Dissertation > Excellent graduate degree dissertation topics show

Study on Automatic Identification Thchnology for Rice Plant Hoppers

Author: LiuDeYing
Tutor: DingWeiMin
School: Nanjing Agricultural College
Course: Agricultural Mechanization Engineering
Keywords: Rice plant hoppers Image segmentation Feature extraction Support vector machine Taxonomy
CLC: TP391.41
Type: PhD thesis
Year: 2011
Downloads: 210
Quote: 0
Read: Download Dissertation


As the most important food crop, rice is the staple food for over60percent of Chinese population. Hence the production of rice plays a decisive part in agricultural production and food security all over the country. Rice plant hopper belongs to the family Delphacidae of Homoptera and is a general term for several kinds of plant hoppers, including Sogatella furcifera,Nilaparvata lugens and Laodelphax striatellus, which work together to damage rice. Rice plant hoppers are the main pests in paddy field and thus threaten the production of rice greatly. The real-time examination of pests is a way for integrated pest control. Only through accurate examination can we reach the aim, i.e. to control the number of insects, without either economic loss caused by pests, or waste and pollution of rice and environment due to excessive prevention. Therefore, it is obvious that only by study on effective automatic identification technology, can we get accurate information of pests in time and then can the scientific decision for integrated insect disease prevention be made.Currently, the real-time detection and prediction of insects in paddy field relies on an identification method like this:the plant protector turns on a200w incandescent lamp at night to trap the insects which are soon killed by DDVP;in the next morning, insects are collected, sorted and counted by hand. Insects poisoned to death come in every shape, what’s more, the location of feet and antennae is at random, and the feet, antennae and wings are likely to be broken. On the other hand, due to the tiny size and pale color of the rice plant hoppers and the sharp difference in the extent of damage to industrial crops, sometimes the insects can not be recognized by the naked eye. The plant protectors should observe them via magnifying glass, anatomical lens or even microscope, and later identify the categories in accordance with the identification key. The identification method is inefficient and depends too much on the experts, moreover, the poisonous gas given off by insects killed by insecticide will damage the staff’s health.The current study on the automatic identification technology for insects is still at the initial stage both at home and abroad. Most of the on-going studies have been conducted under specific circumstances, with the static subject of researches are static. The samples of insects for taxonomic studies are standard samples or samples artificially cultured in the laboratory, which differ more or less from insects in the nature in terms of color, texture and shape etc.However, the real-time and dynamic data of insects in a certain region are needed in the reality to guide the practice. Hence the paper did some research on some key techniques, i.e. how to collect the digital images of insects in the paddy field, esp. rice plant hoppers, how to describe the identification features of rice plant hoppers and how to establish the classification model for insects.To acquire the digital images of insects, esp. rice plant hoppers, in a natural state, an automatic image acquisition device with machine vision has been designed. The acquisition equipment consists of control system, acquisition platform, camera system and a base where acquisition platform and camera system are fixed. The acquisition platform is composed of fixture, white Dacron curtain and curtain driving device. The acquisition platform can move in the X direction and the curtain can move in the Z direction. Rice plant hoppers can be trapped and adsorbed to the curtain by a160w self-ballasted high-pressure mercury vapor lamp. Camera system is composed of color digital camera, ring and cold light source, microscope zoom lens fixed on camera, camera stand, light source stand, camera platform and light sources platform. Light source is installed between the white curtain and camera, while the camera is installed on the camera stand. Both the light source and the camera can be adjusted upwards, downwards. What’s more, the distance between the camera and the curtain is adjustable, while the distance between the camera and the light source can be adjusted separately. Control system is composed of computer, micro-controller, drive and image acquisition card. The micro-controller controls the acquisition platform to move in X direction and the white curtain in Z direction via the drive. The PC captures images from camera and image acquisition card which takes the picture of insects adsorbed on the curtain at regular intervals. And the connection of PC and micro-controller will coordinate the automation of camera shooting and the movement of image acquisition equipment.The color and texture on the back of rice plant hoppers is a relatively stable feature to identify the insects. In the image processing, the image of the back of a single suspected rice plant hopper was extracted. First the paper chose the blue component of the image of rice plant hoppers (B=140) as the color threshold, binarized the image, and then worked out morphology-based filter to remove non-target areas, e.g. feet, antennae and noise. Later the paper marked the connected regions, calculated the size of segments, and selected those suspected rice plant hoppers whose size were similar to that of rice plant hoppers (1398-3847) and so as to reduce the taxonomic workload, it eliminated a majority of insects of which the shape and size were different from that of rice plant hoppers. Next the paper decomposed the images into binary images of a single target area, and captured images with complete information of single insects’back through the comparison with the original images. All the images are128by128pixels.The paper transferred the regional images of insects’back from spatial domain to frequency domain, described the color and texture of insects’back through Fourier spectrum, and then extracted lxl(1≤9)2-D Fourier spectrum window data. With the top left corner of the windows always serving as the center, the paper finally constructed p-dimension eigenvector to identify insects. According to the Fourier spectrum data of the images of the back of four categories insects, the paper calculated the mean of each feature in all the samples. There are certain differences among the means of the81features in four groups of samples, hence the effectiveness of the81spectrum features in identifying the four groups of samples. After the analysis of the variance of each feature in different samples, it turns out that the fluctuation of the81spectrum features is not fierce on the whole, and that means the features are relatively stable when presenting the detailed samples. The examination of the significance level of the81spectrum features in the four groups of samples with single factor analysis of variance shows that only seven spectrum features don’t have a significant difference, which means these spectrum features can be applied to describe the color and texture of the back of insects and it is feasible to identify different samples by these spectrum features.A model for classifying insects has been constructed based on support vector machine. All in all,169images of insects were selected, including34pictures of Sogatella furcifera,34pictures of Nilaparvata,34pictures of Laodelphax striatellus,31pictures of three kinds of leaf hoppers, and36pictures of other insects, i.e.2pictures of ant,10pictures of sap beetle,2pictures of shore fly,19pictures of rice stem maggot,2pictures of lygaeid,1picture of flea hopper. First the paper constructed a p-dimension eigenvector to describe the characteristics of the insects on the basis of data of2-D Fourier spectrum window, and then divided the samples into two parts, i.e. training set and test set. Later the paper utilized standard C-support vector machine, radial base kernel function and the one-against-one classification method, chose the best penalty coefficient C and kernel function parameter σ, and trained the classifier model with training set. Next, in accordance with training set used in the process of algorithm, the paper worked out a decision function via the taxonomic algorithm being assessed. Finally, depending on the accuracy of the decision function tested by test set, the paper got the different test results of training set and test set samples through different Fourier spectrum windows. The accuracy of the prediction is approximately in accordance with the result of Fourier spectrum data analysis of the images of the back of four categories of insects, which means the feasibility of the design of automatic identification for rice plant hoppers.Automatic identification equipment for rice plant hoppers chooses3×32-D Fourier spectrum window, i.e.9eigenvectors are used to describe the color and texture of insects’back, to construct classification model for insects on the basis of support vector machine. With this equipment, the recognition rate of rice plant hoppers reaches above90%. Hence we can come to the conclusion that this equipment can basically reflect the density of rice plant hoppers in fields. To explore the automatic identification technology for rice plant hoppers through the study on some key techniques, i.e. the digital image acquisition of insects in the paddy field, the description of the identification features of rice plant hoppers and the establishment of the classification model for insects, contributes to raising the level of automatic identification technology for rise plant hoppers.

