Research On Classification Technology Of Lidar Echo Data Based On Decision Tree

In the pursuit of laser distance 3D mapping technology with higher distance,higher precision and higher efficiency,obtaining complete data information and rapid processing is one of the key research directions of current surveying and mapping technology.The streak tube laser radar imaging system has the characteristics of large detection field,high detection efficiency,high detection sensitivity and full waveform sampling.Therefore,it has become a research hotspot of current laser radar system.However,the mainstream classification processing method for streak echo data at present is for point cloud data processing,and its processing rate is only one percent of the sampling rate,which is far from meeting the high efficiency requirement of surveying and mapping.The classification and processing of massive laser echo signals has become a bottleneck problem that limits the widespread application of the new system of laser radar.In this paper,we propose to directly classify the original echo signals of the streak tube,and perform in-depth analysis on the simplest original echo data to extract the key data for different needs.At the same time,this paper combines the machine learning method in the field of data mining with the streak-tube lidar mapping system for the first time,and chooses the decision tree classification algorithm with the characteristics of complete theoretical structure,high classification accuracy and good complex terrain classification effect,and proposes a decision tree based on decision tree.The laser radar echo data classification scheme aims to achieve efficient processing of massive callback data.In this paper,by analyzing the typical characteristics of the original fringe echo signals,10 features such as target area,average intensity and elevation difference are obtained.The decision tree is trained and pruned using the obtained feature data set,and the decision tree model suitable for the classification of the original fringe echo signals is obtained.The test accuracy is up to 88%.The time for parallel acceleration processing of 60,000 original echo images was reduced from 4 hours to 0.76 hours.Finally,in order to visually display the classification result of the method,the classification result marks based on the storage of the original streak echo signal and the document-based operation are completed.