The Research of Vehicle Tracking Based on Difference Screening and Coordinate Mapping - Advances in Image and Graphics Technologies - page 323

When the pitch angle of vehicle changes according to the characteristic of the fixed
ratio distribution line and the value of r, we can calculate the value of Δy, and use the
formula (6) to calculate y b , then the actual spatial distance of front vehicle corres-
ponding to y b can be calculated according to formula (5). Thus the effects made by the
changes of pitch angle on distance measure can be eliminated.
PreScan is used to build simulation scenario to verify the correctness of the method
in this research paper. When the vehicle runs on bumpy road, radar and camera sensor
are used to detect the distance from the front vehicle as shown in figure 15. As one
can see from the figure, the effects made by the changes of pitch angle on ranging can
be resisted by this method.

Fig. 15. Distance Measurement Result of Eliminate the Influence of Pitch Angel Change

3.3 Vehicle Tracking by Kalman filtering

The vehicle identification used by classifier (trained by adaboost algorithm) can only
obtain an approximate area of the vehicle in the image, rather than accurate to each
pixel. Therefore when the vehicle is directly located by this mapping relationship, a
phenomenon will appear: the calculated distance from the front vehicle violently
fluctuates (the thin blue line shown in figure 16). To avoid the impact made by the
above phenomenon and guarantee the wanted impact on vehicle location, kalman
filter technology is used for vehicle tracking in this essay. In order to achieve the
kalman filter, mathematical model of the target movement is built first. A dynamic
model of discrete control system is introduced in the following way:
() () ()

= + (7)
In additional to the system dynamic model, the system measurement equation is al-
so introduced to describe the relationship between the measured value and the target
moving state as shown in the following formula:
()

Xt AXt ˉ t

() ()

= + (8)
The above kalman filter method is used to filter the distance from the front vehicle.
The filtered and unfiltered vehicle distances curve are shown in figure 16. As one can
see on figure 16, the filtered vehicle distance is relatively smooth and stable.

Z t Xt ˅ t