Apple detection and instance segmentation in natural environments using an improved Mask Scoring R-CNN Model.

The accurate detection and segmentation of apples during growth stage is essential for yield estimation, timely harvesting, and retrieving growth information. However, factors such as the uncertain illumination, overlaps and occlusions of apples, homochromatic background and the gradual change in the ground color of apples from green to red, bring great challenges to the detection and segmentation of apples. To solve these problems, this study proposed an improved Mask Scoring region-based convolutional neural network (Mask Scoring R-CNN), known as MS-ADS, for accurate apple detection and instance segmentation in a natural environment. First, the ResNeSt, a variant of ResNet, combined with a feature pyramid network was used as backbone network to improve the feature extraction ability. Second, high-level architectures including R-CNN head and mask head were modified to improve the utilization of high-level features. Convolutional layers were added to the original R-CNN head to improve the accuracy of bounding box detection (bbox_mAP), and the Dual Attention Network was added to the original mask head to improve the accuracy of instance segmentation (mask_mAP). The experimental results showed that the proposed MS-ADS model effectively detected and segmented apples under various conditions, such as apples occluded by branches, leaves and other apples, apples with different ground colors and shadows, and apples divided into parts by branches and petioles. The recall, precision, false detection rate, and F1 score were 97.4%, 96.5%, 3.5%, and 96.9%, respectively. A bbox_mAP and mask_mAP of 0.932 and 0.920, respectively, were achieved on the test set, and the average run-time was 0.27 s per image. The experimental results indicated that the MS-ADS method detected and segmented apples in the orchard robustly and accurately with real-time performance. This study lays a foundation for follow-up work, such as yield estimation, harvesting, and automatic and long-term acquisition of apple growth information.

Deep Learning Based Automatic Grape Downy Mildew Detection.

Grape downy mildew (GDM) disease is a common plant leaf disease, and it causes serious damage to grape production, reducing yield and fruit quality. Traditional manual disease detection relies on farm experts and is often time-consuming. Computer vision technologies and artificial intelligence could provide automatic disease detection for real-time controlling the spread of disease on the grapevine in precision viticulture. To achieve the best trade-off between GDM detection accuracy and speed under natural environments, a deep learning based approach named YOLOv5-CA is proposed in this study. Here coordinate attention (CA) mechanism is integrated into YOLOv5, which highlights the downy mildew disease-related visual features to enhance the detection performance. A challenging GDM dataset was acquired in a vineyard under a nature scene (consisting of different illuminations, shadows, and backgrounds) to test the proposed approach. Experimental results show that the proposed YOLOv5-CA achieved a detection precision of 85.59%, a recall of 83.70%, and a mAP@0.5 of 89.55%, which is superior to the popular methods, including Faster R-CNN, YOLOv3, and YOLOv5. Furthermore, our proposed approach with inference occurring at 58.82 frames per second, could be deployed for the real-time disease control requirement. In addition, the proposed YOLOv5-CA based approach could effectively capture leaf disease related visual features resulting in higher GDE detection accuracy. Overall, this study provides a favorable deep learning based approach for the rapid and accurate diagnosis of grape leaf diseases in the field of automatic disease detection.

Soluble solid content and firmness index assessment and maturity discrimination of Malus micromalus Makino based on near-infrared hyperspectral imaging.

Malus micromalus Makino has great commercial and nutritional value. The regression and classification models were investigated by using near-infrared hyperspectral imaging (NIR-HSI) combined with chemometrics to improve the efficiency of non-destructive detection. The successive projections algorithm (SPA), interval random frog, and competitive adaptive reweighted sampling were employed to extract effective wavelengths sensitive to changes of soluble solid content (SSC) and firmness index (FI) information. Two types of assessment models based on full spectrum and effective wavelengths, namely partial least squares regression and extreme learning machine, were established to predict SSC and FI. In addition, the classification models based on the support vector machine improved by the grey wolf optimizer (GWO-SVM) and partial least squares discrimination analysis were constructed to differentiate maturity stage. The SPA-ELM and SPA-GWO-SVM models achieved satisfactory performance. The results illustrate that NIR-HSI is feasible for evaluation of the quality of Malus micromalus Makino.

