Gerie W.A.M. van der Heijden

Calibration and characterisation of imaging spectrographs

Spectrograph-based spectral imaging systems provide images with a large number of contiguous spectral channels per pixel. This paper describes the calibration and characterisation of such systems. The relation between pixel position and measured wavelength has been determined using three different wavelength calibration sources. Results indicate that for spectral calibration, a source with very narrow peaks, such as a HgAr source, is preferred to narrow band filters. A third-order polynomial model gives an appropriate fit for the pixel to wavelength mapping. The signal-to-noise ratio (SNR) is determined per wavelength. In the blue part of the spectrum, the SNR is lower than in the green and red part. This is due to a decreased quantum efficiency of the sensor, a smaller transmission coefficient of the spectrograph, as well as low output power of the illuminant. Increasing the amount of blue light, using an additional fluorescent tube with a special coating considerably increases the SNR. Furthermore, the spatial and spectral resolution of the system has been determined in relation to the wavelength. These can be used to choose appropriate binning factors to decrease the image size without losing information. In our case this could reduce the image size by a factor of 60 or more.

A Mobile Field Robot with Vision-Based Detection of Volunteer Potato Plants in a Corn Crop'

Volunteer potato is a perennial weed that is difficult to control in crop rotations. It was our objective to build a small, low-cost robot capable of detecting volunteer potato plants in a cornfield and thus demonstrate the potential for automatic control of this weed. We used an electric toy truck as the basis for our robot. We developed a fast row-recognition algorithm based on the Hough transform and implemented it using a webcam. We developed an algorithm that detects the presence of a potato plant based on a combination of size, shape, and color of the green elements in an image and implemented it using a second webcam. The robot was able to detect potatoes while navigating autonomously through experimental and commercial cornfields. In a first experiment, 319 out of 324 images were correctly classified (98.5%) as showing, or not showing, a potato plant. In a second experiment, 126 out of 141 images were correctly classified (89.4%). Detection of a potato plant resulted in an acoustic signal, but future robots may be fitted with weed control equipment, or they may use a global positioning system to map the presence of weed plants so that regular equipment can be used for control. Nomenclature: Corn, Zea mays L, Potato, Solanum tuberosum L. Additional index words: Autonomous navigation, autonomous weeding, glyphosate, machine-vision, site-specific weed control. Abbreviations: DIPlib, Delft image-processing library; DSP, digital signal processor; GPS, global positioning system; JPEG, Joint Photographic Experts Group; NiMh, nickle metal hydride; PC, personal computer.

A robot to detect and control broad-leaved dock ( Rumex obtusifolius L.) in grassland

Broad-leaved dock is a common and troublesome grassland weed with a wide geographic distribution. In conventional farming the weed is normally controlled by using a selective herbicide, but in organic farming manual removal is the best option to control this weed. The objective of our work was to develop a robot that can navigate a pasture, detect broad-leaved dock, and remove any weeds found. A prototype robot was constructed that navigates by following a predefined path using centimeter-precision global positioning system (GPS). Broad-leaved dock is detected using a camera and image processing. Once detected, weeds are destroyed by a cutting device. Tests of aspects of the system showed that path following accuracy is adequate but could be improved through tuning of the controller or adoption of a dynamic vehicle model, that the success rate of weed detection is highest when the grass is short and when the broad-leaved dock plants are in rosette form, and that 75% of weeds removed did not grow back. An on-farm field test of the complete system resulted in detection of 124 weeds of 134 encountered (93%), while a weed removal action was performed eight times without a weed being present. Effective weed control is considered to be achieved when the center of the weeder is positioned within 0.1 m of the taproot of the weed—this occurred in 73% of the cases. We conclude that the robot is an effective instrument to detect and control broad-leaved dock under the conditions encountered on a commercial farm.

