Search

In tests for field resistance of potato (Solanum tuberosum L.) to late blight, crop scientists rate the disease severity exclusively using visual examinations of infections on the leaves. However, this visual assessment is generally time-consuming and quite subjective. The objective of this study was to develop a new estimation technique for disease severity in a field using RGB imagery from an unmanned aerial vehicle (UAV). For the assessment of disease resistance of potatoes a test field was designed that consisted of 262 experimental plots on which various cultivars and lines were planted. From mid-July to mid-August in 2012, conventional visual assessment of disease severity was conducted while 11 aerial images of the field were obtained. The disease severity was estimated using an image processing protocol developed in this study. This estimation method was established so that the error of the severity estimated by image processing was minimal when compared with the visual assessment. Comparing the area under the disease progress curves (AUDPCs) calculated from the visual assessment and time series of images, the coefficient of determination was 0.77. A further experiment was conducted to validate the developed method. Eleven images of a field planted the following year were taken, and the resulting coefficient of determination was 0.73. The breeders concluded that these correlations were acceptable and that the UAV image acquisition and the disease severity estimation from the image were more efficient than the conventional visual assessments. Therefore, the developed technique based on aerial imagery allows high throughput, objective, and precise phenotyping with regard to field resistance to potato late blight.

A robot combine harvester for beans was developed using CAN Bus network. Two sensors are connected to the CAN Bus; a RTK-GPS to locate the position, and a GPS Compass to detect the heading direction. This robot is programmed to perform autonomous operation such as moving forward, backward, and turning in right angle. When the robot was tested in harvesting soybeans on a 100m × 30m rectangular field, the deviations between the target path and the robot's traveling path was approximately 0.07m RMS, and the yaw angle error was approximately 1.82° RMS along the long side of the field. Therefore this robot can work autonomously along the target path without overlapping and shortening. In addition, this robot can unload harvested grain to adjacent transport truck during harvesting operation, which improved harvesting capacity to approximately 10%.

The objective of this research is to realize fully robot-operated farming from tillage to harvest in large-scale agriculture such as cultivation of rice, wheat and soybean in Japan. For this purpose, three types of robot have been developed; the first is robot tractor, the second are rice transplanter robots, the third are combine harvester robots. RTK-GPS and inertia measurement unit (IMU), or GPS compass, are used for navigation system. Robots have Controller Area Network (CAN) bus that all sensors and PC can be connected in common among other robots such as tractors, rice transplanters, and combine harvesters. These robots could be done autonomous operation in paddy field. In addition, moving between the fields for effective operations and safe guideline for robot system were discussed.

In this paper, development and standardization of on-vehicle control and communication data network for small size farm machinery is described. There already exist an international standard and many compliant products in the western countries. But, they are too huge and complicated for the average Japanese style of farming, which is represented by the words of small tractors and small paddy fields. In order to achieve an advancement and cost reduction in this field, enough and sufficient low cost simple electronic devices and related software were developed. The communication method adopted basically depends on the ISO 11783 standard technology, and some new features needed for the domestic agricultural conditions are considered. At the same time, a new domestic simple standard for data exchange between ECU's on tractors and attached implements are discussed. From some experiments, it became clear that the developed hardware and software together are strong tools to realize the domestic standardized devices. It can be said that standardization of data communication for lower range of farm machinery just has been started.

We have developed a new Global Positioning System (GPS) guided rice transplanter. We developed GPS and IMU guided rice transplanter in previous researches. Those were guided with GPS position data and inertia measurement unit (IMU) direction data. New one is guided with GPS position data with tilt correction during straight driving and guided with IMU direction data during turning at the headland. The GPS antenna is set to 1.5m height and 0.4m front offset from vehicle front axle. The actuator control command and data communication protocols comply with through the controller area network (CAN) bus. Steering and transmission are controlled by electrical actuators according to the location in a field. We report results of experiment.