Convex Non-Negative Matrix Factorization With Adaptive Graph for Unsupervised Feature Selection.

Unsupervised feature selection (UFS) aims to remove the redundant information and select the most representative feature subset from the original data, so it occupies a core position for high-dimensional data preprocessing. Many proposed approaches use self-expression to explore the correlation between the data samples or use pseudolabel matrix learning to learn the mapping between the data and labels. Furthermore, the existing methods have tried to add constraints to either of these two modules to reduce the redundancy, but no prior literature embeds them into a joint model to select the most representative features by the computed top ranking scores. To address the aforementioned issue, this article presents a novel UFS method via a convex non-negative matrix factorization with an adaptive graph constraint (CNAFS). Through convex matrix factorization with adaptive graph constraint, it can dig up the correlation between the data and keep the local manifold structure of the data. To our knowledge, it is the first work that integrates pseudo label matrix learning into the self-expression module and optimizes them simultaneously for the UFS solution. Besides, two different manifold regularizations are constructed for the pseudolabel matrix and the encoding matrix to keep the local geometrical structure. Eventually, extensive experiments on the benchmark datasets are conducted to prove the effectiveness of our method. The source code is available at: https://github.com/misteru/CNAFS.

Fine-Grained Grape Leaf Diseases Recognition Method Based on Improved Lightweight Attention Network.

Real-time dynamic monitoring of orchard grape leaf diseases can greatly improve the efficiency of disease control and is of great significance to the healthy and stable development of the grape industry. Traditional manual disease-monitoring methods are inefficient, labor-intensive, and ineffective. Therefore, an efficient method is urgently needed for real-time dynamic monitoring of orchard grape diseases. The classical deep learning network can achieve high accuracy in recognizing grape leaf diseases; however, the large amount of model parameters requires huge computing resources, and it is difficult to deploy to actual application scenarios. To solve the above problems, a cross-channel interactive attention mechanism-based lightweight model (ECA-SNet) is proposed. First, based on 6,867 collected images of five common leaf diseases of measles, black rot, downy mildew, leaf blight, powdery mildew, and healthy leaves, image augmentation techniques are used to construct the training, validation, and test set. Then, with ShuffleNet-v2 as the backbone, an efficient channel attention strategy is introduced to strengthen the ability of the model for extracting fine-grained lesion features. Ultimately, the efficient lightweight model ECA-SNet is obtained by further simplifying the network layer structure. The model parameters amount of ECA-SNet 0.5× is only 24.6% of ShuffleNet-v2 1.0×, but the recognition accuracy is increased by 3.66 percentage points to 98.86%, and FLOPs are only 37.4 M, which means the performance is significantly better than other commonly used lightweight methods. Although the similarity of fine-grained features of different diseases image is relatively high, the average F1-score of the proposed lightweight model can still reach 0.988, which means the model has strong stability and anti-interference ability. The results show that the lightweight attention mechanism model proposed in this paper can efficiently use image fine-grained information to diagnose orchard grape leaf diseases at a low computing cost.

Identification of Apple Leaf Diseases by Improved Deep Convolutional Neural Networks With an Attention Mechanism.

