The 10-m cotton maps in Xinjiang, China during 2018-2021.

Cotton maps (10 m) of Xinjiang (XJ_COTTON10), which is the largest cotton production region of China, were produced from 2018 to 2021 through supervised classification. A two-step mapping strategy, i.e., cropland mapping followed by cotton extraction, was employed to improve the accuracy and efficiency of cotton mapping for a large region of about 1.66 million km2 with high heterogeneity. Additionally, the time-series satellite data related to spectral, textural, structural, and phenological features were combined and used in a supervised random forest classifier. The cotton/non-cotton classification model achieved overall accuracies of about 95% and 90% on the test samples of the same and adjacent years, respectively. The proposed two-step cotton mapping strategy proved promising and effective in producing multi-year and consistent cotton maps. XJ_COTTON10 agreed well with the statistical areas of cotton at the county level (R2 = 0.84-0.94). This is the first cotton mapping for the entire Xinjiang at 10-meter resolution, which can provide a basis for high-precision cotton monitoring and policymaking in China.

Integrating environmental and satellite data to estimate county-level cotton yield in Xinjiang Province.

Accurate and timely estimation of cotton yield over large areas is essential for precision agriculture, facilitating the operation of commodity markets and guiding agronomic management practices. Remote sensing (RS) and crop models are effective means to predict cotton yield in the field. The satellite vegetation indices (VIs) can describe crop yield variations over large areas but can't take the exact environmental impact into consideration. Climate variables (CVs), the result of the influence of spatial heterogeneity in large regions, can provide environmental information for better estimation of cotton yield. In this study, the most important VIs and CVs for estimating county-level cotton yield across Xinjiang Province were screened out. We found that the VIs of canopy structure and chlorophyll contents, and the CVs of moisture, were the most significant factors for cotton growth. For yield estimation, we utilized four approaches: least absolute shrinkage and selection operator regression (LASSO), support vector regression (SVR), random forest regression (RFR) and long short-term memory (LSTM). Due to its ability to capture temporal features over the long term, LSTM performed best, with an R2 of 0.76, root mean square error (RMSE) of 150 kg/ha and relative RMSE (rRMSE) of 8.67%; moreover, an additional 10% of the variance could be explained by adding CVs to the VIs. For the within-season yield estimation using LSTM, predictions made 2 months before harvest were the most accurate (R2 = 0.65, RMSE = 220 kg/ha, rRMSE = 15.97%). Our study demonstrated the feasibility of yield estimation and early prediction at the county level over large cotton cultivation areas by integrating satellite and environmental data.

Crop/Weed Discrimination Using a Field Imaging Spectrometer System.

Nowadays, sensors begin to play an essential role in smart-agriculture practices. Spectroscopy and the ground-based sensors have inspired widespread interest in the field of weed detection. Most studies focused on detection under ideal conditions, such as indoor or under artificial lighting, and more studies in the actual field environment are needed to test the applicability of this sensor technology. Meanwhile, hyperspectral image data collected by imaging spectrometer often has hundreds of channels and, thus, are large in size and highly redundant in information. Therefore, a key element in this application is to perform dimensionality reduction and feature extraction. However, the processing of highly dimensional spectral imaging data has not been given due attention in recent studies. In this study, a field imaging spectrometer system (FISS; 380-870 nm and 344 bands) was designed and used to discriminate carrot and three weed species (purslane, humifuse, and goosegrass) in the crop field. Dimensionality reduction was performed on the spectral data based on wavelet transform; the wavelet coefficients were extracted and used as the classification features in the weed detection model, and the results were compared with those obtained by using spectral bands as the classification feature. The classification features were selected using Wilks' statistic-based stepwise selection, and the results of Fisher linear discriminant analysis (LDA) and the highly dimensional data processing-oriented support vector machine (SVM) were compared. The results indicated that multiclass discrimination among weeds or between crops and weeds can be achieved using a limited number of spectral bands (8 bands) with an overall classification accuracy of greater than 85%. When the number of spectral bands increased to 15, the classification accuracy was improved to greater than 90%; further increasing the number of bands did not significantly improve the accuracy. Bands in the red edge region of plant spectra had strong discriminant capability. In terms of classification features, wavelet coefficients outperformed raw spectral bands when there were a limited number of variables. However, the difference between the two was minimal when the number of variables increased to a certain level. Among different discrimination methods, SVM, which is capable of nonlinear classification, performed better.

