The invention discloses a method for intelligent video recognition of multi-work-type workers. The method comprises the steps of collecting video images through a front-end camera; decompressing the video images collected, and decomposing the video images into YUV pictures; converting the YUV pictures into RGB pictures; decomposing the pictures frame by frame, randomly extracting multiple frames to make an analysis comparison; comparing a next picture frame with a last picture frame, updating the background environment in real time by utilizing self learning, and extracting target objects referring in particular to people from the background environment; and after the target objects are locked, performing gray level calculation on human body characteristics and form recognition so as to determine workers of each work type. According to the method, the workers of various work types close to electric power facilities can be recognized conveniently and timely.