Abstract: As the performance of image detection method in the normal PC is not satisfactory nowadays because of the huge amount of high resolution image transmitting in current high-speed network,this paper presented a GPU-based solution for image extraction detection．We obtained feature extraction by using SURF(Speed Up Robust Features) algorithm and classified image features based on SVM(Support Vector Machine) algorithm．Furthermore,we utilized the parallelism of GPU float point arithmetic to optimize the system．As a result,the performance of GPU-based system achieves 5to 9times faster than the CPU-based system．

Key words: Speed up robust feature,Support vector machine,GPU