masarycar

Detecting cars in bikelanes at Masarycka with Open CV.

Disclaimer

Using OpenCV with the provided cars recognition xml file has yielded very limited results. This program was never close to reliably detecting cars in the image, perharps due to low resolution of the webcam image (most cars would take up less than 20x20px on the image).

I have also explored the possibility of comparing the latest webcam image with a image with no cars and detecting the diff. This approach was very susceptible to reporting lighting changes of the street's surface due to weather/daytime as false positives. I have not yet discovered a reliable way to detect cars.

How to use

First, install the prerequisites

pip install -r requirements.txt

Next run the script:

python main.py  -f "src/sample01.jpeg" \
                -c "193,440,529,954" \
                -m "src/mask.png" \
                -u 5 \
                -a 1 \
                -d "src/cars.xml"

List of arguments

• -f --file Specify path to the file you want to analyze.
• -c --croprect Specify cropping rect for the picture in order: left, top, right, bottom. Separated by commas.
• -m --mask Specify path to the mask file
• -u --scaleup Set magnify rate to resize the original image by.
• -a --apply-corrections Set 1 to apply corrections suggested here, otherwise set 0.
• -d --detection Specify car detection XML file path