The accurate identification of apple leaf diseases is of great significance for controlling the spread of diseases and ensuring the healthy and stable development of the apple industry. In order to improve detection accuracy and efficiency, a deep learning model, which is called the Coordination Attention EfficientNet (CA-ENet), is proposed to identify different apple diseases. First, a coordinate attention block is integrated into the EfficientNet-B4 network, which embedded the spatial location information of the feature by channel attention to ensure that the model can learn both the channel and spatial location information of important features. Then, a depth-wise separable convolution is applied to the convolution module to reduce the number of parameters, and the h-swish activation function is introduced to achieve the fast and easy to quantify the process. Afterward, 5,170 images are collected in the field environment at the apple planting base of the Northwest A&F University, while 3,000 images are acquired from the PlantVillage public data set. Also, image augmentation techniques are used to generate an Apple Leaf Disease Identification Data set (ALDID), which contains 81,700 images. The experimental results show that the accuracy of the CA-ENet is 98.92% on the ALDID, and the average F1-score reaches .988, which is better than those of common models such as the ResNet-152, DenseNet-264, and ResNeXt-101. The generated test dataset is used to test the anti-interference ability of the model. The results show that the proposed method can achieve competitive performance on the apple disease identification task.

High-absorption grating-insulator-metal structures.

A thin grating-insulator-metal (GIM) structure consisting of a top metal grating layer on a dielectric layer and a bottom metal layer is proposed, which shows a broadband high absorption at a small thickness. This phenomenon is attributed to the appropriate effective surface permittivity of the top grating layer and the cavity resonance of the middle insulator layer. By optimizing the structural and material parameters, the materials of the GIM structure from top to bottom are Mn, Al2O3, and Mn with thicknesses of 10, 70, and 70 nm, respectively. The structure with these optimum parameters is fabricated and characterized, and an improved performance with absorption exceeding 90% in the visible region is obtained using Mn as the metal layers. The experimental results are in good agreement with the numerical values, depicting an ultrabroad absorption bandwidth. The conclusions presented here could have potential applications in optical devices used for optical displacement detection and visible light absorption.

Contrasting dorsal caudate functional connectivity patterns between frontal and temporal cortex with BMI increase: link to cognitive flexibility.

BACKGROUND: Obesity is associated with brain intrinsic functional reorganization. However, little is known about the BMI-related interhemispheric functional connectivity (IHFC) alterations, and their link with executive function in young healthy adults. METHODS: We examined voxel-mirrored homotopic connectivity (VMHC) patterns in 417 young adults from the Human Connectome Project. Brain regions with significant association between BMI and VMHC were identified using multiple linear regression. Results from these analyses were then used to determine regions for seed-voxel FC analysis, and multiple linear regression was used to explore the brain regions showing significant association between BMI and FC. The correlations between BMI-related executive function measurements and VMHC, as well as seed-voxel FC, were further examined. RESULTS: BMI was negatively associated with scores of Dimensional Change Card Sort Test (DCST) assessing cognitive flexibility (r = -0.14, p = 0.006) and with VMHC of bilateral inferior parietal lobule, insula and dorsal caudate. The dorsal caudate emerged as a nexus for BMI-related findings: greater BMI was associated with greater FC between caudate and hippocampus and lower FC between caudate and several prefrontal nodes (right inferior frontal gyrus, anterior cingulate cortex, and middle frontal gyrus). The FC between right caudate and left hippocampus was negatively associated with scores of DCST (r = -0.15, p = 0.0018). CONCLUSIONS: Higher BMI is associated with poorer cognitive flexibility performance and IHFC in an extensive set of brain regions implicated in cognitive control. Larger BMI was associated with higher caudate-medial temporal lobe FC and lower caudate-dorsolateral prefrontal cortex FC. These findings may have relevance for executive function associated with weight gain among otherwise healthy young adults.

Grape Leaf Disease Identification Using Improved Deep Convolutional Neural Networks.

