21 results about "Natural field" patented technology

The invention belongs to the technical field of remote sensing image intelligent processing and agricultural information extraction, and particularly relates to a remote sensing large model pre-training and multi-granularity boundary supervision natural field type extraction method, which comprises the following steps: firstly, pre-training a model on a large-scale space-time spectrum remote sensing data set, and combining anchor point sensing mask and geographic information coding; secondly, inputting the multi-scale features into a multi-branch structure sensing network, and outputting a semantic segmentation prediction map through high and low resolution double input and dynamic attention fusion; generating a multi-scale field boundary label through morphological operation, and outputting a boundary prediction map through multi-task supervision after domain enhancement of features by a frequency space double-domain enhancement module; and finally, aligning the two images and performing pixel-level operation to obtain a high-precision extraction result. Through the method, the global classification error during cross-region migration is greatly reduced, the method is adaptive to a low-pixel wide ridge, the boundary detection value and the closure rate of a small-scale field are improved, and the extraction precision of a large field and a small field is considered.

The invention relates to the technical field of animal behavioral experiments, and discloses an animal behavioral experiment multi-scene simulation system. According to the invention, through cooperative control of multiple environment modules and seamless switching of a dynamic scene, continuous switching of natural scenes such as illumination, sound, smell, terrain and the like is accurately simulated; on the basis of real-time biological signal detection and intelligent feedback, environment adjustment or scene switching is automatically triggered according to animal behaviors / physiological states, and closed-loop self-adaptive adjustment is formed; in combination with an intelligent decision and an AI scene generation engine, self-definition and iterative optimization of a complex script are supported; the parameter coupling controller simulates the interaction effect between environmental factors, the fault-tolerant mechanism and the effect evaluator guarantee the data reliability, the closed-loop verification loop efficiently establishes a causal model of environmental parameters and behavior response, and a high-authenticity, self-adaptive and reproducible experimental platform is provided for animal behavior research.

Micro-bioenergy system for generating electrical energy from formic acid, comprising at least one capillary tube with a semipermeable membrane for drawing in formic acid from an ant nest, a micropump for transferring the acid into a reaction cell, a reaction cell with magnesium and copper electrodes for electrochemical conversion, an energy storage device for intermediate storage of the generated energy, and a housing for integrating and protecting the components.

The invention discloses a cotton disease detection method and device used in a natural scene, and the method comprises the steps: collecting and marking cotton disease detection data in the natural scene, constructing a self-made data set based on the detection data, and evaluating the self-made data set; an information sharing header, a self-adaptive down-sampling module, a feature extraction network and a YOLO11 algorithm are combined, a feature dynamic redistribution network is established, and model evaluation is carried out on a self-made data set and a public data set; and deploying the feature dynamic redistribution network on edge equipment to realize a disease detection task. The device comprises a processor and a memory. According to the method, the average detection precision of cotton diseases is improved, the size of the model is reduced, and the method is suitable for on-site deployment of cotton disease detection edge equipment in intelligent agriculture.

The invention discloses a safety key scene generalization generation method and system oriented to a natural data set, and relates to the technical field of automatic driving, and the method comprises the steps: collecting scene data of a natural driving scene; labeling all the scene data; inputting the map information and the historical state sequence into a scene encoder for encoding; inputting the future state sequence with noise, the coding condition and the noise level into a diffusion decoder for iterative decoding to obtain a generated future state sequence; iteratively updating the basic scene diffusion model; calculating a safety key score of each scene in the natural scene data set by adopting a layering scheme; comparing the safety key score with a preset threshold value, and screening out a safety key scene; and utilizing the trained scene diffusion model to generate a future state sequence of the safety key scene. According to the method, the safety key scores of the scenes are calculated, and the sufficient number of safety key scenes are extracted and generalized, so that the time cost and the financial cost for generating the safety key scenes are effectively reduced.

The invention relates to the technical field of Chinese chestnut screening, in particular to a Chinese chestnut shell maturity classification method and system and a storage medium. The classification method comprises the steps that a Chinese chestnut image data set in a natural scene is collected, a chestnut shell position frame and maturity grades are marked, and the maturity grades are divided into an immature stage, a half-mature stage of transition from turquoise to yellowish-brown, a yellowish-brown mature stage and an over-mature stage of cracking and deterioration according to chestnut shell color change characteristics; and forcibly marking the target of which the shielding area exceeds 30% as an immature stage, and not marking the long-range target of which the maturity cannot be identified. According to the method, a deep coupling mechanism of the target characteristics and the model architecture is established, a noise transmission path is blocked from a data source, and the tolerance of the model to illumination variation and angle distortion is remarkably enhanced in cooperation with the combined effect of pre-enhancement and online random enhancement.

A geophysical surveying method and assembly applying transient pulses of electric current to an airborne time-domain electromagnetic transmitter to generate a primary controlled electromagnetic field; measuring, using an airborne receiver, a secondary controlled electromagnetic field to generate controlled field data; measuring, using the airborne receiver, a magnetic component of a natural electromagnetic field at an above-ground position to generate first natural field data; measuring, using a ground receiver at a ground station, telluric electrical currents induced by the natural electromagnetic field and / or a magnetic component of the natural electromagnetic field at a ground position to generate second natural field data; merging the first natural field data and the second natural field data into combined natural field data; extracting, from the combined natural field data, off-time natural field data recorded between the pulses; and generating geophysical survey data based on the controlled field data and the off-time natural field data.

The invention belongs to the technical field of pesticides, and particularly discloses application of leaf acetate in prevention and treatment of fusarium graminearum, leaf surface spraying of leaf acetate can significantly reduce the incidence rate of corn stalk rot caused by fusarium graminearum (reduced by 24%), significantly improve the plant height of corn and the fresh weight of the overground part and the underground part, and improve the yield of the corn. It is proved that the leaf acetate has an excellent prevention and control effect on the corn stem rot in the natural field environment, normal growth of corn can be guaranteed, and the leaf acetate has the practical agricultural production application potential.

The invention discloses a portable plant surface potential detection device and method, the device comprises a detection electrode assembly, a portable electrochemical workstation and a terminal computer, the detection electrode assembly is electrically connected with the electrochemical workstation through a shielding conductive wire, and the electrochemical workstation is in bidirectional communication connection with the terminal computer. Plant surface potential monitoring is carried out by creatively adopting an open circuit electrochemical method, traditional high-price patch clamp equipment is replaced with a low-cost portable electrochemical workstation, and all parts of the device are in portable modular design and can be quickly assembled and disassembled, so that the device is convenient to use and popularize. And the field real-time monitoring of the surface potential of plant stalks and leaves in a field natural growth state is realized. According to the device, under set detection and stimulation parameters, personal errors are effectively reduced, detection data are accurate, stability is high, potential response rules of plants under stimulation of fire, damage and the like can be captured, and the device has the advantages of being low in cost, portable, high in adaptability and wide in application prospect.

The embodiment of the invention provides a vehicle atmosphere control method and device, a vehicle and a storage medium. The method comprises the following steps: acquiring a target natural scene mode applied to a vehicle; target environment parameter information corresponding to the target natural scene mode is determined in a first mapping relation, and the first mapping relation records the corresponding relation between the multiple natural scene modes and the multiple pieces of environment parameter information; each piece of environmental parameter information is used for indicating execution parameters of a vehicle-mounted system associated with vision, smell, hearing, touch and climate in the vehicle in the corresponding natural scene mode; and controlling the vehicle to run according to the target environment parameter information. The method is used for providing dynamic changing natural environment simulation in a vehicle, so that a user obtains deep relaxation experience.

The invention discloses a rice leaf folder imago image detection and counting method and system for a complex environment, and belongs to the technical field of image processing and computer vision. Comprising the following steps: collecting rice leaf folder imago images in an actual rice field scene or an insect trap, constructing an image training set, and rejecting seriously fuzzy or seriously shielded samples through an image cleaning strategy to improve data quality and model effectiveness; an efficient attention backbone module and a cross-scale fusion network are introduced on the basis of a YOLOv8 model structure, and the detection performance of the model on small insect bodies and multi-scale targets is enhanced; a freezing-unfreezing two-stage fine tuning training method is adopted for a newly added image, and the stability and recognition robustness of the model in a complex environment are improved; experimental results show that the method has good detection precision, applicability and real-time performance in a natural field environment, and is suitable for an insect situation monitoring terminal, an agricultural image recognition platform and a pest and disease damage intelligent diagnosis scene.

The invention discloses a method for generating a graphic trademark in a natural scene, relates to the technical field of image generation, and can enable a stable diffusion model to learn spatial logic between the graphic trademark and a background object through a multi-strategy training mechanism of spatial semantic relationship modeling, trademark identity feature binding and trademark fusion processing fusion. Therefore, embedding generation of the trademark in various backgrounds is naturally completed, the authenticity and context consistency of image synthesis are remarkably improved, and the technical problems of insufficient fidelity and limited style control capability in a complex scene in the prior art are solved.

The invention provides an improved YOLOv7-based fruit tree fruit detection method in a natural scene, which belongs to the technical field of computer vision and intelligent agriculture crossing and comprises three stages of data acquisition and enhancement and model construction and training. In the data acquisition stage, a multi-modal data set containing RGB and multispectral images is constructed. In the data enhancement stage, data is expanded by adopting three-stage strategies of instance shearing and pasting, geometric enhancement and spectrum enhancement. In the model construction stage, a convolution block attention module is integrated into a YOLOv7 neck network to enhance the feature discrimination capability, and a loss function is replaced by SIoU loss to optimize bounding box regression. According to the method, fruit feature extraction is enhanced through a channel-space double-attention mechanism, the small target positioning precision is improved by using an angle perception loss function, and high-precision detection is realized in a complex natural environment.

This invention relates to the field of geophysical electromagnetic detection technology, specifically disclosing a method and system for noise suppression of electromagnetic detection data from UAV-based natural field sources. The method includes: dividing the original time-domain three-component magnetic field signal into time windows and constructing a data matrix; constructing an overcomplete dictionary matrix containing typical noise morphology characteristics of various UAV platforms; selecting a subset of data as a training set and globally optimizing key hyperparameters of the denoising process to obtain the optimal denoising parameter set; performing sparse entropy adaptive denoising processing on the data of all time windows based on the optimal denoising parameter set, and concatenating and reconstructing the denoising results of each window to obtain the final denoised three-component magnetic field data. By introducing a "sparse entropy adaptive" mechanism, it achieves precise removal of structured interference and adaptive preservation of effective signals; combined with the "gray wolf optimization" algorithm, it overcomes the difficulty of balancing the sparse threshold and iteration depth in traditional methods.

The invention relates to the technical field of coal mine drilling geological exploration, in particular to a frequency-selecting water exploration method and device and electronic equipment, and the frequency-selecting water exploration method comprises the steps: obtaining three target detection frequencies corresponding to a self area, an upper area and a lower area of a horizontal drill hole respectively; a plurality of test points are preset in the extending direction of the horizontal drill hole, each test point is detected, and in the same detection period, target natural field induction electric field signals corresponding to the target detection frequencies are obtained; respectively obtaining a target apparent resistivity set corresponding to each target detection frequency based on each target detection frequency of the plurality of test points and the corresponding target natural field induction electric field signal; and based on the target apparent resistivity set corresponding to each target detection frequency, determining whether a water-containing low-resistance geologic body exists around the horizontal drill hole or not. The underground coal mine tunneling water inrush accident risk can be effectively reduced, and scientific and reliable technical support is provided for underground coal mine horizontal drilling advanced water exploration.

The present application belongs to the technical field of geographic information engineering, and discloses a meadow three-dimensional model boundary optimization method, system, terminal and storage medium, comprising: acquiring a meadow three-dimensional model to be subjected to boundary optimization in a natural scene three-dimensional model; using a distance field-based boundary fusion method to perform gradual blending of material properties of the meadow three-dimensional model and an adjacent area model; dividing a boundary domain and an internal domain of the meadow three-dimensional model, and generating model points corresponding to the boundary domain and the internal domain based on a region density control algorithm and a rejection sampling algorithm; based on the generated model points and the gradually blended material properties, optimizing the meadow three-dimensional model, and outputting an optimized meadow three-dimensional model; the present application enhances the visual consistency of the meadow three-dimensional model at the boundary, and improves the realism of the meadow three-dimensional model.

