Identifying important factors influencing corn yield and grain quality variability using artificial neural networks

Soil, landscape and hybrid factors are known to influence yield and quality of corn (Zea mays L.). This study employed artificial neural network (ANN) analysis to evaluate the relative importance of selected soil, landscape and seed hybrid factors on yield and grain quality in two Illinois, USA fields. About 7 to 13 important factors were identified that could explain from 61% to 99% of the observed yield or quality variability in the study site-years. Hybrid was found to be the most important factor overall for quality in both fields, and for yield as well in Field 1. The relative importance of soil and landscape factors for corn yield and quality and their relationships differed by hybrid and field. Cation exchange capacity (CEC) and relative elevation were consistently identified as among the top four most important soil and landscape factors for both corn yield and quality in both fields in 2000. Aspect and Zn were among the top five most important factors in Fields 1 and 2, respectively. Compound topographic index (CTI), profile curvature and tangential curvature were, in general, not important in the study site-years. The response curves generated by the ANN models were more informative than simple correlation coefficients or coefficients in multiple regression equations. We conclude that hybrid was more important than soil and landscape factors for consideration in precision crop management, especially when grain quality was a management objective.     

Determination of visual quality of tomato paste using computerized inspection system and artificial neural networks

An artificial neural network (ANN) integrated computerized inspection system (CIS) was developed to determine tomato paste color in CIE L^∗, a^∗, and b^∗ color format and the number and size of dark specks which exist in the product. The usability of CIS in the determination of the number and the size of dark specks in tomato paste were investigated by comparing the results of CIS and human inspectors. While the inspectors had difficulties not only in determination of the specks having a diameter less than 0.2 mm but also in correct diameter measurement for all specks, the CIS had good determination and measurement capability. In 99 tomato paste samples, the number of the specks having diameter more than 0.2 mm were found by human inspectors and CIS as 233 and 235, respectively. However, the manual inspection gave inaccurate results for the diameter measurement of the specks. In the color evaluation of the tomato paste, strong correlations (R) were found between the results estimated from ANN-integrated CIS and those obtained from colorimeter (0.889, 0.958, 0.907 and 0.987 for L^∗, a^∗, b^∗ and a^∗/b^∗, respectively). The whole system is adapted to a graphical user interface (GUI) for use by a non-skilled person working in the tomato paste sector. While manual methods need approximately 5 min, GUI needs 20-25 s to determine, count and classify the dark specks and to measure the product color.     

Mapping sunflower yield as affected by Ridolfia segetum patches and elevation by applying evolutionary product unit neural networks to remote sensed data

Recent advances in remote sensing technology have triggered the need for highly flexible modelling methods to estimate several crop parameters in precision farming. The aim of this work was to determine the potential of evolutionary product unit neural networks (EPUNNs) for mapping in-season yield and forecasting systems of sunflower crop in a natural weed-infested farm. Aerial photographs were taken at the late vegetative (mid-May) growth stage. Yield, elevation and weed data were combined with multispectral imagery to obtain the dataset. Statistical and EPUNNs approaches were used to develop different yield prediction models. The results obtained using different EPUNN models show that the functional model and the hybrid algorithms proposed provide very accurate prediction compared to other statistical methodologies used to solve that regression problem.     

Analysis of Andean blackberry (Rubus glaucus) production models obtained by means of artificial neural networks exploiting information collected by small-scale growers in Colombia and publicly available meteorological data

The Andean blackberry (Rubus glaucus) is an important source of income in hillside regions of Colombia. However, growers have little reliable information on the factors that affect the development and yield of the crop, and therefore there is a dearth of information on how to effectively manage the crop. Site specific information recorded by small-scale producers of the Andean blackberry on their production systems and soils coupled with publicly available meteorological data was used to develop models of such production systems. Multilayer perceptrons and Self-Organizing Maps were used as computational models in the identification and visualization of the most important variables for modeling the production of Andean blackberry. Artificial neural networks were trained with information from 20 sites in Colombia where the Andean blackberry is cultivated. Multilayer perceptrons predicted with a reasonable degree of accuracy the production response of the crop. The soil depth, the average temperature, external drainage, and the accumulated precipitation of the first month before harvest were critical determinants of productivity. A proxy variable of location was used to describe overall differences in management between farmers groups. The use of this proxy indicated that, even under essentially similar environmental conditions, large differences in production could be assigned to management effects. The information obtained can be used to determine sites that are suitable for Andean blackberry production, and to transfer of management practices from sites of high productivity to sites with similar environmental conditions which currently have lower levels of productivity.