In order to accurately determine the growth stage of corn, remotely monitor the growth of corn and analyze the relationship between growth stage and field environment elements, this paper proposes a deep local correlation neural network to overcome the multimodal and fuzzy problems in the identification of corn growth stages. In the Oxford VGGNet (Visual Geometry Group Net) model, a new supervised layer, namely the local correlation loss layer, is added to improve the discriminating capability of deep features. Based on the proposed new image recognition algorithm for corn growth stages, the environmental element monitoring function is expanded, and a corn farmland monitoring system based on deep learning is designed. The system consists of a corn farmland monitoring device and a cloud server. The monitoring device collects corn images, meteorological elements and field location data, and sends them to the cloud server through a 4G wireless internet. The cloud server uses the deep local correlation neural network to identify the growth stages, and displays the results and stores them in the database. The simulation experiments show that the average recognition accuracy of the deep local correlation neural network reaches 92.53%, compared with 87.21% of VGGNet and 88.50% of LSTM, and the accuracy rate is increased by 5.32% and 4.03%, respectively. The field test results show that the accuracy rate of the system can reach 91.43% in the field environment, and it can stably monitor the growth of farmland corn, which has important application value.