Abstract: In order to improve the efficiency of finding key information in agricultural technology knowledge video, taking fruit pest knowledge video as an example, firstly, Sobel Edge Local Binary Patterns （Sobel-LBP） of the video image are extracted as features, and the preliminarily selected key frame set is calculated by fusing inter frame difference; then, the position interval of the first extracted key frame in the video sequence is used for secondary optimization to extract the key frame. The test results of four videos of different kinds of fruit diseases and insect pests knowledge show that the average value of the comprehensive index F1 of the key frame extracted by the algorithm can reach 0.925, the average accuracy is 91.35%, the average deletion factor is 2.46, and the average fidelity is 92.18%. It can effectively extract the key frames in the agricultural technology knowledge video, so as to reduce the redundant information in the video, and help to quickly and effectively transfer the new agricultural technology to the majority of farmers.

Key words: fruit knowledge video, key frame extraction, Sobel-LBP, quadratic optimization