[Research on Spectral Scale Effect in the Estimation of Vegetation Leaf Chlorophyll Content].

Spectral indices (SIs) method has been widely applied in the prediction of vegetation biochemical parameters. Take the diversity of spectral response of different sensors into consideration, this study aimed at researching spectral scale effect of SIs for estimating vegetation chlorophyll content (VCC). The 5 nm leaf reflectance data under 16 levels of chlorophyll content was got by the radiation transfer model PROSPECT and then simulated to multiple bandwidths spectrum (10-35 nm), using Gaussian spectral response function. Firstly, the correlation between SIs and VCC was studied. And then the sensitivity of SIs to VCC and bandwidth were analyzed and compared. Lastly, 112 samples were selected to verify the results above mentioned. The results show that Vegetation Index Based on Universal Pattern Decomposition Method (VIUPD) is the best spectral index due to its high sensitivity to VCC but low sensitivity to bandwidth, and can be successfully used to estimate VCC with coefficient of determination R2 of 0.99 and RMSE of 3.52 µg x cm(-2). Followed by VIUPD, Normalized Difference Vegetation Index (NDVI) and Simple Ratio Index (SRI) presented a comparatively good performance for VCC estimation (R2 > 0.89) with their prediction value of chlorophyll content was lower than the true value. The worse accuracy of other indices were also tested. Results demonstrate that spectral scale effect must be well-considered when estimating chlorophyll content, using SIs method. VIUPD introduced in the present study has the best performance, which reaffirms its special feature of comparatively sensor-independent and illustrates its potential ability in the area of estimating vegetation biochemical parameters based on multiple satellite data.

A novel multi-digital camera system based on tilt-shift photography technology.

Multi-digital camera systems (MDCS) are constantly being improved to meet the increasing requirement of high-resolution spatial data. This study identifies the insufficiencies of traditional MDCSs and proposes a new category MDCS based on tilt-shift photography to improve ability of the MDCS to acquire high-accuracy spatial data. A prototype system, including two or four tilt-shift cameras (TSC, camera model: Nikon D90), is developed to validate the feasibility and correctness of proposed MDCS. Similar to the cameras of traditional MDCSs, calibration is also essential for TSC of new MDCS. The study constructs indoor control fields and proposes appropriate calibration methods for TSC, including digital distortion model (DDM) approach and two-step calibrated strategy. The characteristics of TSC are analyzed in detail via a calibration experiment; for example, the edge distortion of TSC. Finally, the ability of the new MDCS to acquire high-accuracy spatial data is verified through flight experiments. The results of flight experiments illustrate that geo-position accuracy of prototype system achieves 0.3 m at a flight height of 800 m, and spatial resolution of 0.15 m. In addition, results of the comparison between the traditional (MADC II) and proposed MDCS demonstrate that the latter (0.3 m) provides spatial data with higher accuracy than the former (only 0.6 m) under the same conditions. We also take the attitude that using higher accuracy TSC in the new MDCS should further improve the accuracy of the photogrammetry senior product.

[Monitoring spatio-temporal spectral characteristics of leaves of karst plant during dehydration using a field imaging spectrometer system].

