Innovation of agricultural digital business model based on remote sensing image target detection and embedded network

With the continuous development of science and technology, the application of remote sensing image technology has gradually penetrated into people's lives. In recent years, compared with the past, the resolution of optical remote sensing image has been significantly improved, and the image data level has also been significantly improved. All these provide the basis for remote sensing technology to detect accurate and small targets in the future, and make its application scope possible in addition to those rough remote sensing image classification. However, there is a certain gap between the technical level of remote sensing image development and the standard we want to achieve in actual operation, whether in terms of accuracy, universality or efficiency. Not only that, computer vision algorithm technology based on deep learning is also in continuous development, and has been applied in many aspects of our life, and has achieved good repercussions. Therefore, in order to solve the problem that the function of remote sensing image target detection technology is not perfect, this paper proposes a multi class target detection method ms-frcnn based on the regional detection learning target vision algorithm. This kind of technology is mainly aimed at the problem that the number of sub targets in the remote sensing image is complex, the range of targets is just wide and the scale changes, which leads to the negligence and omission in the image scanning detection of the technical model. Taking the feature pyramid network as the reference idea of this paper, the final samples obtained by combining various ideas are checked and observed. For the challenges in the process of research, such as getting the complex background environment in the text, uncertain and deformable targets, researchers analyze and extract the targets in the research samples according to the deformable convolution network as the basis of data research and investigation, so as to reduce all kinds of interference problems. The results show that the modified method has higher accuracy and stronger generalization ability compared with the previous method, and has achieved good results in the integration of optical remote sensing imaging technology.