The world has seen may MVPs but vehicle detection accuracy in real world conditions plagued most of commercial deployments. There is a catch and big design gap between a desktop prototype detecting “friends and family” cars and the product that consistently delivers 99% vehicle detection accuracy on city streets.
The secret for improvement is pretty simple - you need more sensors! Not just a magnetormeter or IR range finder but a combination of 2 or 3.
Farnell published my blog about smart parking sensor design which can be interesting to develop further discussion here.
Would appreciate your comments and feedback!
And of course you can buy all necessary sensors (IR, radar, magnetometers, ultrasonic) on seeed!
Nice blog, and thanks for sharing