SURF feature descriptor is proved effective in object tracking applications.However

SURF feature descriptor requires too much complex calculation

which is difficult to run in embedded system for real-time operation.Therefore

improved SURF feature descriptor

which is designed based on triangle and diagonal featrues

is proposed.The evaluation of SURF and improved SURF proposed is performed in the context of matching of same scene under different view conditions.Experiments show that the proposed improved SURF keeps sufficiently distinctive.In order to compare computational efficiency between improved and previous SURF

the test is also performed in DSP

and result shows that performance of the proposed method is improved about 47 %.At last

a target tracking strategy based on SURF feature feature point matching are designed

and the effectiveness of improved SURF feature descriptor in target tracking is verified.