Measurement System Of Fruit Tree Canopy Volume Based On Machine Vision

Canopy volume parameter can provide the based data for many aspects.Such as fruit yield monitor,breeding,evaluation tree structure,irrigating,variable spray,precision fertilization.If the based data is inaccurate,it will accuse those orchard management and assessment losing objectivity and comparability.Eventually lose scientific research value.Now,the fruit tree canopy volume measurement is mainly composed of manual and non-contact automatic measurement.Practice shows that manual measurement technology is rough,time-consuming and precision low.It is difficult to meet the requirement of people.In order to reduce manual labor intensity and improve artificial measurement precision,all kinds of non-contact automatic measurement technology become the hotspot.The automatic measurement system can be divided into multi-angle view and single view measurement.Multi-angle view measurement use to match and reconstruct by canopy cloud point.This method can get higher accuracy result,but used sensor relatively expensive and much data processing.It seriously affected the operation efficiency and automation.Single view measurement obtained tree volume information from a single picture by canopy model.This method greatly reduced operating cost and processing data.But,it did not consider randomness and uncertainty of canopy with many factors.Such as,climate,environment,seasons.We can find those automatic measurement concentrated focus are the application of new technology and the development of processing algorithms.They didn't combine and use tree's feature,so that the method poor repeatability,developing complex and some biological significance unclear.In order to improve accuracy,reproducibility and universality of methods,reduce processing data and cost.We adopt theory of single perspective.Through image analysis method quantitative analysis the relation between tree feature and volume character,combined with tree's feature,canopy growth change,using engineering method and establish universal model.We took canopy image as research object,in depth study the automatic method of canopy volume.The main research contents and results as follows:(1)In order to solve those problems of high cost,large data.We used single-camera imaging method instead of sensor data collection,reduced operating cost.Used image analysis method to study feature extraction algorithm from tree canopy image,design a simple and effective image processing algorithms.The measurement objects were classified according to population distribution,growing situation and established their respective applicable methods.The results show it's feasible to estimate canopy volume by measuring canopy morphology.(2)In order to solve the universal problem of canopy volume model.We quantitative analysis the correlation between morphological characteristics and canopy volume,obtained significant linear relationship between area S and V,R2 was more than 0.90.Canopy characteristics varied with many factors of climate,environment,season,and so on.This made model parameters complex.So,we adopted five-point parameter calibration method to adjust model parameters before measurement.Five-point model compared with classical model.There were 90%sample of the total sample,their volume estimated error within 20%.Two kinds of mode's estimation precision were similar.The results show that parameter calibration method to solve the universal problem of biological model is feasible and effective.(3)In order to solve portability,transferability issues.We used Qt + OpenCV to design canopy volume measurement APP and made APP achieve running on Android system platform.Alternatively 15 osmanthus tree sample testing system,results were compared with PC platform system,R2 reached 0.95,the average absolute error is 0.31m3.The results show the APP was reliability.Running tests on different mobile platforms,we found that they had higher operating consistency,showed that the measuring system has strong transferability.And when the system CPU at 1.6GHz or above it,the processing 3264 ×2448 image within 20s,showed that the measuring system had high timeliness.