Diagnosis Of Crop Disease,Insect Pest And Weed Based On Image Recognition

This study aimed at the quick diagnosis crop disease, insect pest and weed, integrated the knowledge and experiences on diagnosis of crop disease, insect pest and weed from the agriculture experts, synthetical applied technologies of artificial intelligence and Internet. The goal of tele-diagnosis on crop disease, insect pest and weed with the digital image recognition by Internet was came ture. Acquired advances are as follows:1. In the aspect of images pre-processing methods, analyses and with methods of images enhancement according to the features of crop disease, insect pest and weed images. The new methods of using reverse color images in the crop disease, insect pest and weed image enhancement were proposed. The methods have briefness and facility, small operation and strong ability eliminated noise compared with the traditional image enhancement methods.2. In the aspect of images segmentation methods, the methods of analysis gray value of Hue image to fix on the threshold segment image was proposed according to the characteristics of crop disease image. The segment precision was improved obviously with the methods. To segment color images, the segment effect was satisfying using Hue image of HIS color system, following, was the segment effect of Red image of RGB color system. It was put forward that B image was sensitive to light and to segment these images obtained in dark or unclear condition using B images were suited. The new segment aim image methods of using extracted outline were proposed. Firstly, obtained the reverse color image of the original image (aim image) used relevant algorithm. Second, get the two value image used threshold segment, then made the erode operation. Thirdly, extracted outline of the two values image. The segment precision was more veracious, hardly any noise, compare with the traditional methods. The efficiency segment image methods of using the skill of extract outline and seed fill in to process the image with the symptoms such as hole and howe or protuberance. The methods can clear up a little noise but nervations. To segment leaves, as other leaves in image background, the segment method that combined auto-threshold with handwork sampling to modify segment threshold can segment accurately aim leaves.3. In the aspect of extracted the image features, put forward the methods of using the chroma coordinate value of RGB color system and H (hue), I (illumination), S (saturation) of HIS color system as key characteristics to recognize and diagnoses crop disease, insect pest and weed. Six sets simulated models of cotton leave chlorophyll concentration using the color values were founded. The correlations of the B/R , color coordinate b, b/r of the RGB color system and saturation of HIS color system with chlorophyll concentration were significant, and chlorophyll concentration can be forecasted using the value of B/R, b, b/r and s. The correlations of the values of color character from two sides of leaves with chlorophyll concentration were unite, furthermore, there were more significant correlation of color value of back leaves with its chlorophyll concentration. The error percent to forecast chlorophyll concentration were 7.8- 13.65%, and used the model of b/r with chlorophyll b , y=0.8058x+1.2403(R2=0.8056**), the right percent to forecast chlorophyll b concentration was beyond 90%. The leaf green grade (chlorophyll content) forecast models used leaves color value were proposed. The new method and new criterion of decide harm grade by disease or insect pest was put forward. The methods of using the energy...