This repo contains the software components for our Barcode scanning solution implemented with in scope of the Tech-Challenge S22.
As we our task was to come up with an innovative solution to tackle robotizing waste recycling. Our team developed the idea which aims to ultimately create a zero waste stream, this can be done by tracking waste from waste to source. Our prototype aims to showcase the potential of tracking recycable waste present in the collected through computer vision. For that we Built a prototype which consists of a simple conveyer belt, run by a stepper motor. This is combined with a two cameras mounted on the on top of the conveyer belt looking down. Our Software which run on a dedicated Raspberry Pi, processes the camera stream, detects and decodes the barcode present on the products, and finally visualizes the output along with information from our database. Here is a detailed explanation of the Implementation.
For Building the prototype, the following electronics were used:
- Raspberry Pi Model 4B 2Gb+
- Raspberry Pi Model 2B
- Raspberry Pi Cam V1.3, documentation can be found here
- Microsoft lifecam cinema
- Adafruit DC & Stepper Motor HAT for Raspberry Pi - Mini Kit, documentation can be found here
- Stepper motor - NEMA17
- 2 x MicroSD cards
- 12 V DC power supply
- 5.1 V USB C power supply
- 5 V Micro-USB power supply
The parts used for building the conveyer belt were drawn on Inventor. These parts can be found under components directory. The prototype can be completey built using lastercut and 3D printed parts mounted together using bolts. Some components can be bought as well from building supplies stores. The complete assembly can be found here.
The following parts can be cut on 6mm HDF plates and mounted together:
- Sides of the conveyer Belt. Design can be found here.
- The Pi holder in which the Pi and cameras are mounted to. This consists of two parts, the sides and the holder.
- The two gears transfering the movement from the stepper motor to the belt. gear1, and gear2.
- In order to mount and fix the gears together, this part, which should be also laser cut.
- The Pi camera mounting holder, the design was adopted from here.
- The rollers can be printed, but since this can take long times on the 3D printer, pvc pipes from the store can be used instead.
- If PVC pipes are used for the rollers, this part should be printed to mount the pipes on the shafts (Welle).
The Barcode detection algorithm running on a Raspberry Pi Model 4B 2Gb+ consists of the following modules:
- The Scanner: This is the main module executing the barcode detection and decoding algorithm. The module contains several algorithms for processing image frames. The processing is mainly carried out using OpenCV python library, in addition to utilizing the Zbar linux library and it's python wrapper Pyzbar and Dynamsoft Barcode Reader SDK for decoding the detected barcodes.
- Several utility functions are implemented in this file.
- Two mutliple object tracker (MOT) algorithms were adopted to track the detected barcodes along the conveyer belt, to prevent multiple detection and decoding of the same object more than once, as well as assigning unique IDs for each detected object. The two trackers adopted are SORT tracker adopted from here, as well as Centroid adopted from here.
All other python packages needed can be found in the requirements Specific function description is to found in the python files for each function, in addition to the inputs/outputs and other comments.
First of all, a fresh OS image has to be installed on the micro SDcard for the Raspberry Pi Model 4B 2Gb+. A more detailed tutorial can be found here.