This application relates to the fields of computer graphics and artificial intelligence, and in particular to a single-image SVBRDF reconstruction method based on the fusion of lighting cues in natural scenes. The method includes: extracting ambient light distribution in a natural scene using a mirror sphere and performing geometric occlusion modeling; constructing planar light source cues and normal light source cues based on predicted normals; adaptively fusing material appearance features with bi-branch lighting cues using a multi-head cross-attention mechanism; and predicting diffuse albedo, normals, roughness, and specular reflection intensity maps using a parametric regression network. This application effectively decouples complex lighting and material properties, eliminates ambiguity in parameter estimation, and achieves high-precision, physically consistent material reconstruction without requiring a controlled darkroom environment, thus improving the robustness and practical value of single-image material digitization in natural scenes.

The invention belongs to the technical field of geographic information engineering, and discloses a grassland three-dimensional model boundary optimization method and system, a terminal and a storage medium, and the method comprises the steps: obtaining a grassland three-dimensional model to be subjected to boundary optimization in a natural scene three-dimensional model; using a boundary fusion method based on a distance field to carry out gradient mixing of material attributes on the grassland three-dimensional model and the adjacent region model; dividing a boundary domain and an internal domain of the grassland three-dimensional model, and generating model point locations corresponding to the boundary domain and the internal domain based on a region density control algorithm and a sampling rejection algorithm; optimizing the grassland three-dimensional model on the basis of the generated model point location and the material attribute after gradient mixing, and outputting the optimized grassland three-dimensional model; according to the method, the visual consistency of the grassland three-dimensional model at the boundary is enhanced, and the sense of reality of the grassland three-dimensional model is improved.

The invention discloses a disturbance stripping type multi-mode cattle individual identity recognition method used in a natural scene. An existing processing mode cannot adapt to the modal missing problem, quality differences of different modals are not considered, the problem of efficiency and generalization ability imbalance exists, correspondingly, the problem of calculation redundancy caused by operation of all calculation branches cannot be solved, effective feature channels cannot be dynamically selected according to input data feature distribution, and recognition performance fluctuation is large. According to the invention, after an original image is processed, three types of pure modal input data are respectively formed through three-modal feature extraction, and the three types of pure modal input data are respectively processed through a hybrid expert module and a quality scoring module to obtain single-modal features and quality scores. And combining the single-mode features with the mass fraction, and performing dynamic sparse gating and feature fusion processing in sequence to complete the process of cattle individual identity matching.

The present application belongs to the field of agricultural information technology, and particularly relates to a grape leaf disease spot segmentation method based on a cross-resolution Transformer model, which is used to solve the problem of grape leaf disease spot segmentation in a complex background in a natural field environment. Compared with the previous Transformer model based on the consideration of obtaining high-resolution features or semantic information alone, the cross-resolution Transformer model simultaneously considers the preservation of high-resolution features and strong semantic information. The cross-resolution Transformer model is an "encoder-decoder" structure, the encoder is composed of 4 stages, and i parallelly connected multi-resolution Transformer blocks are distributed in the i-th stage, arranged in a pyramid shape; the decoder is a hamburger decoder. Each Transformer block includes a large core attention mechanism, a multi-path feedforward network and a cross-resolution fusion strategy. Finally, the feature information of different scale resolutions is fused through the hamburger decoder.

The present application belongs to the field of agricultural information technology, and particularly relates to a U-shaped perception lightweight Transformer method for segmenting grape leaf small disease spots. A lightweight convolutional neural network MobileNetV2 is adopted to extract multi-scale feature information through U-shaped pyramid down-sampling; a context perception enhancement module is used to extract low-frequency global features and high-frequency local feature maps; a token aggregation strategy is introduced to reduce the loss of detailed information caused by direct aggregation of low-frequency and high-frequency features; and the aggregated tokens are directly transmitted to a lightweight segmentation head to implement a segmentation task. The present application can achieve balanced efficiency and speed to solve the problem of grape leaf small disease spot segmentation in a natural field complex background.