It is known that biomaterials have unique characteristics in their color, shape, size, spectral reflectance, and biological structure which are different from industrial materials. Specific machine vision designs are often required for constructing inspection, monitoring, measuring, or production systems to handle the biomaterials. In this presentation, cuticular layers on fruit surfaces are in focus and a direct lighting device with a polarizing filter, which was developed for a fruit inspection system is introduced, because illumination is one of the most important components. Since the biomaterials have various reflectance in different wavelength bands, characteristics of light from gamma ray to terahertz were described. In the visible region, there are many types of light sources, which have different properties on luminance, lumen maintenance, color rendering, color temperature, life, and cost. A high quality image acquisition requires a proper light source selection. Fruit grading facilities are places where the largest number of machine vision systems are practically used in agricultural fields. In this paper, a conventional fruit grading machine and a fruit grading robot with machine visions are introduced. In addition, a contribution of the imaging technologies of the grading systems to the food traceability system is explained.
This paper summarizes the authors' research in imaging- and optical weed sensors based on studies of plant spectral and morphological characteristics. Location, color, texture, and shape features of weeds and crops were identified as the major criteria for sensor structure and algorithm design. Imaging-based weed sensors were compared with optical weed sensors. Factors affecting detection accuracy were studied through experiments. Laboratory experiments were also conducted to test the accuracy of statistical and neural-network classification models. Extensive field experiments were conducted to test the effectiveness of the weed sensors.
Tomato cultivation in large-scale plant factories is expanding in Japan. It is therefore important to measure the leaf area and growth of tomato accurately in the plant factory for estimating the yield. In this study, the area of tomato leaves was measured using a three-dimensional range meter. The models, which estimated the whole area of the tomato plant, were then developed using the area of a part of non-destructively measured leaf. The parameter necessary for the modeling was designed by taking the leaf position on tomato plant into account. As a result, a practical model for measuring the tomato leaf area non-destructively was developed.
Recent developments of microcomputer-based machine vision systems has offered convenient and non-destructive ways for measurements of plant characteristics that allow plant growth assessment. This paper presents brief reviews and comparisons of machine vision approaches for non-destructive plant growth measurement. The simple approach of using projected silhouette image of a plant was most commonly used in acquiring growth data in various experiments. Plant fresh or dry weight can be indirectly estimated from projected leaf area by calibration with data from standard measurement methods. However, the estimation error usually increases as the extent of overlapped leaves increases. Images acquired from different views or utilizing dual cameras allow the estimation of leaf area without predetermined calibration relationship but the image processing algorithms usually become more sophisticated. By incorporating multiple images of a plant, three-dimensional and structural information may be extracted for more detailed growth analyses and modeling.
Hybrid robust image processing system which extracts visual features of interest during plant cultivation was developed based on the wireless tele-operative man-machine interface. The robustness of processing image was achieved via task sharing between the computer and the operator. Utilizing a man-machine interactive hybrid decision-making system which was composed of three modules such as wireless image transmission, task specification and identification, and man-machine friendly interface, computing burden and the instability of the image processing results caused by the variation of illumination and the complexity of the environment from the ambiguity among stems, leaves, shades, and fruits were overcome. Color and brightness reflectance of various parts at the cultivation site such as soil, mulching vinyl, straw, leaves, and fruits were analyzed. Segmentation of an object of interest was performed utilizing the trend of brightness and color distribution of each part. Hough transform was modified for the image obtained from the locally specified window to extract the geometric shape and position of the elliptic fruit. The processing time was less than 100 ms. The proposed system showed the robustness and practicability in identifying plant status at the cultivation site.
Mass production of moss plant has been expected because of huge demand of roof top greening of factory buildings. A biotechnology for proliferation of sunagoke moss has been developed. It will be produced from plant factory. A bio-response feedback control strategy known as Speaking Plant Approach (SPA) was applied to the automated moss plant production system. Moisture content, water potential and leaf area index were measured and used for an Artificial Neural Network (ANN) model output. Three textural analysis features (Energy, Local homogeneity, Contrast) were obtained for input parameters of the model. The results of the experiment using ANN model show that it is possible to predict the moss water status parameters by using textural features. It was shown that through appropriate selection of the architecture of the network, all parameters of moss water status can be predicted. By using back-propagation supervised learning and inspection data method, ANN prediction model was tested successfully describing the relationship between textural features and water status parameters. It also produced high correlation between measured and predicted value (R2 ranged from 0.90 to 0.98) and minimum absolute error using inspection data. This indicates that SPA will become an attractive strategy for control system for moss production factory.
In this research, growth information of wheat was acquired using multi-spectral imaging sensor (MSIS) mounted on an unmanned helicopter and application of variable rate fertilizer based on this growth information. Growth information of the wheat obtained from normalized difference vegetation index (NDVI) value. NDVI value was calculated by image processing of the acquired image. Calculated NDVI was used to generate a map. This map was used for the application of the variable rate fertilizer in the wheat field. Moreover, the growth variation of the wheat by variable rate fertilizer was analyzed. The variability of SPAD was low before the variable rate fertilizer. Comparison of variable rate and uniform rate fertilizer has been carried out. The cross-correlation coefficient of SPAD and NDVI was low in the variable rate fertilizer compare with uniform rate fertilizer section. There is no correlation before and after variable rate fertilizer because the variable rate fertilizer influences the SPAD and NDVI of the wheat. The protein variability of the variable rate fertilizer section is less than the uniform rate fertilizer section. The yield in variable rate fertilizer section that has been fertilized with nitrogenous fertilizer of 1.7 N-kg 10 acre-1 is almost the same with the uniform rate fertilizer section.
Carbon balance and flower quality of post-harvest flower bud opening (FBO) were studied in cut shoots of the standard red carnation ‘Francesco’. For each FBO treatment, six ‘tight bud’ - cut carnations were placed in a ventilated transparent cylindrical container, separated into two compartments (flower heads, 25±1°C, 85±5% RH and leaves/stems, 23±1°C, 95±5% RH) by a transparent acrylic plate under continuous light with a PPFD of 30 (P30), 90 (P90) or 150 (P150) μmol m-2 s-1 with the cut ends in flower opening solution (25 mg L-1 AgNO3, 200mg L-1 8-hydroxyquinoline citrate, and 30g L-1 sucrose) . The time required for FBO was reduced for higher PPFD for Experiment A (8.0 d in P30, 7.0 d in P90, and 6.0 d in P150, respectively) and Experiment B (6.0 d in P30, 5.0 d in P90, and 5.0 d in P150, respectively) . The mean flower quality scores of opened carnations after FBO closely correlated with total carbon uptake (TCU) with correlation coefficients of 0.94 and 0.98 during 10-d flower quality evaluation for Experiments A and B. Carbon from net CO2 exchange per stem (CNCE) and carbon from gross photosynthesis (CGP) of leaves/stems in P150 was greater than those in P90 and P30.
The effect of electrical conductivity (EC) on the voltage response of ECH2O soil moisture probe (ECH2O probe) has been investigated by laboratory experiments in salt solutions and saline as well as non-saline soils. The output of the probe decreases noticeably with increasing EC up to 0.24 dS m-1 but increases significantly over higher EC in salt solution and saline soils. However, the effect of soil-water EC is relatively small below 0.5 dS m-1 and may be neglected for practical purposes without significant error. Due to higher probe output in coarse textured soils than in fine textured soils for any particular water content, the salinity effect is much higher in sand than in clay. The calibration function between the volumetric soil-water content and probe output for non-saline soils estimates significantly different soil-water contents for soil-water EC>0.5 dS m-1. Due to less contact of dry soil with the probe and reduced width of electromagnetic field on the probe in wet soils, the ECH2O probe is less sensitive both at low and high soil-water content irrespective of the salinity level.