The accurate extraction of cow targets serves as the basis for the behavior analysis such as lameness detection, ruminate and estrus. In order to realize the automatic tracking and monitoring of dairy cows in large-scale farms, the correlation filtering algorithm was integrated into the basic framework of target extraction, and a cow target extraction algorithm (CFED) that combined correlation filtering and edge detection to extract the cow target was proposed. Firstly, the correlation filters constructed by the color names and the Histogram of oriented gradient were applied to obtain the cow target range box. Then 13 edge filter templates in different directions convolved the target image box to get the edge image. Finally, the edge information and color feature were combined to extract the cow target. In order to verify the effectiveness of CFED algorithm, experiments were conducted on nine pieces of video samples of moving cows under different environments and interferences. The results showed that the average overlap rate between the CFED results and the manually marked results reached 92.93%, which was 35.63 percentage points, 32.84 percentage points, 20.28 percentage points and 14.35 percentage points higher than that of Otsu, K-means clustering, frame difference method and Gaussian mixture model method, respectively. The false positive rate and false negative rate of CFED were 5.07% and 5.08%, respectively. The average time cost was 0.70s per frame. This result showed that the proposed CFED algorithm had good target detection ability in complex environments such as weather, scale and occlusion, which can provide an effective method for accurate and rapid extraction of dairy cow targets.