Information Processing in Agriculture最新文献

As a complement to traditional estimates of stem dimensions from numerical models, terrestrial light detection and ranging (Lidar) provides direct stem diameter and volume values using cylindrical models constructed from point clouds. This study used two approaches to estimate total stem volume using Lidar and compared them with two empirical equations, one used by the Forest Inventory Analysis in the Pacific Northwest (FIA-PNW) and one based on a taper equation. We fitted point clouds of 10 Douglas-fir with three sets of cylinder models that are distinguished by their segment length (i.e. 0.5 m, 1 m, and 2 m), then developed three taper equations based on the point-cloud-based diameter estimated previously. We estimated the total stem volume of the tree with eight models: six-point cloud-based (i.e. three taper and three cylinders) and two empirical. Finally, we used simulations to extrapolate the volume estimations of various methods for different diameters at breast height (DBH) classes. We found that all the point-cloud-based taper equations were similar in their performance ($R^2=0.94$, RMSE = 4.6 cm) and produced mean volume estimates greater than mean estimates of the existing models. The cylinder models produced 11–16% greater mean volume estimates than the FIA-PNW estimate, with the 0.5 m segment length producing the greatest values, followed by the 1 m and 2 m segment length. The simulated data suggested that the mean volume estimates of a given DBH class are different when using different computation methods. ANOVA revealed a combined effect of the computation methods and the DBH class on the mean volume estimates. We conclude that the point-cloud-based taper equations, after being symmetrically calibrated, would be consistent with the regional stem volume estimates, whereas the cylinder models would be better in estimating individual stem volume. When constructing Lidar-based cylinder models in future applications, cylinder segment length would need to be adjusted to the length and DBH of the stem, as well as to the objectives of the research.

Several internet-based surveillance systems have been created to monitor the web for animal health surveillance. These systems collect a large amount of news dealing with outbreaks related to animal diseases. Automatically identifying news articles that describe the same outbreak event is a key step to quickly detect relevant epidemiological information while alleviating manual curation of news content. This paper addresses the task of retrieving news articles that are related in epidemiological terms. We tackle this issue using text mining and feature fusion methods. The main objective of this paper is to identify a textual representation in which two articles that share the same epidemiological content are close. We compared two types of representations (i.e., features) to represent the documents: (i) morphosyntactic features (i.e., selection and transformation of all terms from the news, based on classical textual processing steps) and (ii) lexicosemantic features (i.e., selection, transformation and fusion of epidemiological terms including diseases, hosts, locations and dates). We compared two types of term weighing (i.e., Boolean and TF-IDF) for both representations. To combine and transform lexicosemantic features, we compared two data fusion techniques (i.e., early fusion and late fusion) and the effect of features generalisation, while evaluating the relative importance of each type of feature. We conducted our analysis using a corpus composed of a subset of news articles in English related to animal disease outbreaks. Our results showed that the combination of relevant lexicosemantic (epidemiological) features using fusion methods improves classical morphosyntactic representation in the context of disease-related news retrieval. The lexicosemantic representation based on TF-IDF and feature generalisation (F-measure = 0.92, r-precision = 0.58) outperformed the morphosyntactic representation (F-measure = 0.89, r-precision = 0.45), while reducing the features space. Converting the features into lower granular features (i.e., generalisation) contributed to improving the results of the lexicosemantic representation. Our results showed no difference between the early and late fusion approaches. Temporal features performed poorly on their own. Conversely, spatial features were the most discriminative features, highlighting the need for robust methods for spatial entity extraction, disambiguation and representation in internet-based surveillance systems.

