Recognition Of Summer Corn Growth Stage Based On Computer Vision Technology

Accurate information of crop development period can provide guidance for agricultural mechanization operation required by the growth and development stage of crops,so as to obtain productive and high-quality crops.At present,the data source of the crop growth period is mainly obtained through manual observation,which is time-consuming,labor-consuming,discontinuous,and is easily affected by people’s subjective factors.Information technology has promoted agricultural modernization.As one of the hot spots in the field of modern agriculture,because it can monitor large areas and has the advantages of saving time,effort,and continuous observation of crop growth trends,this technology has been used to obtain information on crop growth period,crop height or leaf area,and leaves disease identification and other aspects of agriculture.This paper takes five growth stages of summer maize as the research object: seedling stage,three-leaf stage,sevenleaf stage,jointing stage and tasseling stage.According to different cultivation methods and different growth periods,different segmentation methods were selected.The improved genetic algorithm combined with Otsu algorithm was used to extract maize seedlings at seedling stage.By comparing the change of the number of maize seedlings in two consecutive days,the maize seedling emergence stage was determined.An adaptive AP-HI color segmentation algorithm was proposed for maize at three leaf stage,seven leaf stage and jointing stage.By determining the ratio of the long axis and the short axis of the ellipse with the same second moment as the single maize plant,it is determined whether the ellipse enters the trifoliate stage.The focus of the examination of the buckeye stage and the jointing stage is the change of the coverage.In the tasseling stage,the tassel was enhanced based on YCbCr space,and the tassel was further extracted from the enhanced RGB image using HSV space.Finally,the tassel was successfully extracted by combining with morphological operation.Tasseling stage was determined by the change of tassel number in the region.Based on the detection results of each period,the arrival date of each growth period was recorded to provide a basis for the decision and implementation of precise farming measures.The main conclusions of this paper are as follows:(1)Through the processing and analysis of maize seedling images,the results show that the error rate and miss rate of the algorithm under the spring rotation tillage treatment are 3.10% and 1.96%,respectively,and 2.24% and 1.81% under the spring deep pine treatment.At the same time,by comparing the detection results of the image-based development period with the artificial detection results,the error is found to be very small,in line with the interval of 1 to 3 days,indicating that the algorithm has a high segmentation accuracy,and can effectively complete the segmentation of maize seedlings under a complex background.(2)Through the processing and analysis of the images of the three-leaf stage,the seven-leaf stage and the jointing stage of maize.The results show that the error rate and miss rate are 3.02% and 1.88%,2.47%and 1.91%,respectively.At the same time,the automatic detection results were compared with the manual detection results,the error is found to be very small,in line with the interval of 1 to 3 days,indicating that the algorithm has a high segmentation accuracy,and can effectively complete the segmentation of maize seedlings under a complex background.(3)Through the processing and analysis of maize tasseling images.The results show that the error rate of the algorithm is 7% and the recall rate is 95.3%,which indicates that the segmentation accuracy of the algorithm is high,and the algorithm is simple and easy to operate based on color space.At the same time,the results of the tasseling stage were compared with the results of the manual test,and the error range was found to be in line with the criteria of 1 to 3 days,which indicated that the method could effectively complete the corn seedling segmentation under the complex background.