Anthracnose, brown spot, mites, black rot, downy mildew, and leaf blight are six common grape leaf pests and diseases, which cause severe economic losses to the grape industry. Timely diagnosis and accurate identification of grape leaf diseases are decisive for controlling the spread of disease and ensuring the healthy development of the grape industry. This paper proposes a novel recognition approach that is based on improved convolutional neural networks for the diagnoses of grape leaf diseases. First, based on 4,023 images collected in the field and 3,646 images collected from public data sets, a data set of 107,366 grape leaf images is generated via image enhancement techniques. Afterward, Inception structure is applied for strengthening the performance of multi-dimensional feature extraction. In addition, a dense connectivity strategy is introduced to encourage feature reuse and strengthen feature propagation. Ultimately, a novel CNN-based model, namely, DICNN, is built and trained from scratch. It realizes an overall accuracy of 97.22% under the hold-out test set. Compared to GoogLeNet and ResNet-34, the recognition accuracy increases by 2.97% and 2.55%, respectively. The experimental results demonstrate that the proposed model can efficiently recognize grape leaf diseases. Meanwhile, this study explores a new approach for the rapid and accurate diagnosis of plant diseases that establishes a theoretical foundation for the application of deep learning in the field of agricultural information.

Precision Landing Test and Simulation of the Agricultural UAV on Apron.

Unmanned aerial vehicle (UAV) has been used to assist agricultural production. Precision landing control of UAV is critical for application of it in some specific areas such as greenhouses or livestock/poultry houses. For controlling UAV landing on a fixed or mobile apron/platform accurately, this study proposed an automatic method and tested it under three scenarios: (1) UAV landing at high operating altitude based on the GPS signal of the mobile apron; (2) UAV landing at low operating altitude based on the image recognition on the mobile apron; and (3) UAV landing progress control based on the fixed landing device and image detection to achieve a stable landing action. To verify the effectiveness of the proposed control method, apron at both stationary and mobile (e.g., 3 km/h moving speed) statuses were tested. Besides, a simulation was conducted for the UAV landing on a fixed apron by using a commercial poultry house as a model (135 L × 15 W × 3 H m). Results show that the average landing errors in high altitude and low altitude can be controlled within 6.78 cm and 13.29 cm, respectively. For the poultry house simulation, the landing errors were 6.22 ± 2.59 cm, 6.79 ± 3.26 cm, and 7.14 ± 2.41cm at the running speed of 2 km/h, 3 km/h, and 4 km/h, respectively. This study provides the basis for applying the UAV in agricultural facilities such as poultry or animal houses where requires a stricter landing control than open fields.

Age-Related Decreases in Interhemispheric Resting-State Functional Connectivity and Their Relationship With Executive Function.

Age-related alterations of functional brain networks contribute to cognitive decline. Current theories indicate that age-related intrinsic brain functional reorganization may be a critical marker of cognitive aging. Yet, little is known about how intrinsic interhemispheric functional connectivity changes with age in adults, and how this relates to critical executive functions. To address this, we examined voxel-mirrored homotopic connectivity (VMHC), a metric that quantifies interhemispheric communication, in 93 healthy volunteers (age range: 19-85) with executive function assessment using the Delis-Kaplan Executive Function System (D-KEFS) scales. Resting functional MRI data were analyzed to assess VMHC, and then a multiple linear regression model was employed to evaluate the relationship between age and the whole-brain VMHC. We observed age-related reductions in VMHC of ventromedial prefrontal cortex (vmPFC) and hippocampus in the medial temporal lobe subsystem, dorsal anterior cingulate cortex and insula in salience network, and inferior parietal lobule in frontoparietal control network. Performance on the color-word inhibition task was associated with VMHC of vmPFC and insula, and VMHC of vmPFC mediated the relationship between age and CWIT inhibition reaction times. The percent ratio of correct design scores in design fluency test correlated positively with VMHC of the inferior parietal lobule. The current study suggests that brain interhemispheric functional alterations may be a promising new avenue for understanding age-related cognitive decline.

A Machine Vision-Based Method for Monitoring Scene-Interactive Behaviors of Dairy Calf.

