Implementation of Detection System of Grassland Degradation Indicator Grass Species Based on YOLOv3-SPP Algorithm

Abstract

Abstract Due to climate change and human factors, grasslands in the Sanjiangyuan area have degraded. At present, most grassland workers use manual methods such as visual inspection, measurement and remote sensing technology or neural networks to conduct macro evaluation of grassland. However, the emergence of grassland degradation indicator grass is an important sign for grassland degradation. Therefore, it is simpler and more convenient to provide early warning of grassland degradation through the detection technology of degradation indicator grass species. In this paper, the degradation indicator grass species Stellera chamaejasme was used as an example, and the YOLOv3-SPP algorithm was used to detect the degradation indicator grass species. First of all, collect and process data on the spot, and establish a data set of Stellera chamaejasme for deep learning; secondly, use YOLOv3-SPP algorithm to train and test the data set. By continuously improving the quality of the data set, the accuracy of model detection is improved. Then through the test of the verification set, the accuracy rate reaches 95% and the recall rate reaches 98%. It is proved that the model can be used to detect grassland degradation indicator grass species under complex grassland background. Finally, the detection system of Stellera chamaejasme with uploading and detecting functions is realized.