After booting up the Pi and connecting to a screen or connecting per SSH, open a terminal and clone the repo:
pi@raspberrypi:~$ git clone https://github.com/zizo1111/tech_challenge.gitIn the project's directory, create a python virtual enviroment venv and activate it:
pi@raspberrypi:~$ python3 -m venv tech_env
pi@raspberrypi:~$ source tech_env/bin/activateInstall the required packages listed in the requirements file:
(tech_env)pi@raspberrypi:~$ pip install -r requirements.txtThe scannner is then ready to be run, for this there are several options available that can be chosen by the user such as input source and other settings. For more info run:
(tech_env)pi@raspberrypi:~$ python3 scanner.py -h
usage: scanner.py [-h] [-s [{usb,picam,vid,img}]] [-p PATH] [--database_path DATABASE_PATH] [-d DSTDIR] [--width [WIDTH]] [--height [HEIGHT]] [--show_raw SHOW_RAW] [--show_res SHOW_RES] [--write_vid WRITE_VID]
[--tracker [{sort,centroid,none}]]
optional arguments:
-h, --help show this help message and exit
-s [{usb,picam,vid,img}], --src [{usb,picam,vid,img}]
select input source (default: img)
-p PATH, --path PATH video/image path
--database_path DATABASE_PATH
video/image path
-d DSTDIR, --dstDir DSTDIR
path to the destination
--width [WIDTH] camera frame width
--height [HEIGHT] camera frame height
--show_raw SHOW_RAW show raw frame
--show_res SHOW_RES show processed frame
--write_vid WRITE_VID
write_out video
--tracker [{sort,centroid,none}]
tracker (default: sort)Run the scanner on a already recorded video:
(tech_env)pi@raspberrypi:~$ python scanner.py -s vid -p PATH_TO_VIDRun the scanner using the usb cam:
(tech_env)pi@raspberrypi:~$ python scanner.py -s usbRun the scanner using the picam:
(tech_env)pi@raspberrypi:~$ python scanner.py -s picamThe output window shoulf look like this
On the left corner, the following information is displayed.
- Number of objects detected
- Current collected value in cents
- List of Objects detected and matched with our database, with their names, EAN number, material type and value/kg.
Once this step is done, I2C has to be enabled on the the PI. A more detailed tutorial can be found here.
Install the Adafruit CircuitPython MotorKit library along with the necessary dependencies.
pi@motorpi:~$ sudo pip3 install adafruit-circuitpython-motorkitDownload this python file
pi@motorpi:~$ wget https://raw.githubusercontent.com/zizo1111/tech_challenge/master/motor.pyFinally run the motor by:
pi@motorpi:~$ python motor.pyFor preparation print two rubber plugs in the 3d printer (you can find them in our repo), as it takes some hours. For the other material you can just follow the buy list and get most of the stuff in a regular hardware store.
First of all, the dimensions of the conveyer belt itself needs to be decided. We choose, to cut a 2.00m steel bar in two pieces (=1m length) and cut the rolls in 30cm pieces. With rolls every 16.67cm, we needed 6 rolls. The next step would be cutting everything as needed. Step 2: Prepare the rolls and insert them
For this step you need per cutted roll (not for the motor roll, that means 5/6): a lot of sandpaper, 2 bearings, 2 screws and 2 nuts (M8), 4 flat washers (Unterlegscheiben, inner diameter 8mm) and the roll. That meant we need 10 pieces from everything and the double amount of washers.
First increase the friction on the outside of the rolls (also the motor roll), by sticking on the sandpaper on each of them. You can cut the sandpaper in multiple strips (length= scope of the roll) and stick 3-4 of them on each roll with hot glue.
After everything is dryed, you need to insert the bearings into the roll (it should smuggle in nicely with small pressure, if it’s too loose, add some glue). Then the order is: bearing -> washer -> steel bar -> washer -> nut on both sides. Prepare each position and tighten the screws a little bit. Only if you have all rolls inserted, tight the screws hard.
Now the basic frame is set, span a suitable banner over the whole conveyer. It should be a little longer then needed. Make sure that you build up a tension and stick the beginning and the end with instant adhesive. Cut the too long remains.
For the motor roll, you need one metal rod (4mm), which should be longer then the width of the prototype. Push it through the rubber plug, the roll and the other rubber plugs. Fasten the two rubber plugs at the ends of the roll with hot glue. Then place the metal rod into two holes in the bar.
Here we will discuss the mechanics of the motor construction. Follow our github page for programming instructions. To power the conveyer belt, we use a stepper motor and a speed controller implemented in a raspberry pi. You need to cut the two gears from a stable flat material (we used wood). Fix the smaller gear to the motor and the bigger one to the motor roll. Then place them, so they rotate with each other (see photo).
Connect the motor to the power source and get it run with the raspberry pi. Now you’re finished and everything should work!
The whole construction og the connyever belt looks like this:
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent