查看更多>>摘要：? 2022 Elsevier B.V.The accurate measurement of environmentally-relevant gases such as ammonia (NH3) is increasingly important, especially in agricultural science. Accurate and continuous measurement of gases is essential for evaluating NH3 and other gases as resource-based indicators of air quality in forced ventilation barns and for determining emission rates. To better assess measuring devices under different barn conditions (e.g. relative humidity or concentration range) self-conducted tests should be implemented at the laboratory scale before, during and after (experimental) measurements in the barn, independent of the manufacturer's specifications. Therefore, in this study, a gas calibration unit was set up for measurements at the laboratory scale. Moreover, measurement protocols were developed to investigate the accuracy at different gas concentrations and relative humidity levels and the responsiveness to rapid gas concentration changes. Measurements were performed with photoacoustic gas analysers (INNOVA 1412) and transmitters with electrochemical gas sensors (Polytron 8100 and C300). In addition to the presentation of results of the devices used in this case, this study should above all offer suggestions for quality monitoring to test sensors more intensively at the laboratory scale in order to be able to use them in a more targeted manner according to their optimal suitability.
查看更多>>摘要：? 2022 Elsevier B.V.Utility All-Terrain Vehicles (ATVs) are commonly used on farms for agricultural activities such as carrying implements, applying fertilizers, and transportation. Among all agricultural ATV incidents on farms, rollover crushes are the leading cause of fatalities. Crush Protection Devices (CPDs) are passive safety structures that potentially decrease the severity of injuries and the number of deaths in agricultural ATV rollover incidents. However, there are contradictions in the results of previous studies regarding the effectiveness of CPDs in protecting the operator in rollover incidents. The effectiveness of CPDs should be evaluated experimentally with an autonomous ATV, allowing repeatable and systematic rollover tests. This study aims at developing three automatic control systems on an ATV for future rollover simulations: (1) Global Positioning System (GPS)- based navigation system that allows autonomous steering, (2) remote cruise control module for keeping consistent riding speeds, and (3) remote braking system for safety in future unmanned ATV tests.The performance of each system was examined by operating the autonomous ATV in an outdoor testing terrain. The term autonomy in this study refers to the autonomous steering system, along with the remotely controlled speed and braking systems. Results showed that the developed systems performed within the acceptable range for conducting future rollover tests accurately and safely. Application of the autonomous ATV expects to deliver repeatable results in rollover simulations, and thus increases the reliability of CPD evaluation.
查看更多>>摘要：? 2022 The AuthorsRegular plumage condition assessment in laying hens is essential to monitor the hens’ welfare status and to detect the occurrence of feather pecking activities. However, in commercial farms this is a labor-intensive, manual task. This study proposes a novel approach for automated plumage condition assessment using computer vision and deep learning. It presents ChickenNet, an end-to-end convolutional neural network that detects hens and simultaneously predicts a plumage condition score for each detected hen. To investigate the effect of input image characteristics, the method was evaluated using images with and without depth information in resolutions of 384 × 384, 512 × 512, 896 × 896 and 1216 × 1216 pixels. Further, to determine the impact of subjective human annotations, plumage condition predictions were compared to manual assessments of one observer and to matching annotations of two observers. Among all tested settings, performance metrics based on matching manual annotations of two observers were equal or better than the ones based on annotations of a single observer. The best result obtained among all tested configurations was a mean average precision (mAP) of 98.02% for hen detection while 91.83% of the plumage condition scores were predicted correctly. Moreover, it was revealed that performance of hen detection and plumage condition assessment of ChickenNet was not generally enhanced by depth information. Increasing image resolutions improved plumage assessment up to a resolution of 896 × 896 pixels, while high detection accuracies (mAP > 0.96) could already be achieved using lower resolutions. The results indicate that ChickenNet provides a sufficient basis for automated monitoring of plumage conditions in commercial laying hen farms.
查看更多>>摘要：? 2022 Elsevier B.V.Machine learning methods are increasingly used in analyzing remotely sensed data and studying different aspects of agricultural production. In particular, several of these flexible models are widely adopted to predict regional crop yield during or after the growing season. However, most existing models cannot be applied when dealing with functional covariates. In this paper, an approach based on multidimensional scaling is proposed to generate a set of artificial covariates from empirical density functions of different phenomena captured within specific administrative boundaries through satellites. In contrast to traditional aggregation methods, this approach is designed to reduce the loss of information associated with the use of summary statistics as covariates. The proposed methodology is applied to NASA remote sensing data, combined with information from surveys and USDA's end-of-season county estimates, to study the prediction accuracy of different crop-yield models for three major crops in North Dakota.