Phenomics research, which involves the determination of the relationship between phenotypes and genotypes, is constrained by the difficulties in phenotyping. Typically, the traits or characteristics of plants are manually measured or visually evaluated by humans for purposes such as, breeding and crop science. Therefore, general phenotyping is laborious, expensive, and subjective, whereas genotyping is readily conducted via DNA-sequencing techniques. The objective of this study was to develop a high-throughput phenotyping system that used field imagery from an unmanned aerial vehicle (UAV). The UAV was a multi-rotor drone that flew along a path preset by a global positioning system (GPS)-based navigation system. During the flight, a camera mounted on the UAV captured images of the field every 2 s. Because diseases on the leaves could be detected by image processing, disease severity was assessed via the aerial images of a potato field for breeding purposes. The areas under disease progress curves (AUDPCs) from the aerial images and conventional visual assessment were found to have a correlation; the coefficient of determination was R2 = 0.77. In the large test field for crop studies, the aerial images were assembled, and a composite image of the entire field was generated. Furthermore, a three-dimensional (3D) model of the field was generated as a point cloud with 3D reconstruction techniques called structure from motion and semi-global dense matching. With these techniques, the physical size of the plants in the field, which is one of the growth parameters, was retrievable as digital information.

We propose a system that is able to detect vehicle speed and yaw angle based on vision sensors. In rice fields in Japan, rice is transplanted in an ordered grid pattern. We have already developed a vision system able to estimate plant positions with high precision based on this lattice layout. In this study, the inclination to the crop rows of vehicles (Yaw angle) was calculated from the crop position information acquired using our vision system. Next, two time lag images were compared and the speed of the vehicles was computed by tracing the motion of the same plant. Furthermore, the tracks of vehicles were calculated using these two sets of data. As a result of comparing the computed tracks with the position information obtained by a RTKGPS, when it was assumed that the RTK-GPS’s error was 0, the vehicle’s position was detectable within 30mm error.

We are engaged in research about autonomous rice field operation. We have developed an automated tractor, a rice transplanter and a combine harvester. These automated agricultural machineries are guided by a global positioning system (GPS) and a direction sensor such as inertia measurement unit (IMU) using the controller area network (CAN) bus. Because each machine is used for a limited period of time in a year, we plan to share a main control computer and navigation sensors among automated machineries. We expect to deliver autonomous operation in paddy field at lower cost and to shorten the development period of each machine. The actuator control command and data communication protocols comply with the ISO 11783. In this paper, we report about project outline and explain about a rice transplanter as an example of automated agricultural machinery.

Recently, tomato production in large-sized greenhouse facilities has been popular in Japan. In large-scale greenhouse, however, harvesting tomatoes by hand is tough work. Therefore, robotic agricultural system has been addressed. In order to promote robotic agricultural system, it is important to develop the machine vision system.

The target of our research is to develop a technology for detecting vegetable infected by a disease at an early stage. We developed the experimental setup to capture hyper-spectral images of lettuce and attempted to detect the disease-infected area before the stage that human can easily recognize it. This experimental setup was composed of a hyper-spectral camera, a linear slider, a personal computer (OS:federa core 3), a signal conversion processor (PIC16f873), and sixteen Halogen lamps. This system makes it possible to make 2D spectral image, at regular intervals of about 8nm of wavelength in the range from 339nm to 1014nm. We found it possible to distinguish between healthy and dead part of lettuce plants where human couldn’t find it as a dead part.

