Gao J.,Zhejiang University |
Li X.,Zhejiang University |
Zhu F.,Zhejiang University |
He Y.,Zhejiang University |
He Y.,Key Laboratory of Equipment
Computers and Electronics in Agriculture | Year: 2013
A rapid and non-invasive method was investigated to identify the geographical origin of Jatropha curcas L. seeds in China by using near-infrared hyperspectral imaging technique on the wavelength between 874 and 1734. nm. Two hundred and forty J. curcas L. seed samples from four different geographical origins (Jiangsu, Sichuan, Hainan and Taiwan) in China were studied and all of them were scanned by a pushbroom hyperspectral imaging system. Then the obtained data sets were analyzed by spectral and image processing technique respectively. Successive projections algorithm (SPA) was used for selecting effective wavelengths. Dimension reduction was carried out on the region of interest (ROI) image by principal component analysis (PCA). The first principal component (PC) explained over 92% of variances of all spectral bands. Gray-level co-occurrence matrix (GLCM) analysis was implemented on the principal component (PC) image to extract 5 textural feature variables in total. Moreover, 7 morphological features of samples were computed additionally. Then least squares-support vector machine (LS-SVM) classification models were built based on the extracted spectral, textural, morphological, combined spectral and textural, combined spectral and morphological, combined textural and morphological, combined spectral, textural and morphological features, respectively. The satisfactory results show the correct discrimination rate of 93.75% for the prediction samples based on spectral and morphological features. The study demonstrated that hyperspectral image technique can be a reliable tool for discriminating different geographical origins of J. curcas L. seeds. The above results indicated that this objective and non-destructive method can be utilized for quality control purposes and seed breeding in future. © 2013 Elsevier B.V.
Xiang R.,Zhejiang University |
Xiang R.,China Jiliang University |
Jiang H.,Zhejiang University |
Jiang H.,Key Laboratory of Equipment |
And 2 more authors.
Computers and Electronics in Agriculture | Year: 2014
To improve the applicability of the recognition method for clustered tomatoes, an algorithm based on binocular stereo vision was presented. First, a depth map of clustered tomatoes was acquired using a combination stereo matching method. Second, the noises in the depth map were removed using an eight-neighbor mode denoising method. Third, the clustered regions were classified into two types (i.e., overlapping and adhering regions) based on the depth difference between the front and back regions in a clustered region using an iterative Otsu method. Finally, different recognition methods were used for different types of clustered tomatoes. For adhering tomatoes, a recognition method based on edge curvature analysis was used for the edges in color image. For overlapping tomatoes, the same method was applied for the edges in color image, which were segmented into several parts by the edges in depth map after segmentation. A total of 189 pairs of stereo images were tested, and the recognition accuracy rate of clustered tomatoes was 87.9% when the leaf or branch occlusion rate was less than 25%. The acquisition distance and average execution time of this method were 300-500. mm and approximately 0.5. s, respectively. In conclusion, this method can realize the recognition of the clustered types and different types of clustered tomatoes, despite the serious occlusion of other tomatoes. Moreover, the headmost tomato in clustered tomatoes can be recognized based on depth information. This method can also realize the recognition of clustered tomatoes based on the images taken at different distances. However, the success rate of clustered tomatoes was not satisfactory when the occlusion was serious. Further research should focus on the improvement of the accuracy of stereo matching and the recognition of tomatoes occluded by leaves. © 2014 Elsevier B.V.
Zhang X.,Zhejiang University |
Zhang X.,Key Laboratory of Equipment |
Liu F.,Zhejiang University |
Liu F.,Key Laboratory of Equipment |
And 3 more authors.
Biosystems Engineering | Year: 2013
This study was carried out to investigate the potential of visible and near infrared (VIS-NIR) hyperspectral imaging system for rapid and non-destructive content determination and distribution estimation of nitrogen (N), phosphorus (P) and potassium (K) in oilseed rape leaves. Hyperspectral images of 140 leaf samples were acquired in the wavelength range of 380-1030 nm and their spectral data were extracted from the region of interest (ROI). Partial least square regression (PLSR) and least-squares support vector machines (LS-SVM) were applied to relate the nutrient content to the corresponding spectral data and reasonable estimation results were obtained. The regression coefficients of the resulted PLSR models with full range spectra were used to identify the effective wavelengths and reduce the high dimensionality of the hyperspectral data. LS-SVM model for N with RP of 0.882, LS-SVM model for P with RP of 0.710, and PLSR model for K with RP of 0.746 respectively got the best prediction performance for the determination of the content of these three macronutrients based on the effective wavelengths. Distribution maps of N, P and K content in rape leaves were generated by applying the optimal calibration models in each pixel of reduced hyperspectral images. The different colours represented indicated the change of nutrient content in the leaves under different fertiliser treatments. The results revealed that hyperspectral imaging is a promising technique to detect macronutrients within oilseed rape leaves non-destructively and could be applied to in situ detection in living plants. © 2013 IAgrE.
Jiang Z.,Zhejiang University |
Zhou M.,Zhejiang University |
Tong J.,Zhejiang Sci-Tech University |
Jiang H.,Zhejiang University |
And 4 more authors.
Computers and Electronics in Agriculture | Year: 2015
The important task of replugging bad or missing cells with healthy seedlings in greenhouses is carried out by automatic transplanters. Grippers of such transplanters spend a considerable amount of time shuttling between the source and target trays during replugging. Therefore, work efficiency of transplanters can be significantly improved by tour planning. In this study, performances of the ant colony algorithm (ACA), the genetic algorithm (GA), and the common sequence method (CSM) in replugging tour planning were compared. Two types of seedling trays, with 50 and 200 cells, were used. The ACA and the GA were found to have more advantages than the CSM in total tour lengths for one tray. Moreover, the ACA performed better than the GA when the numbers of empty cells and healthy seedlings in the target and source trays, respectively, increased. When a 20. ×. 10 tray was used, the average length of the ACA decreased by 6000.9. mm compared with that of the GA and by 13058.4. mm compared with that of the CSM after finishing 40 empty cells in one tray. The average run times of the GA and the ACA in MATLAB (R2012a) were 0.32 and 0.94. s, respectively. These results meet real-time operation requirements. © 2015 Elsevier B.V.