Research On Key Technologies For Remote Measurement Of Crop Growth Information

Getting the crop growth information timely and accurately can provide guidance ideasfor field management. Due to the problem of agricultural production’s scattered ingeographical location, the work of information gathering is a big task, so it is unable tocomplete it by artificial methods only. Therefore, realizing the remote access of crop’s growthparameters, not only can reduce the amount of manual labor, but also can improve theefficiency of information gathering and reduce the costs, it has important practicalsignificance. In view of this, aiming at the shortcomings of the existing remote monitoringmethods, this topic mainly researched on the technologies of crop growth conditions’ remotemeasurement which based on machine vision, the main purpose is to achieve real-time andnondestructive monitoring of crop’s growth parameters, such as plant height, leaf area indexet al, and provides a simple, efficient and universal method for crop growing information’sremote access. The main research contents and the results which have been obtained are asfollows:(1) The thesis proposed and realized a method for crop height’s remote measurementwhich based on geometric principles. First, capture the monitor video stream and realize therotation step quantization, then by calculating the rotation angle of camera optical centerwhich from the horizontal position to the highest point of crop as well as from the highestpoint to the lowest point, realize the crop height’s online and real-time calculate. Taking cornas the measurement object, the experimental results show that, this way is simple andefficiency, the average relative error of crop height’s remote measured value is3.75%, and themaximum relative error is6.70%, the accuracy satisfy the requirements of agriculturalproduction.(2) The thesis realized the improvement of C-V model which based on priori information,and this provides a guarantee for the accurate calculation of leaf area index. By introducing anenergy penalty term, in the process of evolution, can avoid the repeatedly initialize operationof the level set function, so it greatly simplifies the calculation workload. At the time ofimplementation, depending on the target to be detected, first render the image into an easy segmentation model, then extract the priori information and perform a simple initialsegmentation, to complete the initialize operation of level set function, at last the final contourcan be got through iterations. The experimental results show that, for the crop images of threedifferent resolutions which taken by IP camera under field conditions, the method this paperproposed can gets a better segmentation result，meanwhile the average value of segmentationefficiency increase about45.40%.(3) The thesis implemented the remote measurement of leaf area index. First establish aremote predictive model, then segment the images which taken by IP camera use the methodthis paper proposed, through the corresponding processing, can calculate the analytical leafarea index values and make it as the model input, eventually gets the remote forecast data.Taking corn as the measurement object, the experimental results show that, the averagerelative error of remote predicted value is2.92%, the prediction accuracy is higher.