Donnerstags, 16:00: IP_Farmrobot

https://dev.azure.com/farmrobot/

Protokolle

• video
• Foto1
• Foto1
• Foto3

(Bild Systemdiagram)

Bill of Materials

Der Farmrobt verfügt über zwei 350W Gleichstrom-Motoren.
Kettenantrieb
Bereifung
Drehmoment
Geländetauglichkeit
Motortreiber
Einstellungen in Pixhawk

• Details ↪ Motorisierung

↪ Batteriemanagement

↪ Manuelle Steuerung mit Fernbedienung

↪ Steuerungs-Kamera

↪ Objektverfolgung

↪ NVIDIA Jetson TX1

YOLO (“you only look once”) - Object detection framework

YOLO can be set up as a pre-trained algorithm or be trained with custom data. Since the pre-trained neural net does not differentiate plants and has problems recognizing them in general, our net must be trained with custom image data.

One of the first steps in working with YOLO is choosing the version. Since custom training seems relatively simple, the main issue will be coordinate extraction, so the version will be chosen based on which one offers the best functionality with this.

V3 coordinate extraction:\ Edit darknet/src/image.c file:\

if(bot > im.h-1) bot = im.h-1;
 
printf("Bounding Box: Left=%d, Top=%d, Right=%d, Bottom=%d\n", left, top, right, bot); 
 
draw_box_width(im, left, top, right, bot, width, red, green, blue);

1. After editing you must “make” your darknet again
2. Running the algorithm now will output the coordinates of the bounding box in console

↪ Missions-Planer

↪ Autonomes Fahren mit GPS
↪ Hinderniserkennung mit Sonar

Delta Arm
Sprüh Mechanismus
Realcode G-code
↪ Unkraut behandeln

↪ Protokollierung

Erster Kamera Test
https://drive.google.com/drive/u/0/folders/1jvYJYdXXpjlxoEx-OkK9A2ChfQgQ70e3