Application of Convolution Neural Network Algorithm Based on Intelligent Sensor Network in Target Recognition of Corn Weeder at Seedling Stage

Abstract

Grass damage in the seedling corn field has always been an important factor affecting the growth and development of crops. The existence of grass not only compresses the living space of corn seedlings but also easily causes insect damage. Therefore, it is essential for weeding in the seedling corn field. The existing weeding methods usually use manual or chemical herbicide spraying, which is not only time-consuming and laborious but also inefficient. With the development of artificial intelligence and modern agricultural technology, the use of robots for field weeding has become an effective means, which has attracted more and more attention of researchers at home and abroad. Therefore, based on the full investigation of the development of relevant technologies at home and abroad, this paper carried out the research on the real-time target recognition and ranging method of field weeding robot and proposed a target recognition method of weeding corn at the seedling stage by using an intelligent sensor network and deep learning convolution neural network. Among them, the intelligent sensor is mainly used for target ranging and obstacle avoidance, and CNN is mainly used for target recognition. Taking the images of corn seedlings and weeds in the seedling stage under natural environmental conditions as samples, the migration training is carried out through the depth network model of the COCO data set, and the convolution features are shared by the CNN depth network model and Fast-CNN depth network model. VGG and ResNet feature extraction networks are compared. The experimental results show that the CNN depth network model based on this paper has obvious advantages in rape and weed target recognition. The target recognition accuracy of rape and weeds can reach 87.64%, and the recall rate can reach 80.23%. Compared with other models, it has obvious advantages, which proves the effectiveness of this model.