A computer vision based car identification and tracking system, while also identifying areas of repeat attendance, then representing it with a heat map that perpetually updates
- OpenCV using Python for computer vision based object idenitification/tracking
- Tensorflow using Python to gain pre trained models to base the object identification on
- matplotlib with Python to visualize data in graphs (heat map)
Using a pre-trained model that identiifies various objects including cars, cars on each frame of the video is identified with the x and y coordinates of the object's border is idenitified. Following this, these borders are mapped onto a new canvas. The x and y coordinates are also presented within the border's, constantly updating as the cars move across the frame
After breaking up the frame's pixels into an array the size of the pixels on the x and y axis, with every movement in an area of pixels, the arrays count goes up throughout the course of the video
Next, the array is visualized as a heat map with the 2-D array representing the frame of the map while higher numbers correlate to a darker shade of red in that rea, indiciating high concentrations of vehicle activity in those areas throughout the course of the video. This map is accessible through clicking a button at the top of the frame
Tensorflow model : https://github.com/tensorflow/tfjs-models/blob/master/coco-ssd/README.md