Aiming at the problems of high intensity, high cost and low efficiency of artificial falling vine of cucumber crops in long season, a greenhouse cucumber automatic falling vine device based on canopy height threshold recognition was designed. In order to ensure the accuracy of cucumber canopy feature extraction, cucumber canopy images collected at 8:00, 12:00 and 18:00 were preprocessed, and smooth cucumber canopy binary images were obtained based on Otsu algorithm. The Hough transform algorithm was used to identify the rope bundle at the top of cucumber canopy, and the value range of each component of the green part of HSV color space in the region of interest was summarized by statistical method, and the change rule of the ratio of green pixels to the total pixels in the region of interest was studied. In order to prevent cucumber tendrils from entangling the horizontal rope bundle at the top of vine canopy, cucumber plant height 184-185 cm was taken as the toppling control target, and the average proportion of green pixels in the region of interest was determined to be 0.162-0.182 as the toppling threshold range of the overall toppling device. The results showed that 90% and 10% of the total test times were from 184-185 cm and above 185 cm, respectively. In the group higher than 185 cm, the distance between the highest plant and the top rope bundle (195 cm) was 9.2 cm, and no cucumber tendrils were entangled in the horizontal rope bundle at the top of the canopy of yellow melon vine. The method of determining the threshold of cucumber vine by the proportion of green pixels in the region of interest could meet the actual requirements of cucumber production.