Many spectral unmixing approaches ranging from geometry, algebra to statistics have been proposed, in which nonnegative matrix factorization (NMF)-based ones form an important family. The original NMF-based algorithm loses the and spatial information between mixed pixels when stacking responses of into observed matrix. Therefore, various constrained NMF methods are developed impose structure, spectral-spatial joint structure enforce estimated endmembers abundances preserve these structures....

Data samples of complicated geometry and nonlinear separability are considered as common challenges to clustering algorithms. In this article, we first construct Mahalanobis distance in the kernel space then propose a novel fuzzy model with kernelized distance, namely KMD-FC. The key contributions KMD-FC include: first, construction KMD matrix is innovatively transformed from Euclidean matrix, which able effectively avoid problem “curse dimensionality” posed by explicitly calculating sample...

The immense representation power of deep learning frameworks has kept them in the spotlight hyperspectral image (HSI) classification. Graph Convolutional Neural Networks (GCNs) can be used to compensate for lack spatial information (CNNs). However, most GCNs construct graph data structures based on pixel points, which requires construction neighborhood matrices all data. Meanwhile, setting similarity relations structure is not fully applicable HSIs. To make network more compatible with HSIs,...

The air-suction precision seeder for small seeds is a planting machine, characterized by precision, high efficiency, and ease of operation, that uses air suction technology to sow grain at set intervals depths into the soil. However, forced vibration, enhanced increase in operating speed, affects seeding accuracy limits efficiency. To study influence vibration conditions on seed performance seeder, we developed computational fluid dynamics–discrete element coupling method construct...

Hyperspectral images (HSIs) obtained from remote sensing contain abundant information of ground objects, and precise analysis landcover depends on effective efficient classification HSIs into homogeneous regions. While many advanced algorithms have been developed for HSI classification, it is a challenge an algorithm to achieve good balance between its effectiveness efficiency due the high dimensionality insufficient labeled training samples. By taking both rich spectral features spatially...

As the quality of life rises, demand for flowers has increased significantly, leading to higher expectations flower sorting system efficiency and speed. This paper presents a real-time, high-precision end-to-end method, which can complete three key tasks in system: localization, classification, grading. In order improve challenging maturity detection, red–green–blue depth (RGBD) images were captured. The multi-task multi-dimension-You Only Look Once (MTMD-YOLO) network was proposed these an...

Accurate early lung cancer detection is essential towards precision oncology and would effectively improve the patients' survival rate. In this work, we explore capacity by learning from deep spatial features. A 3D CNN network architecture constructed with segmented CT volumes as training testing samples. The new model extracts projects features to following hidden layers, which preserves temporal relations between neighboring slices. well-built consists of 11 layers generates 12,544 neurons...

Considering the difficulty of broadcasting in small plots and complex terrain South China, this research aimed to explore a new efficient way figure out advisable operation parameters by using hollow 12-axis, rotor-wing Unmanned Aerial Vehicle (UAV) which is typically made up four groups solid support structure, with each consisting three axes two rotor wings. A 3.7 L reverse pyramid-shape seed hopper 60 mm×13 mm rectangular outlet at bottom was designed realize self-gravity seeding. Rice...

The instability of the motion layer fertilizer particle influences precision and accuracy amount application in working process fluted-roller applicator. characteristics were investigated, a systematic scheme was proposed for designing structures fertilizer-filling cavity, surface fertilizer-delivery cavity. key parameters structure studied by discrete element method. It figured out that to ensure smooth operation applicator, angle 105°, fertilizer-contact should be large than 100°,...

Existing devices for dry direct-seeded rice with film mulching in northern China have limitations such as imprecise sowing, unadjustable sowing depth, and seeding device blocking. In this regard, study proposes a combined method of ‘mini shovel + telescopic pipe’ mulching. A precision seeder was developed through theoretical calculations, discrete element modelling (DEM) simulations, field experiments. The configuration diameter the rollers were obtained. Twelve pipes evenly distributed on...

With the continuous development of hyperspectral image technology and deep learning methods in recent years, an increasing number classification models have been proposed. However, due to numerous spectral dimensions images, most suffer from issues such as breaking continuity poor information. In this paper, we propose a new model called enhanced fusion network (ESFNet), which contains two parts: optimized multi-scale fused attention module (FsSE) 3D convolutional neural (3D CNN) based on...

As the research on deep learning methods gradually progresses, more and classification models are applied in of hyperspectral image. High-dimensional low-resolution characteristics image (HSI), however, make it difficult for conventional to process its data effectively. In this paper, a novel HSI model, namely Spatial Spectral Pyramid Network (SSPN), is designed by combining 3D Convolutional Neural (3D CNN) with feature pyramid structure. SSPN taking advantage convolution coupled multi-scale...

KNN has gained popularity in machine learning due to its simplicity and good performance. However, kNN faces two problems with classification tasks. The first is that an appropriate distance measurement required compute distances between test sample training samples. other the highly computational complexity requirement of searching nearest neighbors whole data. In order mitigate these problems, we propose a novel method named KCNN enhance performance kNN. uses convolutional neural networks...