Research And Realization Of Automated Fruit Fly Identification System-Bactrocera

The automatic identification technology of insects'digital images by computer is one of the hot spot today, is the key to solve the problem of inadequacy of artificially insect species identification ability. Aimed to carry on the automatic identification of insects by computer, this paper introduced its research background and the research status from domestic and abroad, discussed the main principle and method required for automatic recognition. On this basis, it designed and developed the automatic identification system of fruit fly's images, introduced its workflow, examined the system's recognition result under different conditions, such as image's illumination was uneven, images were distorted, specimen had been damaged and different sample size, and so on. Then, it looked forward to the system's prospects for development.The major contents of this paper include:First, this paper introduced the background, the significance and research situation of automatic identification for insects, as well as the automatic identification system's research content, working principle and methods, the key technology needed, etc.Secondly, it introduced the pattern recognition technology, such as the statistical pattern recognition, syntactic pattern recognition and fuzzy pattern recognition. And analyzed several methods which had realized the automatic identification for digital images by computer, such as the artificial neural network (including feed forward networks and supervised learning, competition and Lenovo networks and unsupervised learning, etc.), support vector machine, boost algorithm, etc, and compared the advantages and disadvantages of every methods. Aimed to solve some problems brought by conventional automatic identification methods applied in automatic identification systems of digital images, such as the dimension disaster problems of features, the bad result of classification for many kinds, etc, it improved the existing methods, and put forward a new recognition method, to meet the new requirement of identification.Again, it introduced several features representation methods for images, such as color, texture and local characteristics and so on, compared the advantages and disadvantages of every feature, and selected a valid method for images' feature extraction.Finally, in view of the problem of the automatic recognition for fruit flies' images, this paper designed and developed the automatic identification system of fruit fly, introduced the system's working modes from image acquisition, image cut and pretreatment, image feature extraction, classifier design, image recognition and identification results show. It carried out the analysis of the system's identification performance by experiment, to show that the new identification method was feasible. It added the backend database support module to the system, to display the species information identified by the system for users, so as to reflect the huge advantages of automatic identification technology by computer.