The invention discloses a detection method for monitored video speed anomaly targets based on a motion energy model. The detection method includes the steps: considering the anomaly targets and motion vector information of eight peripheral neighborhoods of the anomaly targets, and building the motion energy model, and enabling sum of a data item and a difference item to express a motion energy value of each space block; extracting a boundary value for each space block of all video frames, enabling each boundary value to express the motion energy value of the space block corresponding to the boundary value, and distinguishing boundary values of normal motion and abnormal motion; performing classification for motion energy values of testing concentration based on a binary classifier of a sigmoid function, taking difference testing the motion energy value of each space block and extracted boundary values as an input, and outputting a [0,1] probability value. The probability values corresponding to the space blocks are larger, the possibility of abnormal motion is higher, and namely, the possibility of speed anomaly is higher. According to the detection method, a motion intensity mode is effectively expressed, complex rate is low, detection accuracy and detection efficiency are balanced, and speed anomaly targets in monitored videos can be detected in real time.