The purpose of this research is to develop a laboratory-based near-infrared (NIR) hyperspectral imaging system to measure the two-dimensional distribution of nitrate concentrations in a vegetable leaf as a tool for precisely analysing nitrate metabolism. Komatsuna leaves were analyzed by hyperspectral reflectance in the range from 607 to 967 nm with a resolution of 9 nm. The reflectance was standardized by a reference plate and was converted to relative reflectance. An algorithm to select the effective wavelength to predict the nitrate concentration was developed in conjunction with partial least squares (PLS) regression and principal components regression (PCR). As for preprocessing methods for the spectra, mean-centre and standard normal variate transformations were examined. Estimation accuracy of the developed models was evaluated by the weighted average of standard error (WSE). The estimation accuracy of the wavelength-selected models was improved and the WSE was smaller than that of the full-spectrum model (41 wavelengths). The calibration model that used 21 wavelengths achieved the best WSE of 1446 ppm with a correlation coefficient of 0.870. The nitrate distribution in Komatsuna leaves were visualized in digital images with a spatial resolution of 2.5×10−4 mm/pixel. These images showed that the transporting route of nutrients contains higher nitrate ion concentration than other areas in the Komatsuna leaf.
The objective of this study was to develop a reliable field monitoring system combining helicopter-based and satellite-based remote sensing. In this research, multi-spectral imageries were used that was taken by SPOT5, QuickBird-2 and helicopter. These imageries were synchronously taken. As every vision covers the same wavelength range and takes same objects, the color information of the satellite and helicopter images would be integrated. The relation of the information on NDVI was examined. There was a high correlation between satellite-based NDVI and helicopter-based NDVI. The R2 value was 0.81 between helicopter-based NDVI and Quickbird-2 satellite-based NDVI. Therefore, it was possible to do revision of the color information on the relative satellite imageries in using helicopter imageries. Using these color information, the field status (moisture content of wheat ear) can be estimated. Combined Helicopter-based NDVI and satellite-based NDVI, it became possible to estimate the wide range in high accuracy and high efficiency.
Stereo cameras have been used as perception sensors for agricultural vehicle navigation for years. One problem impeding their broader application is the difficulty of calibrating the installation poses of a camera using conventional measuring tools, especially when such a system is used in ill-structured agricultural field environments. The research reported in this paper was aimed to develop an automated calibration method for determining the camera installation pose with respect to a vehicle frame. Based on this method, a binocular stereo camera acquired a sequence of field scenery images as the vehicle moved straight forward for a short distance on a relatively flat surface. An image processing algorithm has been developed to detect some static feature points in the ground image and track their three-dimensional (3D) relative motions with respect to the moving vehicle. A plane best fitting to those detected ground features was then used to determine the camera roll and pitch angles, and the tracked motions of those feature points were used to estimate the camera yaw. Field test results validated that the developed auto-calibration method was capable of determining the camera installation pose at a calibration accuracy of ±1° over an approximately 10 m of vehicle traveling distance. The calibrated poses could be used to compensate for the navigation errors induced by camera misalignment.
The values of Normalized Difference Vegetation Index (NDVI) are directly related to the capacity of plant photosynthesis and important to crop growth. In this research, an attempt was performed to develop a real-time NDVI measurement device that eliminated the noise such as shadows and background soil. The developed NDVI sensor has the advantage of portability and the capability of real-time measurement of plant growth. The sensor has flexibility in operation, either manually or remotely. The system can be used with several interfaces to enable the communication using on-board Wi-Fi for image acquisition and NDVI measurement. Thus, the vision of this system is to develop an NDVI measurement sensor for using in field monitoring. The NDVI sensor could be manufactured commercially in a low-cost and would be convenient compare to other expensive similar commercial products.
Over the past two decades, a number of researchers around the world tried to develop a citrus harvesting robot. However, no commercial harvesting robot is yet available in the market. Both technical and economic factors have hindered the commercialization of robotic harvesting. Fruit detection in the orchard under natural daylight condition is still a challenging a task. This paper presents a study of using multispectral imaging to enhance citrus fruit detection in the field under natural daylight condition. The multispectral imaging is composed of a 12-bit monochrome camera fitted with a filter wheel which carries six optical band pass filters covering the spectrum identified to have a high discriminability between orange fruits and leaves. Multispectral images of mature orange fruit targets were acquired in the field under natural lighting condition. Pattern recognition techniques such as linear discriminant classifier and artificial neural networks were developed to segment the fruit by classifying the fruit pixels from the background pixels. A modified watershed transform combined with blob analysis was used to detect the individual fruits in a cluster. In addition, principal component analysis (PCA) was used to transform the multispectral images and to identify the wavelengths or combination of wavelengths that could improve detection of fruit from the canopy background.
The overall goal of this study was to develop a low-cost machine vision system for detecting green fruit on citrus trees. It is difficult to distinguish green citrus fruit from the background leaves using color information because the leaves and fruits are similar in color. Therefore, spatial characteristics such as shape and texture were analyzed. After background extraction, edge detection was performed. Then, circle-template matching was applied to the edge image to detect green fruits. The detection tests revealed that 86% of the fruit were correctly detected.