Study On Extraction And Classification Of Characteristic Band Of Crucifera Crops Based On LCTF Imaging Technique

China is an agricultural country. The quantity and quality of crops were seriously affected by plant diseases and pesticide residues. Therefore, it is meaningful for China’s precision agriculture to find an efficient way to distinguish the agriculture accurately. As a group of daily eaten and nutrient-rich vegetables, cruciferous vegetables were widely planted all over China. Thus, recognizing the information of crops has great significance for take precautions against plant diseases and pesticide residues. Both the spectrometer and the imaging techniques based on vision and RGB can not obtain the target’s spectral information and space information in the mean time. Besides, multi-band imaging and great volume of data led to low rate of accuracy. However, as multi-band imaging technology step forward in recent years, the method has been widely adopted in precision agriculture, and its image-spectrum feature appears to be extremely important in the process of acquiring and handling data information. Consequently, the aim of this thesis is to study extraction and classification of several crucifers’ wave band feature, and provide theoretical support for planting crucifers.First of all, according to imaging spectrometry, this thesis put forward a multi-spectral imaging system which consists of core image device CMOS and LCTF. And also, this thesis introduced the operating principle, multi-spectral data of the system.Second, Using self-built LCTF imaging system to collect multi-spectral images from 5 kinds of crucifers every 5nm in the range of 435 nm to 720 nm. Adopting ABS algorithm and band index method for characteristic bands selection from multi-spectral grayscale images. By analyzing the plants’ feature, correlation among wave bands and different date volume, the experiment makes it clear that the valid wave band of health pakchoi leaves, health cabbages’ leaves, health celery cabbage’s leaves, health wild cabbages’ leaves and health radish’s leaves are ideal. Therefore, using characteristic wave band method to extract spectral information from five kinds of crucifers is advisable.At last, according to Euclidean Distance method and Mahalanobis Distance method, spectral angle method and correlation coefficient method, the researchers did a classification accuracy analysis to each crop’s full-wave band and characteristic wave band. The experiment indicates that all the four classification method can reach 97.5 percent in classifying all-wave band. However, none of the four classification methods can achieve high accuracy in classifying characteristic wave band, it can only be achieved by combining all the four methods. Consequently, the experiment would build the corresponding multi-spectral database of these five crucifers’ characteristic wave band for ideally classification crops’ samples.The experiment result expresses that the system which using ABS algorithm and wave band index method on the basis of LCTF imaging techniques might support a new method and idea for discerning crucifers. This kind of system could manage multi-spectral imaging data more efficient and make fewer mistakes while providing ideal data support and theory support for distinguishing crucifers rapidly and effectively.