---|---|
| S | 22 | 00 | ..... |
| C | 22 | 01 | ..... |
| E | 22 | 02 | ..... |
Where S marks the beginning of an image, C a central part, E the end of an image. For each image there is a random two digit Id created. Each Chunk is numerated. The last 4 bytes of information is used by the receiving side to reconstruct an image.
The repo contains two executable files.
The main.py file starts the server and immediately captures images and sends them to the specified ip:port.
The capture.py program basically opens a live image stream where the capture area is specified with four points by simply clicking on the image starting with the top-left corner in clockwise direction.
future==0.17.1 numpy==1.16.2 opencv-python==4.1.0.25
Either install requirements manually or by pip calling the requirements.txt file.
pip install -r requirements.txt
There is a set of parameters that are necessary:
Specify a json file where a capturing area is specified.
Must meet the following structure:
Specify a json file where a camera config is specified.
Must meet the following structure:
{
"width": 640,
"height": 360,
"fps": 60
}
Specify the receivers ip where the image chunks are sent to.
Specify the receivers port.
Specify the maximum udp packet size.
Specify the target sending fps.