David A. Mortensen

Shape Features for Identifying Young Weeds Using Image Analysis

Shape feature analyses were performed on binary images originally obtained from color images of 10 common weeds, along with corn and soybeans, found in the Midwest. Features studied were roundness, aspect, perimeter/thickness, elongatedness, and seven invariant central moments (ICM), for each plant type and age up to 45 days after emergence. Shape features were generally independent of plant size, image rotation, and plant location within most images. The ability to discriminate between monocots and dicots was most evident between 14 and 23 days using these features. Shape features that best distinguished these plants were aspect and first invariant central moment (ICM1), which classified 60 to 90% of the dicots from the monocots. Using Analysis of Variance and Tukey’s multiple comparison tests, shape features did not change significantly for most species over the study period. This information could be very useful in the future design of advanced spot spraying applications.

Textural imaging and discriminant analysis for distinguishing weeds for spot spraying

Advanced computer vision and statistical methods were employed for identifying living plants from soil/ residue background for two species of grasses (Shattercane, Green Foxtail) and two broadleaf species (Velvetleaf, Red Root Pigweed) weeds. The excess green index method was used as a contrast enhancement for specifically identifying plant from soil regions. Excess green classified plant and soil regions correctly over the entire three-week observation period with high accuracies (99% plus). Plant and soil binary images were derived from excess green images and provided edge boundaries. These boundaries were used with corresponding gray scale images to extract four classical textural features for plants and soil: angular second moment, inertia, entropy, and local homogeneity. These features were derived from the co-occurrence matrix. Stepwise and canonical discriminant analyses were used to test the classification performance of the texture and excess green features. Discrimination models of local homogeneity, inertia, and angular second moment were found to classify grass and broadleaf categories of plants, with classification accuracies of 93 and 85%, respectively. Classification accuracies of individual species only ranged from 30 to 77%. Soil classification accuracies were also high for textural feature algorithms (97%). The time required to produce tokensets ranged from 15 to 20 s on a UNIX computer system. Additional time required for the system to reach a plant/soil classification ranged from 5 to 10 s. This translated into an overall system response time of 20 to 30 s, with the preprocessing step constituting the major part of the system response time.

Plant species identification, size, and enumeration using machine vision techniques on near-binary images

Shape parameters such as aspect, roundness, and the ratio of thickness to perimeter were used to describe plant shape and are different according to the species that they represent. Color slide images of several species of plants were digitized for computer analysis. Three optical methods were tested to separate target plants from the soil and residue background. The separation method that provided the best contrast was the normalized difference index. Subtracting the blue or the red raster from the green raster also provided good separation on soils with little residue. Once the plant image had been isolated from the background, leaf edges were automatically traced using a commercial software package. Analysis of the shape of the plant outline was then performed, resulting in the plant shape parameters. Grasses and broadleaf plants had similar values for each shape parameter during the first ten days after emergence. After this period, differences occurred between grasses and broadleaf plants. The parameter that best discriminated grasses from broadleaf plants was the aspect (major axis length/minor axis length). However, when a grass sends out more than one shoot radially from the stem, the aspect will be similar to broadleaf plants. This study contributes to the design of a system that can determine weed populations and identify plant species without the use of human intervention.

How good is your weed map? A comparison of spatial interpolators

Abstract Recent interest in describing the spatial distribution of weeds and studying their association with site properties has increased the use of interpolation to estimate weed seedling density from spatially referenced data. In addition, farmers and consultants adopting elements of site-specific farming practices are using interpolation methods for mapping weed densities as well as soil properties. This study was conducted to compare the performance of four interpolation methods, namely inverse-distance weighting (IDW), ordinary point kriging (OPK), minimum surface curvature (MC), and multiquadric radial basis function (MUL), with respect to their ability to map weed-seedling densities. These methods were evaluated on data from four weed species, velvetleaf, hemp dogbane, common sunflower, and foxtail species, of contrasting biology and infestation levels in corn and soybean production fields in Nebraska. Mean absolute difference (MAD) and root mean square (RMS) between the observed point sample data and the estimated weed seedling density surfaces were used to evaluate the performance of the interpolation methods. Four neighborhood search types were compared within each interpolation method, and Search3 (12 to 16 neighboring locations) generated an interpolated surface with the smallest MAD and RMS indicating the highest precision. IDW with a power coefficient of p = 4 gave the smallest MAD and RMS, as did a test with an elliptical search and no anisotropy. The level of precision of all four interpolation methods was very poor for weed species with low infestation levels (> 75% of field weed-free; MAD ranged from 100 to 187% of the observed mean density), whereas precision was improved for weed species with high infestation levels (< 25% of field weed-free; MAD ranged from 45 to 85%). No single interpolator appears to be more precise than another. Implications of this study indicate that grid sample spacing and quadrat size are more important than the specific interpolation method chosen. Nomenclature: Common sunflower, Helianthus annuus L. HELAN; foxtail species, Setaria spp. SETSS; hemp dogbane, Apocynum cannabinum L. APCCA; velvetleaf, Abutilon theophrasti Medicus ABUTH; corn, Zea mays L.; soybean, Glycine max (L.) Merr.