In response to the challenges in providing real-time extraction of crop biophysical signatures, computer vision in computational crop phenotyping highlights the opportunities of computational intelligence solutions. Shadow and angular brightness due to the presence of photosynthetic light unevenly illuminate crop canopy. In this study, a novel vegetation index named artificial bee colony-optimized visible band oblique dipyramid greenness index (vODGI_abc) was proposed to enhance vegetation pixels by correcting the saturation and brightness levels, and the ratio of visible RGB reflectance intensities. Consumer-grade smartphone was used to acquire indoor and outdoor aquaponic lettuce images daily for full 6-week crop life cycle. The introduced saturation rectification coefficient (Ω), value rectification coefficient (ν), green–red wavelength adjustment factor (α), and green–blue wavelength adjustment factor (β) on the original triangular greenness index resulted in 3D canopy reflectance spectrum with two oblique tetrahedrons formed by connecting the vertices of visible RGB band reflectance and maximum wavelength point map to corresponding saturation and value of lettuce-captured images. Hybrid neighborhood component analysis (NCA), minimum redundancy maximum relevance (MRMR), Pearson’s correlation coefficient (PCC), and analysis of variance (ANOVA) weighted most of the canopy area, energy, and homogeneity. Strong linear relationships were exhibited by using vODGI_abc in estimating lettuce crop fresh weight, height, number of spanning leaves, leaf area index, and growth stage with R² values of 0.936 8 for InceptionV3, 0.957 4 for ResNet101, 0.961 2 for ResNet101, 0.999 9 for Gaussian processing regression, and accuracy of 88.89% for ResNet101, respectively. This low-cost approach on developing greenness index for biophysical signatures estimation proved to be more accurate than the previously established triangular greenness index (TGI) using RGB smartphone camera.

Classification of sheep behaviour from a sequence of tri-axial accelerometer data has the potential to enhance sheep management. Sheep behaviour is inherently imbalanced (e.g., more ruminating than walking) resulting in underperforming classification for the minority activities which hold importance. Existing works have not addressed class imbalance and use traditional machine learning techniques, e.g., Random Forest (RF). We investigated Deep Learning (DL) models, namely, Long Short Term Memory (LSTM) and Bidirectional LSTM (BLSTM), appropriate for sequential data, from imbalanced data. Two data sets were collected in normal grazing conditions using jaw-mounted and ear-mounted sensors. Novel to this study, alongside typical single classes, e.g., walking, depending on the behaviours, data samples were labelled with compound classes, e.g., walking_grazing. The number of steps a sheep performed in the observed 10 s time window was also recorded and incorporated in the models. We designed several multi-class classification studies with imbalance being addressed using synthetic data. DL models achieved superior performance to traditional ML models, especially with augmented data (e.g., 4-Class + Steps: LSTM 88.0%, RF 82.5%). DL methods showed superior generalisability on unseen sheep (i.e., F1-score: BLSTM 0.84, LSTM 0.83, RF 0.65). LSTM, BLSTM and RF achieved sub-millisecond average inference time, making them suitable for real-time applications. The results demonstrate the effectiveness of DL models for sheep behaviour classification in grazing conditions. The results also demonstrate the DL techniques can generalise across different sheep. The study presents a strong foundation of the development of such models for real-time animal monitoring.

Energy consumption in the agricultural sector is significant, reaching 20% of the total energy consumption in China. Agricultural Energy Internet, an important extension of Energy Internet in the agricultural field, significantly contributes to agricultural modernization. Key technologies of Agricultural Energy Internet are vital factors supporting its development. This article systematically reviews the key technologies of Agricultural Energy Internet for two areas: agriculture and fishery. The working mechanisms and power consumption characteristics of some state-of-the-art new-energy agricultural intelligent equipment are described. In addition, the principles and profit methods underlying the agro-industrial complementary operation model are introduced. Moreover, against the Agricultural Energy Internet background, the development trends of some state-of-the-art new energy agricultural intelligent equipment, agro-industrial complementary, and carbon–neutral technology are proposed in this paper, providing novel perspectives on the promotion of the development of Agricultural Energy Internet and related technological innovation research. An unmanned farm is the main form of the future agricultural system, which is powered by the Agricultural Energy Internet based on smart agriculture and a smart grid. It will become the inevitable trend of modern agriculture to replace oil agriculture with electric farms. The electricity in farming is mainly generated by renewable energy. Renewable energy power generation has low carbon emissions and is the future direction for the development of agricultural energy systems. In addition, the Internet of Things will be further strengthened to realize automation and intelligence of agricultural energy systems.