The number of rice seedlings in the field is one of the main agronomic components for determining rice yield. This counting task, however, is still mainly performed using human vision rather than computer vision and is thus cumbersome and time-consuming. A fast and accurate alternative method of acquiring such data may contribute to monitoring the efficiency of crop management practices, to earlier estimations of rice yield, and as a phenotyping trait in breeding programs. In this paper, we propose an efficient method that uses computer vision to accurately count rice seedlings in a digital image. First, an unmanned aerial vehicle (UAV) equipped with red-green-blue (RGB) cameras was used to acquire field images at the seedling stage. Next, we use a regression network (Basic Network) inspired by a deep fully convolutional neural network to regress the density map and estimate the number of rice seedlings for a given UAV image. Finally, an improved version of the Basic Network, the Combined Network, is also proposed to further improve counting accuracy. To explore the efficacy of the proposed method, a novel rice seedling counting (RSC) dataset was built, which consisted of 40 images (where the number of seedlings varied between 3732 and 16,173) and corresponding manually-dotted annotations. The results demonstrated high average accuracy (higher than 93%) between counts according to the proposed method and manual (UAV image-based) rice seedling counts, and very good performance, with a high coefficient of determination (R²) (around 0.94). In conclusion, the results indicate that the proposed method is an efficient alternative for large-scale counting of rice seedlings, and offers a new opportunity for yield estimation. The RSC dataset and source code are available online.

The leaf area index (LAI) is a key parameter for describing crop canopy structure, and is of great importance for early nutrition diagnosis and breeding research. Light detection and ranging (LiDAR) is an active remote sensing technology that can detect the vertical distribution of a crop canopy. To quantitatively analyze the influence of the occlusion effect, three flights of multi-route high-density LiDAR dataset were acquired at two time points, using an Unmanned Aerial Vehicle (UAV)-mounted RIEGL VUX-1 laser scanner at an altitude of 15 m, to evaluate the validity of LAI estimation, in different layers, under different planting densities. The result revealed that normalized root-mean-square error (NRMSE) for the upper, middle, and lower layers were 10.8%, 12.4%, 42.8%, for 27,495 plants/ha, respectively. The relationship between the route direction and ridge direction was compared, and found that the direction of flight perpendicular to the maize planting ridge was better than that parallel to the maize planting ridge. The voxel-based method was used to invert the LAI, and we concluded that the optimal voxel size were concentrated on 0.040 m to 0.055 m, which was approximately 1.7 to 2.3 times of the average ground point distance. The detection of the occlusion effect in different layers under different planting densities, the relationship between the route and ridge directions, and the optimal voxel size could provide a guideline for UAV–LiDAR application in the crop canopy structure analysis.

The planting year of apple orchard not only determines the fruit output but also provides information for the governmental management of the fruit industry. However, considering that different orchards use different management and cultivation methods, this may result in some trees having similar outlines but different planting years, and it is, therefore, difficult to effectively determine the actual planting year based on textural or structural characteristics. Therefore, the monitoring method provided in this paper is not to monitor the growing year positively from the planting of orchard seedlings but to use time series remote sensing data to reverse determine the continuous growth age of each existing orchard. The city of Qixia, Shandong Province, China, was used as a case study. Firstly, the spatial distribution of apple orchards was accurately extracted using the Sentinel-2 normalized difference vegetation index (NDVI) spatiotemporally fused images and phenological vegetation information. Secondly, using region of interest (ROI) data for different vegetation types obtained from a field survey, NDVI time series were extracted from the Sentinel-2 NDVI spatiotemporally fused image. Among them, three characteristic phenological periods were selected, and the NDVI time series for apple orchards was used as a template to extract the apple orchard distribution area from 2000 to 2017. Then, the distribution area of apple orchards was defined as the area of interest in the planting year, combined with the Landsat NDVI time series image composed of three characteristic phenological periods each year from 2000 to 2017, and the apple orchard phenological curve. Subsequently, a Euclidean distance (ED) method was used to calculate the distribution area of apple orchards for each year between 2000 and 2017. Finally, a pixel-by-pixel inverse time series calculation method was used to obtain the planting year of apple orchards in the study area. This study provides a new way to accurately identify the planting year of apple orchards using satellite remote sensing images.

Recommender systems (RSs) have proven to be highly effective in guiding consumers towards well-informed purchase decisions for electronics. These systems can provide personalised recommendations that consider individual preferences, past purchases and current market trends by collecting and analysing massive amounts of consumer data. However, RSs have traditionally employed centralised storage of users consumption records and item interactions, which may potentially lead to privacy concerns. In particular, centralised data storage may prove unworkable in the future with the advent of regulations such as the General Data Protection Regulation. In turn, this can lead to an urgent need for decentralised recommendation frameworks for consumer electronics. In this study, we propose a federated learning recommender system (FRS) for the recommendation task in the consumer electronics industry. However, this is rather challenging due to its privacy protection, model scalability and personalisation requirements. First, the federated recommender system for consumer electronics (FRS-CE) adopts an outer product and two proposed feature fusion operations to construct an interaction map between users and items. Second, the FRS-CE uses a lightweight convolution operation to extract high-order features from the interaction map. Finally, the proposed model employs an adaptive aggregation mechanism to update the global model, which enhances the scalability of the system. Extensive experiments conducted on two real-world datasets have demonstrated the effectiveness of the FRS-CE in generating consumer electronics recommendations with privacy protection.

Mobile edge computing (MEC) considerably enhances the capabilities and performance of connected autonomous vehicles (CAVs) by deploying edge servers (ESs) on roadside units (RSUs) near CAVs, thereby ensuring low-latency services. Given the constrained and costly nature of ES resources (computing, storage, and bandwidth), equitable ES utilization is critical for CAV operations. However, fairness considerations are often overlooked in current budgeted edge server placement (ESP) strategies, potentially worsening resource imbalances and compromising user experience. This paper investigates the fairness-aware budgeted edge server placement (FESP) problem within RSUs, proving its NP-hardness. To address FESP, we first propose FESP-O, an integer programming-based optimal approach for small-scale problems, followed by FESP-APX, an approximation approach for large-scale scenarios that provides near-optimal solutions. We analyze the time complexity and approximation ratio of our proposed algorithms and validate their efficacy through experiments on real-world datasets. Extensive experimental results demonstrate significant performance improvements over baseline and state-of-the-art methods, indicating practical suitability and efficiency.