Research On Laver Raft Detection Based On Waveform Feature Image Of Airborne LiDAR

Laver is an important marine agricultural cash crop.In recent years,the breeding range has expanded rapidly,creating economic wealth while also generating related problems.For example,unreasonable high-density breeding will reduce the production of laver,occupy shipping lanes to affect maritime traffic,occupy marine protected areas in violation of regulations,and cause green tides to damage marine ecology.Obtaining information on the number,location and overall distribution of laver rafts and breeding area and yield of laver is of great significance for the healthy development of the laver aquaculture industry.At present,the agricultural and rural departments in coastal provinces and cities have normalized the detection and monitoring of laver rafts.The raft detection methods mainly include manual navigation GPS detection methods and satellite image interpretation methods.Manual navigation measurement is time-consuming,laborious,and inefficient,and the real-time and continuity of the measured data are poor.Satellite image interpretation methods are susceptible to external environments such as clouds and fog,resulting in low accuracy and resolution.Airborne Li DAR has the advantages of flexibility,efficiency,high resolution,and resistance to external environmental influences,enabling high-precision and automated detection of laver rafts.Therefore,this article conducted research on the detection of laver rafts in northern shallow shoals of Jiangsu province based on airborne Li DAR.The main work and results of the paper are as follows:(1)A method for optimizing infrared laser waveform features based on the probability density overlapping ratio is proposedMultiple features in the full waveform of airborne infrared lasers can be used to distinguish between ocean and laver rafts,but the effectiveness of various waveform features in distinguishing the two has not been compared and verified.Therefore,waveform feature optimization research need to be conducted.First,the original measurement data of the airborne laser are decoded to obtain the full waveform data of the laser and data preprocessing is performed.Second,feature extraction(amplitude,area,and full width at half peak)are performed on the full waveform.Then,a probability density overlapping ratio is proposed and used to select the optimal waveform features.Finally,waveform feature images are created based on the selected optimal waveform features to provide a foundation for subsequent detection of laver rafts.The experiment shows that the probability density overlapping ratio of the amplitude,area,and full width at half peak extracted from the full waveform of the infrared laser is 1.50%,3.77%,and 4.28%,respectively.The probability density overlapping ratio of amplitude is the smallest,indicating amplitude is the best feature among the selected features.(2)A method for detecting and estimating the yield of laver rafts based on laser waveform feature images is proposedTo address the problems of low efficiency,accuracy,and resolution of traditional manual GPS on-site measurement and satellite image interpretation methods for detectinglaver rafts,a study of high-precision,automated detection of laver rafts and laver yield estimation based on the waveform feature images of the infrared laser and object detection models was carried out.First,a dataset is created through waveform feature image preprocessing and annotation.Second,the detection models for laver rafts were trained and experimentally validated and compared based on the YOLOv5 s,Faster R-CNN,and improved YOLOv5 s.Then,the improved YOLOv5 s was used to obtain information on the number,location,and area of laver rafts and to estimate the laver yield.The ultimate goal is to achieve automated detection of laver rafts and laver yield estimation based on the laser waveform feature images and object detection models.The experiment shows that,the improved YOLOv5 s model has an m AP,speed,and model size of 99.58%,71.15 frames/second,and 12.8 MB for detecting laver rafts based on infrared laser waveform feature images,respectively.Compared to the YOLOv5 s model,the improved YOLOv5 s model has similar detection accuracy for laver rafts,but the detection speed has increased by 29.3% and the model size has decreased by 11.1%.The number,area,and estimated yield of laver rafts in the experimental area derived from the waveform feature images by using the improved YOLOv5 s model were 853,285.55 hectares,and 326.865 tons,respectively.(3)An automated detection software for the laver raft based on the waveform feature images of infrared laser is developedA complete set of automated detection systems for laver rafts based on waveform feature images of infrared laser and object detection models has been designed and developed for the convenience of practical application.The system includes a login module,an infrared laser waveform feature image selection module,a model selection and parameter setting module,and a raft detection module,achieving fast and automated detection of laver rafts.