An effect of PL filter on imaging that eliminates halation on object surface is well known. However, it has not been easy to use the filter in front of halogen, incandescent, or HID lamps which radiate a lot of heat, because PL film is melted with a temperature of about 60 degree centigrade. Although fluorescence lamps or LEDs can be used with PL filters, their intensity and color rendering are inferior to them. Diffusers or domes with reflection plates are often used when halogen lamps are used, but unavoidable halation and surroundings reflection occur on glossy surface. In this presentation, a direct lighting device (called DL) developed for fruit grading systems is introduced. DL mainly consisted of a PL filter, two heat absorption filters, a halogen lamp, and a fan. Since the absorption wavelength bands of the two filters were different, major heat did not reach the PL filter. In addition, cold air was introduced to the PL filter by its fan. Most parts of plants have cuticular layers on their surfaces to keep moisture within themselves. In this study, three categories are created for objects corresponding to degree of gloss on surfaces and surface materials: category 1 (apples, tomato, eggplants, bell peppers), category 2 (pear, kiwi, peach), and category 3 (potatoes, radish). A color CCD camera was set at the top of object with four DLs and images were acquired. To compare the PL filtering images, a dome and diffusers were also used for the image acquisition. From experiments through three types of illumination devices, it was observed that PL filtering images not only could eliminate halation on products of the category 1 but also could express true colors of the products. Microscopic images showed proportional relation between thickness of cuticular layers and halation on surfaces.

A method to measure the per-tree yield of citrus fruits was developed based on three approaches. Initially, the problem was viewed as a variation on the theme of particle counting as performed earlier on flows of fertilizer particles. The main difference was that 1) citrus fruits are much larger than fertilizer particles 2) they travel at a significantly lower speeds an 3) their flow density may be higher than that of fertilizer. The earlier research for fertilizer mass flows used a dedicated time of flight device that employed optical magnification to measure small particles with relatively large sensors. A second advantage was that the image of the particles could be projected in focus on the sensor plane. In the case of citrus fruit counting there was no need for magnification, since citrus fruits are much larger than the sensors. However, for citrus fruits this led to defocus problems, when direct lighting was used. The theory used in the fertilizer application was capable of dealing with multiple particles (clumps) interrupting the sensors simultaneously, which made the method non-intrusive and eliminated the need for particle singulation. Although this method was viewed to be transferable to the citrus counting problem, an easier and relatively low-tech method was assumed to be more reliable. A laser beam was projected just above the surface of a plate on which the fruits roll down into a collection bin. Theoretically, if the laser beam is placed low enough, each fruit will be counted individually, even if they are clumped together. Tests revealed that this assumption does not hold under high-density flow regimes. Therefore, an improved gutter system was designed which effectively transforms the flow of fruits into different streams of fruits in single file formation. This method, combined with low-placed laser beams, allowed for accurate counting of the fruits in the laboratory. In future research, the system will be tested on a canopy shaker machine.

Machine vision is widely applied in automated inspection and monitoring systems. One of the most important components of machine vision systems is illumination. To remove and avoid halation on images, indirect lighting devices such as diffusers or domes with several lamps have been studied for post-harvest operations. However, it is difficult to control natural illumination in the field, because sunlight constitutes a point source, which causes shiny and shady parts on objects, and its intensity fluctuates with time. In addition, most parts of plants have a cuticular layer, causing halation effects, which reduce image quality. In this study, a tractor-mounted imaging system was developed to acquire high quality crop images. The system consisted of a frame box (1m X 1m X 1m), covered by a polarizing film (1m X 1m), 2 identical cameras in the box, and a PC with two image capture boards. All 4 sides of the frame box were covered with black curtains so that only polarized sunlight could enter. Polarizing filters were also fitted to the lens of one camera to compare 2 images from the cameras at a time. In experiments, images of coffee plants were acquired under varying sunlight conditions. From the results, it was observed that the polarizing filtering technique was capable not only reduce by 97% of halation on the surfaces of leaves but also of achieving high accuracy color representation of the leaves.

To recognize crop positions by machine vision system is a basic technology to achieve a non-herbicidal, mechanical weeding method. But in fields, where lighting conditions change kaleidoscopically, it is very difficult to recognize the position stably. We propose a method to estimate crop position using a template matching. A template that scans matching area was designed including a crop model, intra- and inter-row spacings. A cross-correlation image was calculated between the template and a field image captured by a CCD camera. Local maximums in the cross correlation image were regarded as crop positions. Our method had high robustness to field conditions and ran stably with various images in rice fields.