Seedling height is an important feature in the process of seedling cultivation, and it is also an important reference index for seedling growth and screening of excellent features. In view of the problem that professional measurement tools and marked measurement methods are mostly used in the current research, a measurement method of the seedling height based on monocular image depth estimation technology was proposed in this study. Firstly, the NYU Depth Dataset V2 depth dataset was enhanced to make the model have better expression ability. The depth estimation network structure is a U-shaped network structure, which is divided into encoder and decoder. The encoder part took ResNeXt 101 network as the main body to extract the depth feature information of plant image. The decoder was mainly based on the up sampling, and a jump connection module was added between the encoder and the decoder to increase the detail information of the depth image. Compared with different depth estimation models, the depth estimation model achieved the best Root Mean Square Error (RMSE), which was 0.165. It showed that the depth estimation model can better complete the estimation task of depth information. Through the calibration of the maximum depth value, the real distance from the shooting point to the plant can be calculated according to the depth information, and the seedling height can be measured in combination with the pixel height of the seedling plant in the image and the calibrated field angle. In order to verify the effectiveness of this method, we collected 1 728 images of tomato seedlings, 160 images of pepper seedlings and 160 images of cabbage seedlings at different distances for plant height measurement. The results showed that within the shooting distance of 105 cm, the Mean Absolute Error (MAE) of tomato seedlings was 0.569 cm, the RMSE was 0.829 cm, and the average plant height ratio was 1.005. The MAE of pepper and cabbage seedlings were 0.616 and 0.326 cm, and the RMSE were 0.672 and 0.389 cm. The average calculation time of the height of each seedling was 2.01 s. The experimental results showed that this method was feasible and universal for seedling height detection usage. The results of plant height measurement under different light intensities showed that when the sensitivity was less than 160, the MAE of plant height measurement result was 0.81 cm, which still had good measurement accuracy. In order to realize the height measurement of multi-target plants, YOLOv5s was used to train and test the images with 2-6 plants. The test results showed that the accuracy of the model was 98.20%, the recall was 0.98, and the mean Average Precision was 84.6%.When the number of plants in a single image was within 5 (less than 6) targets, the average values of MAE and RMSE were 0.652 and 0.829 cm respectively. The results of this study showed that the model can accurately detect the height of multiple plants from a single image, and can accurately measure the heights of a variety of seedlings within different distances and certain light intensity changes, which can provide a non-destructive plant height measurement method for the study of seedling cultivation and growth period judgment.