Related Dissertations

  1. Research on Automatic Detection Algorithm for Substructure Distress of Highway Pavement Based on SVM,U418.6
  2. ISAR Imaging Simulation of Space Targets and Target Recognition Based on ISAR Images,TN957.52
  3. Research on Feature Extraction and Classification of Pulse Waveform for Cholecystitis and Nephrotic Syndrome Diagnosis,TP391.41
  4. Application of Q-Learning in the Content-Based Image Retrieval Technology,TP391.41
  5. Research on Image Recognition Algorithm in the Forest Fire Prevention System,TP391.41
  6. Research on Transductive Support Vector Machine and Its Application in Image Retrieval,TP391.41
  7. Research on Feature Extraction and Classification of Tongue Shape and Tooth-Marked Tongue in TCM Tongue Diagnosis,TP391.41
  8. Research on Visual Measurement for Spacecraft Rendezvous and Approach,TP391.41
  9. Research on the Image Real-Time Acquisition, Storage and Image Processing System,TP391.41
  10. Feature Extraction, Selection and Combination in Lipreading,TP391.41
  11. Multi-currency Notes Technology Research and Implementation,TP391.41
  12. The Research on Paper Currency Classification Method Based on Harr-Like Feature and Minimal Ball Including Samples,TP391.41
  13. Pavement Distress Recognition Based on Image,TP391.41
  14. Research on Visual Detection and Tracking of Mobile Robots,TP242.62
  15. Research on Fusion Algorithm of Hyper Spectral and High Spatial Resolution Remote Sensing Image,TP751
  16. Study on the Design and Biomechanical Behaviors of Individualized Artificial Knee Joint,R318.1
  17. An Approach for Identifying a Plant Resistance Gene Based on the Random Forest,Q943
  18. A Research on Determinating Soil Types and Soil Boundaries under Chinese Soil Taxonomy,S155
  19. Tobacco Diseases Auto-Recognition Research Based on Image Processing Technology,S435.72
  20. Research on Nondestructive Detection Technology for External Qualities of Papayas Based-on Vision,S667.9
  21. Research on Identification System of Cashmere and Wool Fiber,TS101.921

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile