Key Laboratory for Information Technology in Agriculture

Beijing, China

Key Laboratory for Information Technology in Agriculture

Beijing, China

Time filter

Source Type

Miao T.,Beijing Research Center for Information Technology in Agriculture | Miao T.,Chinese National Engineering Research Center for Information Technology in Agriculture | Miao T.,Key Laboratory for Information Technology in Agriculture | Miao T.,Beijing Key Laboratory of Digital Plant | And 21 more authors.
Nongye Gongcheng Xuebao/Transactions of the Chinese Society of Agricultural Engineering | Year: 2016

Three-dimensional (3D) Plant modeling and visualization is a key research issue in both digital plant and agricultural application. Leaf is one of the vital organs in a plant, so the 3D modeling and shading of plant leaves is an important and fundamental work for achieving the goals of digital plant. Appearance simulation of plant leaves is still a challenging issue because of its intricate underlying structure and complex and subtle interaction with light. Texture mapping using leaf photo is a common method for appearance simulation, however, it could bring noise caused by light environment and camera position in lighting simulation step. This paper presents a technique for simulating the appearance of plant leaves with multiple images. Our method can estimate the spatially-varying reflectance properties of plant leaf surface based on a few images, which capture leaves' appearance transition information with different light directions. An apparent image acquisition system using linear light source is built for capturing 400 images with a fixed camera viewpoint and a single direction of motion for the linear light source. This system is composed of a driving module, a linear source module, a background module and a camera. Using a linear light rather than a point light source as the illuminant, we can obtain a piece of area with more intensive illumination. With these image data, we develop a fitting method, which is able to estimate the diffuse color, specular color and specular roughness of each point on the leaf surface. In our method, the isotropic ward model is utilized as the appearance model for specifying that how the leaf surface reflects light. Our fitting technique first simulates the change of reflectance attributes of diffuse and specular reflectance lobes under moving linear light source. In this process, a rectangle is employed to simulate the linear light source and Monte Carlo integration method is used to calculate the radiation transmission process. When we have the simulating results, the appearance parameters of each pixel are determined by comparing its actual parameter values to the simulating results. By above fitting method, 3 kinds of spatially-varying appearance parameters are saved into 3 parameter images for rendering leaf appearance. For quickly shading, multipoint point light sources are used for simulating various illumination conditions instead of complex radiative transfer integral. Using appearance parameter images and shading method, static appearance or dynamic appearance transition of plant leaves can be generated realistically. From the results obtained by this method, we find that it can render more accurate and real appearance texture of leaves compared to traditional texture mapping methods. The advantages of our method are that the appearance parameter images for rendering have removed the light and viewport noise, and only contained the appearance material information. In order to prove this conclusion, we quantitatively analyze the reason for this advantage by some formula derivations in this paper. But for obtaining these advantages, our method needs more complex data acquisition process and parameter fitting algorithm, which will reduce the efficiency of simulation. For improving the efficiency of our method, 2 approaches are discussed in this paper, including reducing image resolution and fitting the specular parameters of the whole leaf by a few sample points. Our method can estimate some appearance parameters which are plant leaf own intrinsic properties. We believe this characteristic will make these appearance parameters used not only for visualization, but also as some important phenotypes instead of so-called color data. In the future work, we will extend the application of our method in agriculture, such as monitoring plant growth status with the appearance parameters, or analyzing the differences among plant varieties. © 2016, Chinese Society of Agricultural Engineering. All right reserved.


Miao T.,Beijing Research Center for Information Technology in Agriculture | Miao T.,ShenYang Agricultural University | Miao T.,Chinese National Engineering Research Center for Information Technology in Agriculture | Miao T.,Key Laboratory for Information Technology in Agriculture | And 12 more authors.
Nongye Gongcheng Xuebao/Transactions of the Chinese Society of Agricultural Engineering | Year: 2016

Simulation of three-dimensional (3D) crop scene infected by crop disease is a tough task, because the related appearance data information is difficult to obtain. To obtain specific disease appearance information, careful bacteria culture and continuous observation may be needed with long-time experimental work and precise environmental control. This paper presents a general method to simulate the appearance transition of crop leaves infected by common diseases based on existing image in the Internet. We assume that a disease image contains some key appearance information in the process of disease infection. Based on this assumption, a set of static properties are extracted from image including shape and color of disease spots on the crop surface, and meanwhile the relevant dynamic transition processes of these properties are also deduced. For analyzing color transition, K-MEANS is firstly used to classify the color vectors of pixels in disease image into 8 categories and the average color vector of each category is computed which is called disease color feature vector. Then, these 8 vectors are sorted based on their proportions of green channel. To get a continual color aging simulation result, 7 linear functions are generated by interpolation between adjacent vectors. Finally, 141 discrete color vectors are sampled from these functions and used to generate the disease color transition texture. In order to obtain dynamic morphogenesis process of disease spot, the threshold segmentation method is firstly applied to segment the disease spot pixels from the pixels of normal crop leaves. Then a gray value is computed for each disease spot pixel based on the mimimum Euclidean distance between pixel's color vector and each disease color feature vector. These gray values of each disease spot pixel are recorded into the texture called morphogenesis texture. The distribution of disease spot on the crop organ surface is complex and random. A interactive interface tool has been developed for designing the distribution. With the tool, users can put some morphogenesis textures onto any location of the crop 3D models and change the size and direction of morphogenesis textures according to users' experience. The operating result is also saved as the texture called distribution texture. The disease color transition texture and distribution texture contain the necessary dynamic appearance information of disease spot and are used in the visualization step. For simulating a dynamic and continual appearance transition process of crop disease, a group of degree parameters for arbitrary 3D position on the crop surface are applied to generate the disease appearance which is computed using the distribution texture and the interactive parameter called general disease degree parameter. With the general degree parameter, user can get a simulation result under any infected state. In order to better define the disease appearance, we decompose it into the symptom appearance for describing the ageing status of the crop organ and the mildew layer appearance caused by the accumulation of mycelium. We consider the crop organ as a homogeneous structure and use the isotropic ward BRDF (bidirectional reflectance distribution function) model to simulate the symptom appearance. The diffuse reflection of ward model at arbitrary position on crop is selected from the color transition texture based on the degree parameter of this 3D position. In order to simulate the volumetric nature of the mildew layers, the shell model is integrated into our approach and the attributes of shell model are all controlled by the degree parameter. We have realized the algorithm in this paper using OpenGL, and found that the method can realistically render the appearance of the crop infected by the disease using only one or a few images. Our strategy is to use existing disease image from Internet to generate plant disease 3D animation, and it can solve the problem of the lack of related apparent data information of plant diseases. This research can provide a powerful tool to produce animations for agricultural science training. © 2016, Chinese Society of Agricultural Engineering. All right reserved.


Wu H.,Chinese National Engineering Research Center for Information Technology in Agriculture | Wu H.,Key Laboratory for Information Technology in Agriculture | Gao R.,Chinese National Engineering Research Center for Information Technology in Agriculture | Gao R.,Key Laboratory for Information Technology in Agriculture
ICIC Express Letters, Part B: Applications | Year: 2013

Intelligent system is a hardware and software entity, which is able to understand and learn complex information and can make decisions and analyze behaviors. It has the capability of identifying objects and events, storing rich knowledge, reasoning and prediction. Considering the demand of intelligent system processing complex task, the method of multi-agent task cooperation process is proposed in this paper. Based on the idea of distributed artificial intelligence, a cooperative decision organizational framework of agent is organized, which is as the centre of project and task. Subsequently, a task description method of multi-agent is proposed, and immune memory, clone selection and affinity calculation all have been applied to solving the coordination problem and achieve low complexity and multi-task agent coordination mechanism. © 2013 ICIC International.


Qian J.-P.,Chinese National Engineering Research Center for Information Technology in Agriculture | Qian J.-P.,Key Laboratory for Information Technology in Agriculture | Yang X.-T.,Chinese National Engineering Research Center for Information Technology in Agriculture | Yang X.-T.,Key Laboratory for Information Technology in Agriculture | And 8 more authors.
Computers and Electronics in Agriculture | Year: 2012

Wheat flour undergoes several processing steps in its transformation from raw wheat in the mill, which differentiates wheat flour from other farm products. At each step, various wheat sources are combined into one batch of wheat flour. This study primarily aimed to develop a Wheat Flour Milling Traceability System (WFMTS), incorporating 2D barcode and radio frequency identification (RFID) technology, and to validate the system in a wheat flour mill in China. We designed the encoding rules for the raw material, processing and traceability batches. Labels with a Quick Response Code (QR Code) were attached to small packages of wheat flour to link them to their processing information, and RFID tags were affixed to the storage bins to record logistics information. A traceability system was developed based on batch identification and record keeping. The system was applied and supported in a wheat flour mill for one year. The WFMTS management and traceability capacity was evaluated using a contrast experiment. The experiment was divided into five parts, including raw material data recording, processing data recording, package data recording, logistics data recording and traceability query. The results show that although time consumption using WFMTS in package data recording was more than that with paper recording, WFMTS was dominant in its total time consumption: five parts were reduced by 113%, and the mean accuracy of the five parts increased by 8%. The QR Code and RFID recognition accuracy was evaluated using experiments with different reading distances. The cost and income variations in application WFMTS were analyzed based on the survey. The results show that the total cost increased by 17.2% to apply the system. Compared to the cost, the sales income increase was obvious, and it reached 32.5%. Considering the good evaluation results, the system has good application potential in medium or large wheat mill enterprises. © 2012 Elsevier B.V.


Qian J.-P.,Chinese National Engineering Research Center for Information Technology in Agriculture | Qian J.-P.,Key Laboratory for Information Technology in Agriculture | Yang X.-T.,Chinese National Engineering Research Center for Information Technology in Agriculture | Yang X.-T.,Key Laboratory for Information Technology in Agriculture | And 8 more authors.
Journal of Food, Agriculture and Environment | Year: 2013

Traceability system is an effective measure to guarantee food quality and safety. Recently, IT-based vegetable traceability system with different information technology is focused on academic research and applied as pilot projects in some cities in China. This paper analyzes a structure of ITbased traceability system, including production identification, supply chain management system and central database. Based on the structure, traceability systems on different operating mechanisms are applied in two cities in China. A sample investigation with five parts is designed to analyze process and barriers of the operating mechanisms. Fifteen agribusinesses with the authorities-driven mechanism in Tianjin and 15 agribusinesses with the enterprise-driven mechanism in Guangzhou are selected for the investigation. The results show that the traceability systems with different driven mechanism have their own strengths and limitations. A framework for a vegetable traceability system integrated the strengths of two different operating mechanisms is proposed. Authorities department is responsible for the establishment of traceability service platform and setting uniform rules. Agribusiness enterprises are responsible for the development of information record system which is suitable for custom requirement on the uniform rules. Some measures are given to guarantee the system working well.


Wang C.,Beijing Research Center for Information Technology in Agriculture | Wang C.,Chinese National Engineering Research Center for Information Technology in Agriculture | Wang C.,Key Laboratory for Information Technology in Agriculture | Guo X.,Beijing Research Center for Information Technology in Agriculture | And 11 more authors.
Nongye Gongcheng Xuebao/Transactions of the Chinese Society of Agricultural Engineering | Year: 2016

Hemispherical photography (HP) has already proven to be a powerful indirect method for measuring various components of canopy structure. One of the main problems using HP for the determination of canopy structure such as leaf area index (LAI) and mean leaf angle (MLA), is the selection of the optimal brightness threshold to distinguish leaf area from sky area and thus produce a binary image. In this process, one of the challenges is how to overcome various natural light conditions which sometimes strongly affect the profile of the crop images taken outdoors. In this paper, a fusion and mapping method of illumination invariant multiple exposure images was proposed in order to eliminate negative impact of variant illumination. Firstly, a series of multiple-exposure maize canopy hemispherical images were captured under natural light condition. The camera (Canon EOS 5D Mark III, sigma 8 mm f 3.5 ex DG FISHEYE) was placed in the bottom of canopy towards the sky, and it provided vertical 180° and horizontal 360° canopy images. The images were captured at different exposure time such as 1/800, 1/400, 1/200, 1/125, 1/30 and 1 s. Secondly, the multiple photographs were fused into a single radiance map, so shadow and highlights parts of original images were extended to a large range. We were able to determine the irradiance value for each pixel, and plot it against the measured pixel value discretized according to the 256 pixel values commonly observed in 8-bit images for each exposure time. The pixel values were proportional to the true irradiance values in the scene. The pixel values, exposure time, and irradiance values formed a problem of least square. Finally, we also employed a histogram equalization method to map irradiance values to RGB color space. After mapping processing, the brightness of image had a more proper distribution, the dark regions were lit more brightly and the saturated regions were depressed to normal brightness condition; moreover, the histograms of images shared the similar distribution, that meant the pixels of variant illumination images after mapping processing had a high similarity. The comparison results showed that plant pixel of HP acquired at 14: 00 and 17: 00 with the threshold value of 180 had a difference of 15.4%, and our method reduced the difference, which was only 2.8%. In the analysis phase of canopy structure parameters, plant pixels were extracted from those photos, and then LAI and MLA could be inversed by Beer-Lambert theory based on the quantitative relationship between radiation condition and canopy structure. The experiment was conducted in 2013, and the planting density was 60000 plants/hm2 with normal water and fertilizer management. The hemispherical images were obtained on August 6th, 13th, 19th, 22nd and 26th and September 12th, and the distribution of LAI and MLA was consistent with the rules of growth and development of maize canopy. Moreover, a performance comparison of direct surveying method and our method was carried out, and the LAI and MLA values of 13 samples were collected with the 2 methods. The results of regression analysis showed that our method had a high consistency with direct surveying method of canopy structure parameter, and correlation coefficient between the values from the 2 methods hit 0.94. The line slope was 1.463, which indicated the values from our method were lower than the direct surveying method. The method proposed in this paper expands the application range of HP, and provides a solution for automatic monitoring of canopy structure parameters. © 2016, Editorial Department of the Transactions of the Chinese Society of Agricultural Engineering. All right reserved.


Wang C.,Beijing Research Center for Information Technology in Agriculture | Wang C.,Chinese National Engineering Research Center for Information Technology in Agriculture | Wang C.,Key Laboratory for Information Technology in Agriculture | Guo X.,Beijing Research Center for Information Technology in Agriculture | And 11 more authors.
Nongye Gongcheng Xuebao/Transactions of the Chinese Society of Agricultural Engineering | Year: 2014

The missing amount of planted corn seedlings plays an important role in corn yield, to acquire it automatically, a new system based on machine vision has been developed. System hardware includes: one Industrial Personal Computer, a Central Processing nit: Intel (r) CPU i5@3.4GHz, 4 Gb of memory, one mvc3000 high speed Industrial camera (24FPS), and one Pentax len 8.5 mm f/1.5. The software development environment includes: Win7 Operating System, Microsoft Visual Studio 2010 Professional, and OpenCV1.0. The core of the system is the image processing method. Firstly, image sequences obtained along plant rows from a top view under in-field lighting conditions were registered to the uniform coordinate system. Secondly, plant pixel (vegetation) was segmented from the background with a pixel classifier trained by a neural network. The segmentation method employed a decision surface in color space that was defined by only three parameters. This surface was a a truncated ellipsoidal surface which was robust in outdoor field images under varying lighting conditions. A simple parallel algorithm working on 8-connectivity was implemented, whereby skeletonization extracts a network of thin curves that describe the overall shape or "skeleton" of objects in a binary image. Due to limitations in camera resolution and non-ideal lighting conditions, the minimum gray level point along the plant skeleton is the best estimation of the actual stem location. The minimum gray pixel area was searched along the plant skeleton, and the center of minimum gray pixel area was marked as the stem center. Finally, a plant row line was fitted by stem centers; a model that predicts a linear relationship between the stem centers and the corn plant row was defined, and the parameters of linear function was estimated by a least-squares fit. Stem centers were projected onto the row line, and the average plant spacing was calculated by a projected point. The number of missing plants between two neighbored seedlings has a linear relationship of plant average spacing. On three varieties of 10 repeats each, a 10 m long row field experiment was performed, In a low density experiment, measurement results of the method agree with manual measurements of 7 in 10 and 3 in 10 have a difference of one plant. In a high density experiment, measurement results of the method agree with a manual measurement 6 in 10 and 4 in 10 have a difference up to two plants. Comparison with a manual measurement and our method, a high correlation on the two methods was found; this method can replace manual measurement, reduce time cost and human labor effort, and improve the degree of automation of the corn seedling missing survey.


Zhao C.-J.,Chinese National Engineering Research Center for Information Technology in Agriculture | Zhao C.-J.,Key Laboratory for Information Technology in Agriculture | Li M.,Chinese National Engineering Research Center for Information Technology in Agriculture | Li M.,Key Laboratory for Information Technology in Agriculture | And 8 more authors.
Computers and Electronics in Agriculture | Year: 2011

Treatment during the primary infection phase is essential for controlling cucumber downy mildew in solar greenhouses. An early warning model applicable to this phase would represent a foundation for early warning systems for managing the disease and reducing pesticide usage. Based on the input parameters that were both readily available and appropriately limited in number, EWMPICDW (early warning model for primary infection of cucumber downy mildew in solar greenhouses) was developed based on monitoring data, early warning theory and plant disease epidemiology. The elaboration of this model included clarification of the meaning of warning, monitoring the warning indicators, forecasting the warning situation, tracing the warning sources and controlling the warning situation. The definition of warning included disease occurrence (yes or no) and probability. Because the leaf wetness duration (LWD) played an important role in disease warning systems for crops in solar greenhouses and was difficult to monitor, the leaf wetness sensor and RH threshold model were investigated and combined to form a practical estimation solution for LWD. Within the warning situation forecasting model system, the infection condition and incubation early warning submodels received the most attention. The infection condition early warning submodel was developed by using a threshold method based on the combination of LWD and mean temperature in LWD. The temperature was chosen as the warning indicator for incubation, and the incubation early warning submodel was defined using nonlinear regression methods. The warning sources traceability algorithm was developed in relation to expert knowledge and in terms of a mode of disaster mitigation that involved cutting the disaster chain from the headstream. The method for controlling the warning situation was based on good agricultural practices (GAP). The early warning model was implemented as a system and was evaluated using data for 4 years at two sites in Beijing, China. The warnings can be provided more than 2 d before symptoms appear. Using EWMPICDW, a positive early warning is associated with a change in the chance of disease occurrence from 0.68 to 0.96. Accordingly, the probability of disease occurrence calculated for the early warning model was 96%. These results demonstrate that the data-driven model will support the development of early warning systems for primary infection by cucumber downy mildew in solar greenhouses. © 2011 Elsevier B.V.


Wang C.,Beijing Academy of Agriculture and Forestry Sciences | Wang C.,Chinese National Engineering Research Center for Information Technology in Agriculture | Wang C.,Key Laboratory for Information Technology in Agriculture | Guo X.,Beijing Academy of Agriculture and Forestry Sciences | And 11 more authors.
Nongye Gongcheng Xuebao/Transactions of the Chinese Society of Agricultural Engineering | Year: 2013