Requirements for animal and dairy products are increasing gradually in emerging economic bodies. However, it is critical and challenging to maintain the health and welfare of the increasing population of dairy cattle, especially the dairy calf (up to 20% mortality in China). Animal behaviors reflect considerable information and are used to estimate animal health and welfare. In recent years, machine vision-based methods have been applied to monitor animal behaviors worldwide. Collected image or video information containing animal behaviors can be analyzed with computer languages to estimate animal welfare or health indicators. In this proposed study, a new deep learning method (i.e., an integration of background-subtraction and inter-frame difference) was developed for automatically recognizing dairy calf scene-interactive behaviors (e.g., entering or leaving the resting area, and stationary and turning behaviors in the inlet and outlet area of the resting area) based on computer vision-based technology. Results show that the recognition success rates for the calf's science-interactive behaviors of pen entering, pen leaving, staying (standing or laying static behavior), and turning were 94.38%, 92.86%, 96.85%, and 93.51%, respectively. The recognition success rates for feeding and drinking were 79.69% and 81.73%, respectively. This newly developed method provides a basis for inventing evaluation tools to monitor calves' health and welfare on dairy farms.

Early Visual Detection of Wheat Stripe Rust Using Visible/Near-Infrared Hyperspectral Imaging.

Wheat stripe rust is one of the most important and devastating diseases in wheat production. In order to detect wheat stripe rust at an early stage, a visual detection method based on hyperspectral imaging is proposed in this paper. Hyperspectral images of wheat leaves infected by stripe rust for 15 consecutive days were collected, and their corresponding chlorophyll content (SPAD value) were measured using a handheld SPAD-502 chlorophyll meter. The spectral reflectance of the samples were then extracted from the hyperspectral images, using image segmentation based on a leaf mask. The effective wavebands were selected by the loadings of principal component analysis (PCA-loadings) and the successive projections algorithm (SPA). Next, the regression model of the SPAD values in wheat leaves was established, based on the back propagation neural network (BPNN), using the full spectra and the selected effective wavelengths as inputs, respectively. The results showed that the PCA-loadings⁻BPNN model had the best performance, which modeling accuracy (RC²) and validation accuracy (RP²) were 0.921 and 0.918, respectively, and the RPD was 3.363. The number of effective wavelengths extracted by this model accounted for only 3.12% of the total number of wavelengths, thus simplifying the models and improving the rate of operation greatly. Finally, the optimal models were used to estimate the SPAD of each pixel within the wheat leaf images, to generate spatial distribution maps of chlorophyll content. The visualized distribution map showed that wheat leaves infected by stripe rust could be identified six days after inoculation, and at least three days before the appearance of visible symptoms, which provides a new method for the early detection of wheat stripe rust.

Automatic detection and counting of urediniospores of Puccinia striiformis f. sp. tritici using spore traps and image processing.

The quantitative monitoring of airborne urediniospores of Puccinia striiformis f. sp. tritici (Pst) using spore trap devices in wheat fields is an important process for devising strategies early and effectively controlling wheat stripe rust. The traditional microscopic spore counting method mainly relies on naked-eye observation. Because of the great number of trapped spores, this method is labour intensive and time-consuming and has low counting efficiency, sometimes leading to huge errors; thus, an alternative method is required. In this paper, a new algorithm was proposed for the automatic detection and counting of urediniospores of Pst, based on digital image processing. First, images of urediniospores were collected using portable volumetric spore traps in an indoor simulation. Then, the urediniospores were automatically detected and counted using a series of image processing approaches, including image segmentation using the K-means clustering algorithm, image pre-processing, the identification of touching urediniospores based on their shape factor and area, and touching urediniospore contour segmentation based on concavity and contour segment merging. This automatic counting algorithm was compared with the watershed transformation algorithm. The results show that the proposed algorithm is efficient and accurate for the automatic detection and counting of trapped urediniospores. It can provide technical support for the development of online airborne urediniospore monitoring equipment.

Target tracking and 3D trajectory acquisition of cabbage butterfly (P. rapae) based on the KCF-BS algorithm.

