Lack of automatic target recognition in existing transmission line protection technology causes poor performance on crane collision monitoring. Therefore

we propose a forewarning algorithm based on feature fusion and random forest to solve this problem. By analyzing the unique characteristics of the cranes

we design a ―Shape-Texture-Color‖ fusion feature extraction method. By the method

firstly

the shape features of the arm are extracted by using improved edge line detection method; secondly

the texture features are extracted by using the LBP operator; thirdly

the color features in the two-step quantified HSV color space are extracted. After extracting the fusion features

the target will be identified by the random forest classifier. If the target is identified as a crane

a danger level will be generated according to the parameters of posture and distance. The experimental results show that the recognition accuracy of this algorithm on the crane image library averages 90.6%

increased by 17.3%

7.9% and 12.6% compared to those by a single feature extracting method

respectively. Meanwhile

this algorithm is capable of real-time processing when it is applied to intelligent video surveillance system of the transmission lines.