查看更多>>摘要：? 2022Remote sensing image is becoming an increasingly popular tool for crop lodging detection because it conveniently provides features for building machine learning models and predicting lodging. However, difficulties in interpreting machine learning models and their predictions limit the confidence of using remote sensing images to detect lodging. In addition, the lodging datasets used for modeling are difficult to balance under natural conditions. Designing a robust and interpretable classification model for the detection of lodging in an imbalanced distribution dataset poses a particularly difficult challenge. In this study, visible and multi-spectral images were collected with a UAV to extract relevant features from remote sensing images. In a preliminary step, Synthetic Minority Oversampling Technique (SMOTE) and Edited Nearest Neighbors (ENN) method were used to treat imbalanced datasets. The SMOTE-ENN-XGBoost model is proposed for the efficient identification of maize lodging at the plot scale. The SMOTE-ENN-XGBoost model achieved an F1-score of 0.930 and a recall of 0.899 on a testing set, suggesting that it can be used for modeling lodging detection. Additionally, the SHapley Additive exPlanations (SHAP) approach was employed to interpret the identification and prioritization of features that determine lodging classification and activity prediction. The results showed that canopy structure and textural features are relatively stable compared with spectral features, which are susceptible to the external environment when modeling is employed to detect lodging. This work also showed that canopy structural, spectral, and textural information should be considered simultaneously rather than separately when detecting crop lodging in a crop breeding program in order to prevent differences in expression controlled by the interaction between genotype and environment obscuring the change in a single feature before and after lodging. For practical applications of machine learning models in crop lodging detection, such insights are of critical relevance. Taken together, the results of this study encourage further applications of remote sensing techniques to build interpretable machine learning models.
查看更多>>摘要：? 2022Weeds compete with productive crops for soil, nutrients and sunlight and are therefore a major contributor to crop yield loss, which is why safer and more effective herbicide products are continually being developed. Digital evaluation tools to automate and homogenize field measurements are of vital importance to accelerate their development. However, the development of these tools requires the generation of semantic segmentation datasets, which is a complex, time-consuming and not easily affordable task. In this paper, we present a deep learning segmentation model that is able to distinguish between different plant species at the pixel level. First, we have generated three extensive datasets targeting one crop species (Zea mays), three grass species (Setaria verticillata, Digitaria sanguinalis, Echinochloa crus-galli) and three broadleaf species (Abutilon theophrasti, Chenopodium albums, Amaranthus retroflexus). The first dataset consists of real field images that were manually annotated. The second dataset is composed of images of plots where only one species is present at a time and the third type of dataset was synthetically generated from images of individual plants mimicking the distribution of real field images. Second, we have proposed a semantic segmentation architecture by extending a PSPNet architecture with an auxiliary classification loss to aid model convergence. Our results show that the network performance increases when supplementing the real field image dataset with the other types of datasets without increasing the manual annotation effort. More specifically, the use of the real field dataset obtains a Dice-S?ensen Coefficient (DSC) score of 25.32. This performance increases when this dataset is combined with the single-species class dataset (DSC=47.97) or the synthetic dataset (DSC=45.20). As for the proposed model, the ablation method shows that by removing the proposed auxiliary classification loss, the segmentation performance decreases (DSC=45.96) compared to the proposed architecture method (DSC=47.97). The proposed method shows better performance than the current state of the art. In addition, the use of proposed single-species or synthetic datasets can double the performance of the algorithm than when using real datasets without additional manual annotation effort.
查看更多>>摘要：? 2022For robotic harvesting of table grape in clusters usually by gripping and cutting the peduncle, the accurate cutting point positioning on the peduncle is crucial. In this paper, a new method for accurate positioning of picking-point based on the combination of far-view and near-view depth data features for horizontal-trellis cultivated grape was proposed. First, a far-near combination strategy for picking-point positioning which makes full use of the features of grape cluster and horizontal-trellis environment obtained from depth point cloud data was put forward. Then the special “eye-under-finger” structure to meet the needs of far-near combination was proposed, and, three key points of far-view point, near-view point and picking-point that determine the hand-eye path for far-near combination were defined. Finally, the far-near combined method composed of grape cluster detection in far view, far-near viewing switching and picking-point positioning based on depth data histogram in near view was established, which was realized by selecting the grape cluster bottom as the key clue. In field experiment of picking-point positioning in near view, the average running time of the algorithm is 0.29 s, and only 5 samples in total 100 failed to achieve accurate positioning. In lab experiments of far-near combined picking-point positioning, the success rate of accurate picking-point positioning reached 100%. This method is hopeful to overcome the deficiency of black-box processing of traditional image processing methods.
