Abstract: In our report, aimed at the frequent human destruction of regional weather stations in China, combined with the construction of regional weather stations, based on the intelligent video monitoring analysis technology and the limb structure characteristics of pedestrian climbing fence, a fence climbing image detection method based on HOG-LBP joint feature and AdaBoost-SVM classifier was proposed. Firstly, the interest region is set up in the monitoring screen of regional weather station, and the frame difference method is used to detect the moving object based on the external scale of the moving object. The AdaBoost classifier trained offline is used to preliminarily select the suspicious moving target. Secondly, the HOG-LBP feature is extracted from the initially selected moving target, and the fence climbing event is determined by combining linear SVM. The results showed that the method not only has fast detection speed, but also has high detection accuracy and low misjudgment rate, which can quickly identify the abnormal behavior of fence climbing, and can be widely used in regional weather station scenes.