Instance segmentation of fruit tree canopies from images acquired by unmanned aerial vehicles (UAVs) is of significance for the precise management of orchards. Although deep learning methods have been widely used in the fields of feature extraction and classification, there are still phenomena of complex data and strong dependence on software performances. This paper proposes a deep learning-based instance segmentation method of litchi trees, which has a simple structure and lower requirements for data form. Considering that deep learning models require a large amount of training data, a labor-friendly semi-auto method for image annotation is introduced. The introduction of this method allows for a significant improvement in the efficiency of data pre-processing. Facing the high requirement of a deep learning method for computing resources, a partition-based method is presented for the segmentation of high-resolution digital orthophoto maps (DOMs). Citrus data is added to the training set to alleviate the lack of diversity of the original litchi dataset. The average precision (AP) is selected to evaluate the metric of the proposed model. The results show that with the help of training with the litchi-citrus datasets, the best AP on the test set reaches 96.25%.

Citrus Huanglongbing (HLB), also named citrus greening disease, occurs worldwide and is known as a citrus cancer without an effective treatment. The symptoms of HLB are similar to those of nutritional deficiency or other disease. The methods based on single-source information, such as RGB images or hyperspectral data, are not able to achieve great detection performance. In this study, a multi-modal feature fusion network, combining a RGB image network and hyperspectral band extraction network, was proposed to recognize HLB from four categories (HLB, suspected HLB, Zn-deficient, and healthy). Three contributions including a dimension-reduction scheme for hyperspectral data based on a soft attention mechanism, a feature fusion proposal based on a bilinear fusion method, and auxiliary classifiers to extract more useful information are introduced in this manuscript. The multi-modal feature fusion network can effectively classify the above four types of citrus leaves and is better than single-modal classifiers. In experiments, the highest accuracy of multi-modal network recognition was 97.89% when the amount of data was not very abundant (1,325 images of the four aforementioned types and 1,325 pieces of hyperspectral data), while the single-modal network with RGB images only achieved 87.98% recognition and the single-modal network using hyperspectral information only 89%. Results show that the proposed multi-modal network implementing the concept of multi-source information fusion provides a better way to detect citrus HLB and citrus deficiency.

Citrus psyllid is the only insect vector of citrus Huanglongbing (HLB), which is the most destructive disease in the citrus industry. There is no effective treatment for HLB, so detecting citrus psyllids as soon as possible is the key prevention measure for citrus HLB. It is time-consuming and laborious to search for citrus psyllids through artificial patrol, which is inconvenient for the management of citrus orchards. With the development of artificial intelligence technology, a computer vision method instead of the artificial patrol can be adopted for orchard management to reduce the cost and time. The citrus psyllid is small in shape and gray in color, similar to the stem, stump, and withered part of the leaves, leading to difficulty for the traditional target detection algorithm to achieve a good recognition effect. In this work, in order to make the model have good generalization ability under outdoor light condition, a high-definition camera to collect data set of citrus psyllids and citrus fruit flies under natural light condition was used, a method to increase the number of small target pests in citrus based on semantic segmentation algorithm was proposed, and the cascade region-based convolution neural networks (R-CNN) (convolutional neural network) algorithm was improved to enhance the recognition effect of small target pests using multiscale training, combining CBAM attention mechanism with high-resolution feature retention network high-resoultion network (HRNet) as feature extraction network, adding sawtooth atrous spatial pyramid pooling (ASPP) structure to fully extract high-resolution features from different scales, and adding feature pyramid networks (FPN) structure for feature fusion at different scales. To mine difficult samples more deeply, an online hard sample mining strategy was adopted in the process of model sampling. The results show that the improved cascade R-CNN algorithm after training has an average recognition accuracy of 88.78% for citrus psyllids. Compared with VGG16, ResNet50, and other common networks, the improved small target recognition algorithm obtains the highest recognition performance. Experimental results also show that the improved cascade R-CNN algorithm not only performs well in citrus psylla identification but also in other small targets such as citrus fruit flies, which makes it possible and feasible to detect small target pests with a field high-definition camera.

Objective: Bone suppression of chest radiograph holds great promise to improve the localization accuracy in Image-Guided Radiation Therapy (IGRT). However, data scarcity has long been considered as the prime culprit of developing Convolutional Neural Networks (CNNs) models for the task of bone suppression. In this study, we explored the effectiveness of various data augmentation techniques for the task of bone suppression. Methods: In this study, chest radiograph and bone-free chest radiograph are derived from 59 high-resolution CT scans. Two CNN models (U-Net and Generative Adversarial Network (GAN)) were adapted to explore the effectiveness of various data augmentation techniques for bone signal suppression in the chest radiograph. Lung radiograph and bone-free radiograph were used as the input and target label, respectively. Impacts of six typical data augmentation techniques (flip, cropping, noise injection, rotation, shift and zoom) on model performance were investigated. A series of statistical evaluating metrics, including Peak Signal-To-Noise Ratio (PSNR), Structural Similarity (SSIM) and Mean Absolute Error (MAR), were deployed to comprehensively assess the prediction performance of the two networks under the six data augmentation strategies. Quantitative comparative results showed that different data augmentation techniques exhibited a varying degree of influence on the performance of CNN models in the task of CR bone signal suppression. Results: For the U-Net model, flips, rotation (10 to 20 degrees), all the shifts, and zoom (1/8) resulted in improved model prediction accuracy. By contrast, other studied augmentation techniques showed adverse impacts on the model performance. For the GAN model, it was found to be more sensitive to the studied augmentation techniques than the U-Net. Vertical flip was the only augmentation method that yielded enhanced model performance. Conclusion: In this study, we found that different data augmentation techniques resulted in a varying degree of impacts on the prediction performance of U-Net and GAN models in the task of bone suppression in CR. However, it remains challenging to determine the optimal parameter settings for each augmentation technique. In the future, a more comprehensive evaluation is still warranted to evaluate the effectiveness of different augmentation techniques in task-specific image synthesis.