查看更多>>摘要：? 2022 Elsevier B.V.Automatic identification of individual animals is an important step towards achieving accurate breeding histories, significant contributions to breeding and genetic management programmers. Currently, a different type of tags, tattoos, paint brands and microchips are used to uniquely identify livestock animals. However, the manual identification system is time-consuming, expensive and unreliable. In this paper, we present a deep learning approach that aims to fully automated pipeline for face detection and recognition of goats. Due to the high similarity and the lack of adequate dataset this problem is more complex than human face recognition. We composed two different publicly available datasets for detection and recognition. State-of-the-art convolutional neural networks (CNN) model are trained on this dataset. To evaluate the robustness of our approach, we compared it with different face recognition methods. The results show better performance with an accuracy of 96.4%. Furthermore, this paper reports 93%, 83%, 92% and 85% detection accuracy (average precision) for face, eye, nose and ear, respectively. The findings of this research could be helpful to improve animal health and welfare, individual monitoring, activity monitoring and phenotypic data collection. All the dataset and the related outcome are publicly available (https://doi.org/10.17632/4skwhnrscr.2).
查看更多>>摘要：? 2022 Elsevier B.V.With the widespread application of machine vision technology in agriculture, the intelligent management of banana orchards is urgent. Accurate detection of banana bunches and stalks is a precondition for orchard yield estimation and automatic harvesting. In a complex banana orchard environment, banana bunches and stalks are similar to the leaves in color, and banana stalks are similar to the petiole in texture, making the detection of banana bunches and stalks in banana orchards challenging. This study proposes an accurate and fast multiclass detection method for banana bunches and stalks. A regular RGB camera was used to collect images. The well-known YOLOv4 network was used to detect the banana bunches and stalks, and the input image resolution was discussed by training and comparison. The banana bunch and stalk detection model showed excellent reliability and generalization ability in different illumination and occlusion scenarios. The AP of the banana bunch and stalk detection was 99.55% and 87.82%, respectively, and the mAP of the detection model was 93.69%. The average execution time was 44.96 ms. The detection of small-sized banana bunches and stalks was discussed, and its significance in banana orchard applications was analyzed. The experimental results show that the fast real-time detection of banana bunches and stalks in the natural environment is helpful for the intelligent management of banana orchards.
查看更多>>摘要：? 2022 Elsevier B.V.Accurate estimation of sheep feed intake is essential for optimal pasture management and understanding the patterns of grass-livestock ecosystems. Although many studies have been reported, fewer studies have comprehensively considered animal body weight, grass species, and grass moisture content. This study designed a 29-test experiment with three classes of sheep weight, two grass species, and three levels of grass moisture content combinations. Segment samples were constructed from each test, and many explanatory variables (44 in total) associated with chewing were extracted from each segment sample. Correlations between each explanatory variable and intake were examined under every test, and the slopes of their regression lines were also recorded. A statistical analysis was then used to reveal the effects of sheep bodyweight class, grass species, and grass moisture content class on the slopes of the explanatory variables. On the test sample set (one test sample corresponding to one test), intake was predicted using a single explanatory variable, all explanatory variables, all explanatory variables and factor variables (sheep weight, grass moisture content, grass NDF content, grass ADF content). The results showed that the chew_ZCTOverLegth variable (number of chewing waveforms over the mean divided by waveform length and then accumulated over all chews) and intake had the strongest correlation with a mean R2 score of 0.8796. The influence of grass moisture content class on the slope of the variable was approximately equal to that of the sheep weight class and less than that of the differences in individual sheep. In the test sample set, the LogMel_power variable (sum of squares of waveform log-Mel) was the best single explanatory variable for predicting intake (when the type of intake is fresh grass matter). Adding factor variables significantly increased the R2 score of the model and reduced the number of variables used. Introducing sheep bodyweight to intake further increased the R2 score of the model and reduced the number of variables used. When the type of intake was FMI_W (fresh matter multiplied by the square of the sheep bodyweight value), intake was accurately estimated using all explanatory and factor variables with an R2 score of 0.9734. This result demonstrates that chewing-derived acoustic variables acoustically can accurately estimate intake despite the complexity of the experimental conditions. This study will make an outstanding contribution to the development of general and accurate models for estimating intake.
NSTL
Elsevier