Actually this is quite inaccurate, because the tyre is flattened at the bottom, so the radius is not what you would expect it to be from the wheel's circumference.
There are two ways to deal with this, and I've done both (although on a motorcycle).
First method: put the car on a large, flat area of ground (like an empty car park). Choose which wheel you will observe, and push the car until the valve is exactly at the bottom of the wheel. Mark the ground with a chalk line - this represents your starting point. It helps to mark a line on the tyre, next to the valve, for even better accuracy.
With the steering straight, push the car along, counting how many times the wheel turns. After (say) five turns, line the valve up at the bottom and draw another line. Then measure the distance between the lines with a tape measure. Divide by five to get the distance travelled per revolution of the wheel. Then, when you know the RPMs of the wheel, you can calculate the speed.
It helps if there is another person who can monitor how many pulses are produced by the sensor as you push the car along for five (or whatever) full revolutions of the wheel. Then you will know how many pulses you get for each turn of the wheel, and you have all the information you need to make your speedometer.
Second method: On a long, straight, empty road, use a GPS to measure your speed. Start at, say, 10kph and go up in 10kph jumps. A passenger monitors the pulses from the sensor and notes their frequency - this normally requires a portable oscilloscope, of course. Then you can plot a graph of frequency against speed, and from that work out the formula to convert from the pulse frequency to the speed in kph.