Maize ear morphological characteristics have important applications in breeding, germplasm, and cultivation areas, subject to the extent of technology development in relevant areas, but the approach of surveying morphological characteristics is not highly automatic. In this paper, we present a new machine vision based method and a supporting device for maize ear morphological characteristic surveying. First, the maize ear was placed on a rotating component, which rotates the maize ear in a fixed angle interval in order to capture 16 images more or less. A preprocess was carried out of maize ear image sequences to remove the image background, and the remaining part of the maize ear image was passed to the next process. The SIFT (Scale Invariant Feature Transform) was used to extract image feature points, and the feature points in the neighboring images could be matched up according to SIFT feature points. The relative motion between the two images could be described by a homography, and an overdetermined equations composed of matching points and homography make specific values of homography available. Mismatched feature points will reduce the accuracy of the homography equation solution dramatically. We adopted a RANSAC (random sample consensus) method to remove the outlier of the matching points during the homography solving process. Secondly, according to the motion described by homography, the first image and the next image are registered to the same coordinate system, using the dynamic programming method to find the seam-line in the two images, cutting the redundancy region in the two images along the seam-line. Since the exposure of the two images had certain differences which led to image brightness near seam-line being slightly different, a weighted Gaussian filter was imposed on both sides of the stitching image to eliminate exposure difference. Finally, the fusion image according to the order in sequence generated the ear panorama, row number, number in a row, kernel number, and other parameters were extracted by processing the maize ear panorama. The experimental results showed that: there is no significant difference between the method proposed by this paper and manual measurement, and the method proposed can greatly strengthen the automation of the maize ear traits investigation.


Miao T.,Beijing Research Center for Information Technology in Agriculture | Miao T.,Key Laboratory for Information Technology in Agriculture | Miao T.,Shenyang University | Zhao C.,Beijing Research Center for Information Technology in Agriculture | And 7 more authors.
Nongye Gongcheng Xuebao/Transactions of the Chinese Society of Agricultural Engineering | Year: 2014

3D agricultural scene under the condition of plant disease and insect pests is very difficult to simulate because of the complex appearance characteristics and severe apparent changes of the disease spots. The realistic appearance of plant leaves infected by the disease can't be obtained by the current methods. This paper presents a method to simulate the appearance of plant leaves infected by the disease. We assume that the disease spots uniformly distribute on the blade surface, spread from the spots center to the surrounding, and the shapes of the same kind of spots are similar. Based on these assumptions, the celluar bias function is used for controlling the shape, distribution and diffusion movement of the disease spot, and also for generating a 2D celluar texture image whose pixels represent the disease degree of any point on the blade surface. A degree parameter (in the range of 0 to 1) is used to adjust the pixel value of celluar texture to control the disease status, and the degree parameter equals 0 means there is no disease, and vice versa. We observed that some diseases can produce mildew layers on the leaf blade surface and which has volumetric, granular and arch form surface nature. In order to simulate the volumetric nature, the shell model is integrated into the approach. We use 15 passes to construct the shell model and use the degree parameter to discard the pixels which are not the mildew layers. For realistically rendering the grainy nature, the Perlin noise is applied to disturb the degree parameter for removing some pixels which belong to the mildew layers. With the purpose of generating an arch form mildew layer surface, we use the degree parameter to discard the pixels which belong to the larger passes of the shell model. Through this operation, the shell will present the height characteristics due to the gradual accumulation of the disease hyphae, middle part of the mildew layer is higher and the marginal part is lower. The optical property of the mildew layer is very hard to modeling because of the heterogeneous internal structure and the subsurface scattering property. In the approach, we construct a parameterized BRDF model to approximate the actual appearance. Owing to covering of the mildew, plant leaves ageing phenomenon happens. For rendering it, a leaf optical model with physiological factors is adopted, which can simulate the aging process by controlling some physiological parameters such as chlorophyll content and carotene content. The new method can be easily integrated with disease early warning model to simulate the disease appearance with different disease index or different environment parameters such as temperature and humidity. We realized the algorithm in this paper using OpenGL, and by comparing the rendering results to some actual disease images, we found that the method can realistically rendering the appearance of the plant leaves infected by the disease and insect pest. The research can provide a powerful tool to produce animations for agricultural science training. In the future work, we will focus on observing and analyzing some actual disease spread process to construct a more accurate parameter model for calculating the shape and the distribution of the disease spots.

Loading Key Laboratory for Information Technology in Agriculture collaborators
Loading Key Laboratory for Information Technology in Agriculture collaborators