The MCRA model for probabilistic single-compound and cumulative risk assessment of pesticides

Pesticide risk assessment is hampered by worst-case assumptions leading to overly pessimistic assessments. On the other hand, cumulative health effects of similar pesticides are often not taken into account. This paper describes models and a web-based software system developed in the European research project ACROPOLIS. The models are appropriate for both acute and chronic exposure assessments of single compounds and of multiple compounds in cumulative assessment groups. The software system MCRA (Monte Carlo Risk Assessment) is available for stakeholders in pesticide risk assessment at mcra.rivm.nl. We describe the MCRA implementation of the methods as advised in the 2012 EFSA Guidance on probabilistic modelling, as well as more refined methods developed in the ACROPOLIS project. The emphasis is on cumulative assessments. Two approaches, sample-based and compound-based, are contrasted. It is shown that additional data on agricultural use of pesticides may give more realistic risk assessments. Examples are given of model and software validation of acute and chronic assessments, using both simulated data and comparisons against the previous release of MCRA and against the standard software DEEM-FCID used by the Environmental Protection Agency in the USA. It is shown that the EFSA Guidance pessimistic model may not always give an appropriate modelling of exposure.

Visualization of spectral images

Spectral image sensors provide images with a large number of contiguous spectral channels per pixel. Visualization of these huge data sets is not a straightforward issue. There are three principal ways in which spectral data can be presented; as spectra, as image and in feature space. This paper describes several visualization methods and their suitability in the different steps in the research cycle. Combinations of the three presentation methods and dynamic interaction between them, adds significant to the usability. Examples of some software implementations are given. Also the application of volume visualization methods to display spectral images is shown to be valuable.

Combining stereo and time-of-flight images with application to automatic plant phenotyping

This paper shows how stereo and Time-of-Flight (ToF) images can be combined to estimate dense depth maps in order to automate plant phenotyping. We focus on some challenging plant images captured in a glasshouse environment, and show that even the state-of-the-art stereo methods produce unsatisfactory results. By developing a geometric approach which transforms depth information in a ToF image to a localised search range for dense stereo, a global optimisation strategy is adopted for producing smooth and discontinuity-preserving results. Since pixel-by-pixel depth data are unavailable for our images and many other applications, a quantitative method accounting for the surface smoothness and the edge sharpness to evaluate estimation results is proposed. We compare our method with and without ToF against other state-of-the-art stereo methods, and demonstrate that combining stereo and ToF images gives superior results.

Non-destructive automatic leaf area measurements by combining stereo and time-of-flight images

Leaf area measurements are commonly obtained by destructive and laborious practice. This study shows how stereo and time-of-flight (ToF) images can be combined for non-destructive automatic leaf area measurements. The authors focus on some challenging plant images captured in a greenhouse environment, and show that even the state-of-the-art stereo methods produce unsatisfactory results. By transforming depth information in a ToF image to a localised search range for dense stereo, a global optimisation strategy is adopted for producing smooth results that preserve discontinuity. They also use edges of colour and disparity images for automatic leaf detection and develop a smoothing method necessary for accurately estimating surface area. In addition to show that combining stereo and ToF images gives superior qualitative and quantitative results, 149 automatic measurements on leaf area using the authors system in a validation trial have a correlation of 0.97 with true values and the root-mean-square error is 10.97 cm(2), which is 9.3% of the average leaf area. Their approach could potentially be applied for combining other modalities of images with large difference in image resolutions and camera positions.

Advances in automatic detection of tulip breaking Virus (TBV) using machine vision

Abstract Tulip breaking virus (TBV) causes severe economic losses in flower bulbs in the Netherlands. To prevent further spread by aphids, infected plants must be removed from the field as soon as possible. Until now screening is done by visual inspection in the field. As the availability of human experts is limited there is an urgent need for a rapid, automated and objective method of screening. Based on laboratory experiments, we developed a vision method for use in the open field. From 2009 - 2012 field trials were carried out and the techniques were tested and improved. First field trails were on single plant density where plants does not overlap in the images. In 2012 the experiment conducted a tulip field planted at production density of 100 and 125 plants per square meter, resulting in images with overlapping plants. The final score of our system in this production density experiment approached the scores obtained by the experienced crop experts.

FLORES: A JAVA Based Image Database for Ornamentals

Flores is a database system for finding and retrieving plant varieties with similar appearance. The system allows color-images to be segmented interactively using JAVA Applets. The alpha-channel of the image is thereby used for the segmentation result, i.e. it contains the object. The image with object mask is transferred to a server-system, which extracts information from the object in the image. The information can be color histograms or a set of size, shape, texture and color features. The set of features depends on the type of object, as indicated by the user. The histogram is binarized and, using fast algorithms, matched with the binarized database histograms. For the feature set, nearest neighbors are searched in a multi-dimensional features space using an SR-tree algorithm. The server returns information of the most similar objects, like similarity, color-image and object mask to the Applet.