As the supplement of spaceborne and airborne imaging spectrometer system, field Imaging spectrometer system spans a very broad range of applications. Imaging spectrometer system of this new kind could provide vital information especially for which spaceborne or airborne remote sensing could not be competent, such as proximal detection of plant population, individual plant or plant organs for site-specific management in precision agriculture. A new self-developed imaging spectrometer system was utilized to monitor spatio-temporal dynamics of spectral changes of plant leaves in response to dehydration. lThe phenomenon of blue shift of red edge of plant leaves was successfully detected and visualized in the form of image series. The patterns of photochemical reflectance index (PRI) of leaves during dehydration were compared and confirmed by fluorescence parameter quantum yield. Our results show that FISS has good spectral and radiometric properties and could be used in quantitative researches and precise information mapping.

[Quantitative inversion of rock SiO2 content based on thermal infrared emissivity spectrum].

The present paper used the emissivity of non-processed rocks measured by M304, a hyperspectral Fourier transform infrared (FTIR) spectroradiometer, and SiO2 content by the X-ray fluorescence spectrometry. After continuum removal and normalization, stepwise regress method was employed to select the feature bands of rocks emissivity. And then quantitative relationship between SiO2 content and continuum removal emissivity of feature bands was analysed. Based on that, by comparing twelve SiO2 indices models, the optimal model for predicting SiO2 content was built. The result showed that the SiO2 indices can predict SiO2 content efficiently, and especially the normalization silicon dioxide index (NSDI) about 11.18 and 12.36 microm is the best; compared with regression models, NSDI is simpler and has higher practicality; the result has an important application value in rock classification and SiO2 content extraction with high precision.

Laboratory calibration of a field imaging spectrometer system.

A new Field Imaging Spectrometer System (FISS) based on a cooling area CCD was developed. This paper describes the imaging principle, structural design, and main parameters of the FISS sensor. The FISS was spectrally calibrated with a double grating monochromator to determine the center wavelength and FWHM of each band. Calibration results showed that the spectral range of the FISS system is 437-902 nm, the number of channels is 344 and the spectral resolution of each channel is better than 5 nm. An integrating sphere was used to achieve absolute radiometric calibration of the FISS with less than 5% calibration error for each band. There are 215 channels with signal to noise ratios (SNRs) greater than 500 (62.5% of the bands). The results demonstrated that the FISS has achieved high performance that assures the feasibility of its practical use in various fields.

[Fresh and frozen-thawed meat discrimination based on FISS imaging spectral data].

In the present paper, a self-developed Field imaging spectrometer system (FISS) was used to detect whether pork has been frozen and thawed. The preservation time of fresh pork has also been identified. Fresh and frozen-thawed pork was scanned and imaged and hyperspectral image cubes were acquired using FISS. To eliminate high-frequency random noise and baseline offset and improve the multi-collinearity, all samples were preprocessed by MNF (Minimum noise fraction) transform and first derivative. Multiple analysis models were built by using Wilks' lambda stepwise method to select proper wavelengths. Fisher LDA (linear discriminant analysis) was performed to discriminate fresh and frozen-thawed pork. Eight selected bands gave 99% correct results of fresh or frozen-thawed pork samples. For the freshness by the day, classification accuracy reached 98% with 6 selected bands, while for the freshness by the hour, classification accuracy reached 93.6% with all 28 selected bands. The results showed that FISS might be used as a screening method to identify the quality of meat.

[Study on discrimination of varieties of milk based on FISS imaging spectral data].

Using a self-developed field imaging spectrometer system (FISS), hyperspectral images of 14 typical kinds of milk were acquired, based on which the discrimination of varieties of milk was studied. Firstly, removing 2 abnormal samples, the remaining 12 kinds of milk were randomly sampled, a total of 1 200 pixel samples. To eliminating high-frequency random noises and baseline offset and decrease the multi-collinearity, all samples were preprocessed by smooth-moving average and first derivative. Secondly, multiple discriminant analysis models for milk were built using characteristic wavelengths selected by the stepwise method. Results demonstrated that the overall identification accuracy for 1 200 spectral samples put together reached 95.5%, of which the overall distinguishing rate of Mengniu, Yili and Guangming acidophilous milk was 88.3%. The discriminant models for the three kinds of acidophilous milk subset, 300 spectral samples in all, were built, with the overall distinguishing rate of 88.7%. This explicated that FISS would be useful for discriminating milk varieties, and to accomplish specific discrimination of milk varieties, it would be best for milk of the same type from different manufacturers to form a subset, which may not only reduce the model variables, improving operational efficiency and the stability of the model, but improve their overall discriminant accuracy.

