Aiming at the problem that the target occlusion is serious and the scale changes greatly in the pedestrian tracking of the UAV

a stable real-time tracking algorithm based on the designated pedestrians of the UAV is proposed which is ECO-HC tracking algorithm. The histogram of oriented gradient（ f HOG） and color name（ CN） features are extracted

and the FHOG feature and the CN feature are merged by using the interpolation model of the training samples in efficient convolution operator（ ECO） tracking algorithm. Using decomposition convolution operator

compact training sample model and sparse model updating strategy to reduce risk of over-fitting and reduce complexity of the algorithm. The position of the maximum value of the response fraction is the target center position. Fast scale filtering method is introduced to adaptively obtain the optimal estimation of the target size and improve the tracking precision. The experiment selects 60 segment public pedestrian databases

and the results are better than the other four outstanding tracking algorithms in recent years. Among them

the average distance precision（ mean DP） and average overlap precision（ mean OP） are 83. 2 % and 87. 3 %. The average speed is 56.3 frames per second. For the self built pedestrian database captured by UAV

it also has good robustness and real time processing speed.