|
With limited spatial and temporal resolution of video recording camera, it is very difficult task to analyze object motion with higher accuracy. In addition, target object or target points of object in one-frame can be hardly recognized automatically in ordinary two-dimensional video recordings. If correlation or template matching methods is used to detect object position, initial object area or template should be decided manually. So fully automatic detection process is difficult. But with linear scan camera (LSC) utilizing line- image sensor, object position detection can be done easily with much higher accuracy. This article describes this feature of LSC in motion analysis. First d-value is introduced to evaluate correlation of two line images. This d-value is calculated in quarter pixel width for higher accuracy. Object displacement between given two-line image can be decided by monitoring this d-value. Steep drop of d-value means match of lines and higher correlation. Because d-value increases as line separation becomes bigger, this variation should be checked simultaneously for consistency. Typically over about 100-line separation this value increases, so pivoting line (template line) should be changed. Selecting suitable combination of two lines based on d-value will be important for overall accuracy of time-displacement analysis. Two lines should be selected to decide pixel shift where spatial quantum error is minimum. Suitable to line images can be selected by searching less d-value. Prior to that, two-dimensional d-value table should be made. Then one-pass trace of suitable two-line selection is carried out for time-velocity analysis. Fully automatic analysis method of position detection with LSC was proposed. To estimate object displacement, d-value was introduced and its validity was proved experimentally. In addition, by selecting suitably separated two points, resultant velocity at middle point can be much accurate than direct differentiation of time-displacement data.
|
