Dissertation > Excellent graduate degree dissertation topics show

Measurement Methods of Forewing Shape Characteristics for Ricehopper

Author: ZhaoSanQin
Tutor: DingWeiMin
School: Nanjing Agricultural College
Course: Agricultural Mechanization Engineering
Keywords: Boundary points form deviation boundary reconstruction Fourierdescriptors veins watershed transform
CLC: TP391.41
Type: PhD thesis
Year: 2010
Downloads: 0
Quote: 3
Read: Download Dissertation


Accurate identification of insect species was research basis of other disciplines, including insect ecology, morphology, physiology, biochemistry, behavior, toxicology and applied entomology (such as agricultural entomology, forest entomology, medical entomology). Inaccurate insect species identification could lead to the loss of research objectivity, comparability and repeatability, thus losing scientific value. Currently, insect species identification was accomplished by few experienced experts. About80%of Earth’s species were yet unknown because of many insects. The actual demand of insect species identification was far more than expert burden. To reduce the burden of routine identification and accelerate taxonomic discovery, insect automatic identification technology had been an active research area. Characteristics measurement and extraction was an essential step to implement insect automatic identification. For most insects, they had morphology identification characteristics. More importantly, insect wings were important taxonomic significance and easy to describe, so wings characteristics extraction software was devised greatly. However, insect species were more than100million, and each species was high morphological diversity. It was almost impossible to achieve all species identification only using several programs or an algorithm. Consequently, some new methods were developed necessarily according to different insect species.Ricehopper, including laodelphax striatellus, nilaparvata lugens and sogatella furcifera, was major rice pests. The ricehopper identification had always been a key for pest predicting. However, ricehopper species identification was completed mainly by professional classifiers, who observed the vertex shape and the color of frontal, buccal and mesonotum by a magnifying glass, dissecting microscope and the microscope, and then compared these characteristics with the keys to determine the species. This identification method was inefficient and dependent greatly on experts. Predicting accuracy and timeliness was seriously affected.To improve the efficiency of ricehopper identification, integrating the advantages of image processing technolody, such as automatic, rapidity and repeatability measurements, measurement methods of forewing shape characteristics were further investigated to discuss whether wing shape could be discarded as ricehopper identification characteristics. The main contents and conclusions were as follows:(1) Using Fourier descriptors to identify or retrieve shapes, the number of boundary points was the only uncertain parameter. To select suitable boundary points, relationship between boundary points and accuracy of Fourier descriptors was tested.Taking circular boundary as an example, the boundary was constructed by circle trigonometric formula. The number of boundary points was controlled by adjusting the sampling interval of central angle. Complex Fourier descriptors and shape similarity were introduced briefly. To establish relationship between boundary points and accuracy of Fourier descriptors, shape similarity was tested. The results showed the more boundry points were, the higher the accuracy of Fourier descriptors was. When the boundary points were greater than64, the accuracy of Fourier descriptors was little change. Finally, circular binary images were used to verify the results correctness, and the accuracy of Fourier descriptors was also related to the accuracy of image collection system. Under the same image collection system, this conclusion was applied to determine the boundary points before calculating the boundary Fourier descriptors.(2) Ricehopper samples preparation, image acquisition and shape normalization.Ten samples for each species were selected. To ensure the wing surface clean and integrity, the dust or dirt on the wing surface were washed away with95%alcohol, and all samples were marked separately. The forewing of each individual was detached from the thorax, fixed on a glass slide with few drops of glue, and secured under a cover slip. Wing images were captured using stereomicroscope and were saved in BMP (Windows bitmap) format. To prepare for comparative study of Fourier descriptors, forewing boundary were extracted and normalized to make all samples be the same boundary position, orientation and size. Meanwhile, all boundary points were re-sampled to64.(3) To investigate whether wing boundary was regarded as identification characteristics for ricehopper, polar radius Fourier descriptors, complex Fourier descriptors and elliptical Fourier descriptors were further compared, and reconstruction accuracy and recognition accuracy were tested.Three Fourier descriptors were briefly introduced. The number of low-frequency coefficients used to reconstruct forewing boundary was obtained by comparing the reconstruction accuracy of three Fourier descriptors. Low-frequency coefficients and selected coefficients were respectively as characteristics to calculate the recognition accuracy based on hierarchical clustering algorithm. The results showed15low frequency coefficients of each Fourier descriptors could be used to reconstruct completely the forewing boundary. Under the same reconstruction accuracy, the requiring number of three descriptors was in order of elliptical descriptors, complex descriptors and polar radius descriptors. Recognition rate of elliptical descriptors was higher, while recognition rate of complex descriptors and polar radius descriptors was almost the same. However, recognition rates of three descriptors were all not high. When low frequency coefficients were used as characteristics, it was difficult to identify ricehopper samples. Taking selected coefficients as characteristics, clustering results of three Fourier descriptors were the same, the forewing characteristics of laodelphax striatellus were not stable, while shape difference between nilaparvata lugens and sogatella furcifera was larger. Because forewing boundaries of ricehopper were very similar, the recognition accuracy was lower only taking forewing boundaries as characteristics.(4) To investigate whether wing venation characteristics were regarded as identification characteristics for ricehopper, extraction method for venation characteristics was studied.The watershed algorithm based on distance transform was introduced. The venation skeleton was extracted by the above method. To compare with the existing study, venation skeleton was also extracted using a thinning algorithm. Finally, all wing images were used to assess the repeatability and reliability of the methods. The result showed the skeletons generated with the watershed transform algorithm were smoother, and the number of junctions was more stable. The proposed method was not limited by the quality and method of capturing of the wing images. Wing venation characteristics of Drosophila and honeybee could also been extracted effectively using this method.(5) On the basis of extraction the coordinates of vein junctions, venation characteristics whether to use for ricehopper identification, the vein junction distances were analyzed adopting Fisher stepwise discriminant.First, vein junction distances were obtained.16vein distances were analyzed by Fisher stepwise discriminant to select significant difference distance, and thus establishing identification model of ricehopper. To test the reliability of distances and compare the differences of different methods, vein junction distances were obtained using world-famous TPSDig software and were also analyzed by Fisher stepwise discriminant. The result showed the vein distances obtained by the watershed algorithm were more reliable. The vein distances of3,4,5,6,12were significant differences. Taking these distances as characteristics, recognition rate was83.33%. Compared with the boundary Fourier descriptors, the recognition accuracy of the wing venation characteristics was higher. Consequently, it was feasible that ricehopper identification could be accomplished using wing venation characteristics.