Insect behaviour is an important research topic in plant protection. To study insect behaviour accurately, it is necessary to observe and record their flight trajectory quantitatively and precisely in three dimensions (3D). The goal of this research was to analyse frames extracted from videos using Kernelized Correlation Filters (KCF) and Background Subtraction (BS) (KCF-BS) to plot the 3D trajectory of cabbage butterfly (P. rapae). Considering the experimental environment with a wind tunnel, a quadrature binocular vision insect video capture system was designed and applied in this study. The KCF-BS algorithm was used to track the butterfly in video frames and obtain coordinates of the target centroid in two videos. Finally the 3D trajectory was calculated according to the matching relationship in the corresponding frames of two angles in the video. To verify the validity of the KCF-BS algorithm, Compressive Tracking (CT) and Spatio-Temporal Context Learning (STC) algorithms were performed. The results revealed that the KCF-BS tracking algorithm performed more favourably than CT and STC in terms of accuracy and robustness.

Realistic texture synthesis for point-based fruitage phenotype.

Although current 3D scanner technology can acquire textural images from a point model, visible seams in the image, inconvenient data acquisition and occupancy of a large space during use are points of concern for outdoor fruit models. In this paper, an SPSDW (simplification and perception based subdivision followed by down-sampling weighted average) method is proposed to balance memory usage and texture synthesis quality using a crop fruit, such as apples, as a research subject for a point-based fruit model. First, the quadtree method is improved to make splitting more efficient, and a reasonable texton descriptor is defined to promote query efficiency. Then, the color perception feature is extracted from the image for all pixels. Next, an advanced sub-division scheme and down-sampling strategy are designed to optimize memory space. Finally, a weighted oversampling method is proposed for high-quality texture mixing. This experiment demonstrates that the SPSDW method preserves the mixed texture more realistically and smoothly and preserves color memory up to 94%, 84.7% and 85.7% better than the two-dimesional processing, truncating scalar quantitative and color vision model methods, respectively.

Intrinsic brain subsystem associated with dietary restraint, disinhibition and hunger: an fMRI study.

Eating behaviors are closely related to body weight, and eating traits are depicted in three dimensions: dietary restraint, disinhibition, and hunger. The current study aims to explore whether these aspects of eating behaviors are related to intrinsic brain activation, and to further investigate the relationship between the brain activation relating to these eating traits and body weight, as well as the link between function connectivity (FC) of the correlative brain regions and body weight. Our results demonstrated positive associations between dietary restraint and baseline activation of the frontal and the temporal regions (i.e., food reward encoding) and the limbic regions (i.e., homeostatic control, including the hypothalamus). Disinhibition was positively associated with the activation of the frontal motivational system (i.e., OFC) and the premotor cortex. Hunger was positively related to extensive activations in the prefrontal, temporal, and limbic, as well as in the cerebellum. Within the brain regions relating to dietary restraint, weight status was negatively correlated with FC of the left middle temporal gyrus and left inferior temporal gyrus, and was positively associated with the FC of regions in the anterior temporal gyrus and fusiform visual cortex. Weight status was positively associated with the FC within regions in the prefrontal motor cortex and the right ACC serving inhibition, and was negatively related with the FC of regions in the frontal cortical-basal ganglia-thalamic circuits responding to hunger control. Our data depicted an association between intrinsic brain activation and dietary restraint, disinhibition, and hunger, and presented the links of their activations and FCs with weight status.

Semi-Supervised Fuzzy Clustering with Feature Discrimination.

Semi-supervised clustering algorithms are increasingly employed for discovering hidden structure in data with partially labelled patterns. In order to make the clustering approach useful and acceptable to users, the information provided must be simple, natural and limited in number. To improve recognition capability, we apply an effective feature enhancement procedure to the entire data-set to obtain a single set of features or weights by weighting and discriminating the information provided by the user. By taking pairwise constraints into account, we propose a semi-supervised fuzzy clustering algorithm with feature discrimination (SFFD) incorporating a fully adaptive distance function. Experiments on several standard benchmark data sets demonstrate the effectiveness of the proposed method.