[Study on the relation between fraction cover and mixed spectra in karst environment].

Karst rocky desertification is one of the most serious eco-environmental problems of land degradation in karst regions, southwest China. The fractional cover of vegetation and exposed bedrock are the main land surface symptoms and essential assessing indicators of karst rocky desertification. To assess the extent of rocky desertification in complex karst environments, the information of multiple land cover types fraction is needed. Based on in situ spectral reflectance data, this study proposed several spectral indices and explored the relationship between spectral features of main land cover types and their responding fractional cover. The results showed that spectral indices have much higher correlation coefficients with fractional cover than does spectral reflectance. Vegetation indices have good linear relation with fractional cover of photosynthetic vegetation (PV). The proposed spectral indices have high correlation coefficients with fractional cover of non-photosynthetic vegetation (NPV) and bare soil, with R2 0.70 and 0.73, respectively. Lower correlation coefficients (R2 = 0.55) with the factional cover of exposed bedrock, were observed. The absorption depth of four forms of the proposed indices has the highest correlation coefficient with the fractional cover of NPV, bare soil, and exposed rock. This study indicates that hyperspectral remote sensing has the potential for the extraction of karst rocky desertification information.

[Research on crop-weed discrimination using a field imaging spectrometer].

Discrimination of weeds from crop is the first and important step for variable herbicides application and precise physical weed control. Using a new field imaging spectrometer developed by our group, hyperspectral images in the wavelength range 380-870 nm were taken in the wild for the investigation of crop-weed discrimination. After normalizing the data to reduce or eliminate the influence of varying illuminance, stepwise forward variable selection was employed to select the proper band sets and fisher linear discriminant analysis (LDA) was performed to discriminate crop and weeds. For the case of considering each species as a different class, classification accuracy reached 85% with eight selected bands while for the case of considering overall weed species as a class, classification accuracy was higher than 91% with seven selected bands. In order to develop a low-cost device and system in future, all combinations of two and three bands were evaluated to find the best combinations. The result showed that the best three bands can achieve a performance of 89% comparable to the performance achieved by five bands selected using stepwise selection. The authors also found that "red edge" could afford abundant information in the discrimination of weed and crop.

Simulation of EO-1 Hyperion Data from ALI Multispectral Data Based on the Spectral Reconstruction Approach.

Data simulation is widely used in remote sensing to produce imagery for a new sensor in the design stage, for scale issues of some special applications, or for testing of novel algorithms. Hyperspectral data could provide more abundant information than traditional multispectral data and thus greatly extend the range of remote sensing applications. Unfortunately, hyperspectral data are much more difficult and expensive to acquire and were not available prior to the development of operational hyperspectral instruments, while large amounts of accumulated multispectral data have been collected around the world over the past several decades. Therefore, it is reasonable to examine means of using these multispectral data to simulate or construct hyperspectral data, especially in situations where hyperspectral data are necessary but hard to acquire. Here, a method based on spectral reconstruction is proposed to simulate hyperspectral data (Hyperion data) from multispectral Advanced Land Imager data (ALI data). This method involves extraction of the inherent information of source data and reassignment to newly simulated data. A total of 106 bands of Hyperion data were simulated from ALI data covering the same area. To evaluate this method, we compare the simulated and original Hyperion data by visual interpretation, statistical comparison, and classification. The results generally showed good performance of this method and indicated that most bands were well simulated, and the information both preserved and presented well. This makes it possible to simulate hyperspectral data from multispectral data for testing the performance of algorithms, extend the use of multispectral data and help the design of a virtual sensor.