Study On The Information Of Ramie Leaves Based On Digital Image Processing Technology

For achieving the rapid diagnosis of ramie (Boehmeria nivea L.Gaud) nitrogen content, three nitrogen levels of ramie were planted in pots, and the SPAD value, color features of standard images and the nitrogen, phosphorus, potassium content of ramie leaves were obtained respectively by using the information technique and the conventional method at different growth stages.The relationships among them were analyzed. The primary research conclusions and contributions were listed as follows.(1) The optimum method of measuring SPAD value of ramie leaf was to multi-point measure starting from the leaf stalk base, within 20% -80% of the leaf surface along the main vein, avoid the leaf veins, and then average the results.(2) Leaf images captured under following standard were clear and accurate:use CANON EOS 40D SLR camera and CANON EF-S 17-85mm F/4-5.6 IS USM zoom lens, fix vertical distance between the lens and the background of 50cm, set the lens focal length at 35mm, the shutter speed at 1/60s, the aperture at F=5.6, with ISO 100, auto white balance and flash off. Film in darkroom and use red flannel as background. Use circular light source with color temperature of 5400K, Ra>90 and 22W power.(3) The 3×3 template median filtering method effectively eliminated image noise while maintaining the clarity of the original ramie leaf image.(4) A segmentation method for ramie leaf images was proposed. First, use the Otsu threshold method to segment the green channel image of ramie leaf. Second, fill and label each area in the image. Third, select the maximum area as the aim and set the others as background.(5) Image zoom out has significant effect on calculated perimeter of ramie leaf, but there's no significant effect on calculated leaf area, leaf width and leaf length.(6) There were significant differences among total nitrogen content of ramie leaves with different nitrogen treatments, while total phosphorus and potassium contents were not significantly different.(7) There were significant differences among G color values of ramie leaf images with different nitrogen treatments.(8) The SPAD values and total nitrogen contents of ramie function leaves were significantly and positively correlated at each growth stage. Six regression models were built using SPAD value to predict the total nitrogen content of ramie leaves at each growth stage and the average prediction accuracies were from 85.23% to 92.08%.(9) Digital image processing technology can be used to diagnose total nitrogen nutrient status of ramie leaf. Six regression models were built by using the selected color features to predict the total nitrogen contents of ramie leaves at each growth stage, and the average prediction accuracies were from 86.35% to 93.83%.(10) The image processing software system for ramie leaves (RLIPS) was developed based on MATLAB. The functions of the RLIPS were not only to pre-process and segment digital images of ramie leaf, extract color features and morphological features of ramie leaf images, diagnose nitrogen status of ramie leaf, but it also included some basic image processing functions.