Optical reflectance sensor for detecting plants

The reflectance ratio meter, an optical device for detecting weeds by measuring the ratio of reflected red and near-infrared light, is described. Experiments were conducted to evaluate the accuracy of the sensor in detecting weeds in the inter-row of growing soybeans and to characterize its sensitivity. Several methods for interpreting the signal were evaluated. The reflectance ratio meter is shown to have potential for estimating local weed populations.

Within-field soil heterogeneity effects on herbicide-mediated crop injury and weed biomass

Abstract Soil organic carbon (OC), clay content, water content, and pH often influence the bioactivity of soil-applied herbicides, and these soil properties can vary greatly within fields. The purpose of this work was to determine the influence of within-field soil heterogeneity on the efficacy of RPA-201772 where corn, shattercane, and velvetleaf were seeded as bioassays. An experimental approach was developed to quantify RPA-201772 dose–response across a range of soil conditions in an agricultural field. Based on a logistic model, crop injury was quantified with the I20 parameter, the dose eliciting 20% greenness reduction, using a series of photographic standards. Weed biomass was quantified with the I80 parameter, the dose eliciting 80% biomass reduction, relative to the untreated control. Crop and weed responses varied by two orders of magnitude. Significant correlation, as high as 0.76, was observed between measures of plant response and soil properties, namely particle size and OC. Furthermore, native velvetleaf spatial distribution at the study site was heterogeneous, and seedlings were observed in plots where seeded velvetleaf biomass was high. Spatial heterogeneity of soil affinity for herbicide results in differential weed fitness and contributes to weed “patchiness.” Nomenclature: RPA-201772, 5-cyclopropyl-4-(2-methylsulphonyl-4-trifluoromethyl-benzoyl)isoxazole; corn, Zea mays L.; shattercane, Sorghum bicolor (L.) Moench. SORBI; velvetleaf, Abutilon theophrasti Medik. ABUTH.

Two-Year Weed Seedling Population Responses to a Post-Emergent Method of Site-Specific Weed Management

Field experiments were conducted to determine how a site-specific weed management practice in Zea mays L. influenced the numerical and spatial distribution of a naturally occurring weed infestation in Z. mays and the succeeding Beta vulgaris L. crop. Compared to conventional broadcast herbicide applications, site-specific herbicide applications reduced herbicide load by 11.5 and 98.0% in two separate Z. mays fields. The broad range in outcomes was attributed to the spatial aggregation and density of target weed populations. While herbicide use was successfully reduced at field locations with low weed density, most survivors of multiple control tactics were in locations with the highest initial density. A greater understanding of interactions between weed management and weed density would increase the likelihood that site-specific weed management offers long-term improvements over conventional approaches.

Predicting Weed Species Occurrence Based on Site Properties and Previous Year's Weed Presence

Probabilities of Setaria spp., Solanum ptycanthum, Helianthus annuus and Abutilon theophrasti occurrence were predicted based on two site property factors and weed species presence in a previous year using logistic regression models. Weed seedling surveys were conducted just prior to post-emergence weed management in two grower-managed fields in the central Platte River Valley of Nebraska, USA at Alda in 1995 and 1996 and at Shelton in 1994, 1995 and 1997. Weed species density data were re-classified as present or absent at each pair of points on the sampling grid, representing quadrat locations either in the pre-emergence herbicide band or between the crop rows. Site property data were collected in March 1995 at Alda and March 1994 at Shelton. Using factor analysis, two independent factors were derived from correlated attributes of relative elevation, percent organic carbon, pH, nitrate, phosphate, and soil texture measured at Alda. Logistic regression models were estimated and parameterized for each weed species at Alda in 1996 based on the two factors (“topography and soil type” and “soil fertility status”) and weed species presence in 1995. Performance of these models for each weed species was evaluated using the independent data set from Shelton. Between and on crop row Setaria spp. and Solanum ptycanthum models described these populations at Alda. At Shelton, on row Setaria spp. occurrence and between row Solanum ptycanthum occurrence were adequately predicted. Helianthus annuus or Abutilon theophrasti occurrence was not well predicted even with knowledge of their presence in the previous year, probably as a result of low actual occurrence within a given year. Maps of predicted occurrence have value in directing weed scouting to field locations where the species is most likely to occur.

Red/near-infrared reflectance sensor system for detecting plants

Growing plants, soil types, and surfaces and residues on a soil surface have distinct natural light reflectances. These reflectance characteristics have been determined using current spectroradiometry technology. Detection of plants is possible based upon the distinct reflectance characteristics of plants, soil, and residues. An optical plant reflectance sensor was developed which utilizes a pair of red and near infrared sensitive photodetectors to measure the radiancy from the plant and soil. Another pair of sensors measures radiancy from a highly radiant reference surface to accommodate varying intensities of the natural light. The ratio of the target and reference radiancies is the target reflectance. Optical filters were used to select the spectral bandwidth sensitivities for the red and NIR photodetectors. The reflectance values were digitized for incorporation into a normalized difference index in order to provide a stronger indication that a live plant is present within the field of view of the sensor. This sensor system was combined with a microcontroller for activating a solenoid controlled spray nozzle on a single unit prototype spot agricultural sprayer.