At present, green plant walls are widely used in urban streets, but the area of green walls is usually large and the environment of each street in which they are located is complex, so the task of monitoring the growth status of plants in them is not easy. The use of manual methods for daily plant maintenance and detection of plant water deficiency status is time-consuming and labor-intensive. In recent years, many researchers have tried to use a combination of deep learning methods to achieve a method that is both efficient and can detect whether plants are in water shortage status in real time, but most of them use a single target detection network, thus cannot effectively exclude the interference caused by complex and variable backgrounds, so an integrated network-based plant status detection network method is proposed. Experiments prove that the method can effectively avoid the false identification of non-plant wall areas, thus improving the mAP value, and has practical engineering application value.