Related Dissertations

  1. Analysis on the Surface Structure of ACF Micro-Image,TP391.41
  2. Application of Intraoperative Indocyanine Green Angiography for Giant Parasagittal Meningioma,R739.45
  3. Research of Infrared Image Segmentation for Power Equipment in Substation,TP391.41
  4. West Tianshan high-pressure mineral veins and host rocks typical electron microprobe analysis and fluid Tracing,P575.1
  5. Minimally invasive surgical treatment of varicose veins combined with traditional Chinese medicine clinical research,R654.4
  6. Qi Yang Xin medicine in the treatment of chronic systolic heart failure clinical study,R259
  7. Research and Implementation of CAD on the Pathological Slices of Gastric Adenocarcinoma,TP391.41
  8. Study on Seal Matching and Identification,TP391.41
  9. Parameter grid clustering algorithm,TP311.13
  10. Edge Points Based Feature Extraction Method for Medical Image Classification,TP391.41
  11. Nelin, the New Gene’s Expression and Roles in Varicose Veins,R543.6
  12. Ophthalmic Blood from the Veins and Organs of the Relationship between the Eye,R276.7
  13. The Study of Boundary Detecting Algorithm Based on Joint Entropy,TP311.13
  14. The Research of Nonparametric Clustering Boundary Detection Algorithm,TP311.13
  15. Research on Investigation Technical System of Rural Housing Land in Hubei Province,P272
  16. Endoscopic Anatomy and Its Signification on Outflow Segment of the Cerebral Bridging Veins into the Superior Sagittal Sinus,R322
  17. A Comparative Study to the Female Tragic Consciousness in the Novels of Zhang Ai Ling and Cui Zhen Xi,I207.42
  18. Image segmentation algorithm based on mathematical morphology studies,TP391.41
  19. Clinical Studies on EVLT Combined with Operation for Venous Varicosis of Lower Limbs,R654.4
  20. Angiography of Superficial Cerebral Veins with the 64 Detector Row Helical CT,R816.1
  21. A New Model of Unilateral Cerebral Cortical Draining Vein Occlusion in the Rabbits and Preliminary Observation of Drug Intervention,R743.9

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