Research On Key Issues Of Automatic Observation Of Crop Based On Visual Anaysis

China is one of the most important agricultural production countries in the world.Agricultural production has been paid more attention in China since a long time.Many of the experiences and records of early agrometeorological changes reveal the relationship between the growth of crops and meteorological changes,which provide a response to cope with different weather changes and natural disasters for future generations.In short,it provided support to the development of Chinese agriculture.In recent years,computer vision has been utilized in seed quality testing,agricultural grading and processing,prevention of crop disease and insect control,plant growth detection and other aspects of research and so on in the developed countries.In comparison,the related researches developed relatively slow in China.Therefore,it is necessary to improve the automatic observation capabilities of crop and agrometeorological changes.In this thesis,the research on agricultural meteorological automatic observation was carried out.The research was based on the coverage of green crop and the measurement of crop height.The details are as follows.1.A crop coverage extraction algorithm based on priori threshold optimization CNN was proposed.Firstly,the pixels were preserved in a rough way by priori threshold segmentation method.Secondly,the rough segmentation result was input into the image by the trained convolution neural network.Using Pixel classification,the crop coverage was extracted.2.Domain knowledge-guided crop height stereoscopic vision measurement method is proposed.Firstly,both the left and right two images are divided into two different regions,upper half region and lower half region.In the upper half region,Harris corner features were extracted and the same pixel points were found in the corresponding region.The pixel parallax was also calculated for the measure of crop height according to the trigonometric function.Secondly,the lower half region was used to extract the crop coverage and the height of the crop was calculated.Finally,the plant height calculated by these two methods was combined to solve the effect of the observation that the crop target was too small at the early stage of growth and the decrease of the coverage in the later growth period,so that the results of plant height measuring was accurate.3.According to crop automation observation,the relationship among crop coverage,plant height and other growth parameters was analyzedto solve the problem of crop growth period observation.