39 results about "Terrain classification" patented technology

The invention discloses a terrain classification method based on a random forest algorithm, a server and a storage medium, and belongs to the technical field of terrain classification, and the method comprises the steps: obtaining DEM data and remote sensing image data of a research area, and extracting terrain features, spectral features, index features and texture features; designing a feature combination scheme, and generating an optimal feature combination data set as input data of terrain classification; calculating an optimal parameter combination of a random forest algorithm, and generating an optimal random forest classifier; establishing a terrain classification system suitable for the research area, and constructing a terrain classification training sample set and a verification sample set; performing terrain classification by using a random forest classifier and the optimal feature combination data set to obtain a terrain classification result of the research area; and calculating and evaluating the precision of a terrain classification result by utilizing a verification sample set and a Kappa coefficient evaluation method. By adopting the method, the terrain classification refinement degree and the terrain classification calculation efficiency and classification efficiency can be improved.

The invention discloses a spectrum-space depth fusion hyperspectral image classification method oriented to a small sample condition, and the method comprises the steps: firstly carrying out the multi-scale hole convolution processing of input hyperspectral data through a range attention convolution SAC module, and extracting the multi-scale context features; then, a spatial normalization attention SNA mechanism is utilized to carry out adaptive weighting adjustment of spatial dimensions on the feature map, and spatial feature representation of the key area is enhanced; the method comprises the following steps: constructing a lightweight hybrid expert model LMOE, carrying out parallel processing and gating weighting through a multi-path expert network, carrying out efficient refining and mapping on features, finally fusing processed spectral features and spatial features, and carrying out pixel-level prediction through a classifier to obtain a terrain classification result map of a hyperspectral image. The method solves the problems that in the prior art, overfitting is prone to occurring under the small sample condition, the spectrum-space collaborative modeling capacity is insufficient, the long-range dependence obtaining efficiency is low, and the recognition precision is reduced under the class imbalance scene.

The application provides a terrain recognition method, device and equipment for a legged robot and a medium. The method comprises the following steps: acquiring pose data collected by a body sensor of a target legged robot, and calculating state parameters of the target legged robot; inputting the pose data and the state parameters into a foot-ground contact force model to determine three-dimensional foot-ground contact force; generating a contact mask sequence representing a foot-end ground contact state of the target legged robot based on the three-dimensional foot-ground contact force, and segmenting the pose data to obtain a segmented pose data set; inputting the pose data set into a pre-trained terrain classification model to determine a confidence vector of the current ground under multiple terrain attributes, and determining a terrain category of the current ground according to multiple confidence vectors. Through the method and device, the autonomous adaptation capability and environmental perception reliability of the legged robot in complex terrain are significantly improved.

The invention discloses a power transmission line microtopography recognition method, system and device based on a multi-modal attention mechanism and a medium, and the method comprises the steps: obtaining power transmission line space corridor data and multi-source geographic information data of a target region, carrying out the preprocessing, and generating a line feature map and a multi-channel ground feature map; constructing a double-branch deep learning network model, inputting the line feature map into a topographic feature branch, inputting the multi-channel ground feature map into a line feature branch, and respectively extracting deep feature information; through a cross-branch space attention module, weighting the features extracted by the topographic feature branches by using a space attention weight generated by extracting the features by the line feature branches to obtain fusion features; and inputting the fused features into a decoder for up-sampling and feature reconstruction, and outputting a classification recognition result for the microtopography along the power transmission line corridor. According to the invention, full-process intelligent processing is realized, the identification efficiency is obviously improved, and the method is suitable for wide-range line census and risk assessment.

This application proposes a method, device, and electronic device for acquiring fire rescue tactics. The method includes: acquiring the fire line classification features and terrain classification features of any first grid cell in a set of grid cells corresponding to the fire line; acquiring the feature vector corresponding to any first grid cell based on the first feature vector corresponding to the fire line classification features, the second feature vector corresponding to the terrain classification features, and the third feature vector corresponding to the environmental features of any first grid cell; and analyzing the feature vector corresponding to any first grid cell using a target graph convolutional neural network model and a target temporal Transformer model to acquire the fire rescue tactics corresponding to any first grid cell. This solves the technical problem in the prior art that fire relief methods cannot adapt to changes, resulting in a mismatch between the fire relief methods and the current fire situation, and poor fire extinguishing effect.

The application discloses a passability prediction method for unstructured terrain environment, and is applied to the technical field of vehicle traffic capacity prediction. The application constructs a three-dimensional grid map based on a multi-source sensor, combines semantic terrain classification and elevation structure feature extraction, and forms a geometric and semantic joint representation of the environment. By comparing the control instruction with the motion response of the vehicle in the actual terrain, a traffic capacity coefficient is constructed as a supervision signal, and a conditional variational autoencoder is trained to model the traffic capacity coefficient distribution and its uncertainty under different terrain conditions. Further, intermediate layer features are extracted from the semantic recognition network to establish a category statistical model for identifying terrain categories that do not appear in the training set. Finally, the traffic capacity confidence and terrain reliability are fused to output a grid-level traffic probability, realizing the passability prediction in the unstructured environment.

The application provides a topographic classification method and system based on typhoon wind speed prediction, and the method comprises the following steps: obtaining topographic basic data of a classification area; constructing a topographic multi-classification set based on regional meteorological model wind speed and the topographic basic data; obtaining typhoon wind speed prediction data and performing risk assessment on the topographic multi-classification set to obtain a risk index grading classification result corresponding to the topography; and obtaining a microtopography distribution map of the classification area based on the risk index grading classification result. The topographic classification method based on typhoon wind speed prediction can realize the risk assessment of a future line under a current wind speed environment by adding the typhoon wind speed prediction data into the topographic classification process for dynamic topographic classification, and the line operation and scheduling efficiency is improved.

The application relates to a photovoltaic pile driver terrain identification method and system, a terminal and a storage medium, and relates to the technical field of intelligent equipment. The application comprises the following steps: analyzing camera pixel data, laser radar data, a camera internal parameter matrix, a camera radar external parameter matrix and a radar satellite external parameter matrix to determine pixel camera data, radar world data and camera world data; fusing the radar world data and the camera world data to determine simultaneous space fusion data and coordinate point gray values; analyzing the coordinate point gray values and the simultaneous space fusion data to determine spatial terrain features and pile driving terrain classification; and analyzing the radar world data, the camera world data, the spatial terrain features and the pile driving terrain classification to determine real-time terrain modeling. The application has the effects of improving the photovoltaic pile driving terrain identification accuracy and terrain updating efficiency.

The land classification method based on deep and shallow feature and multi-attention mechanism fusion, electronic equipment and storage medium belong to the technical field of remote sensing image processing and deep learning semantic segmentation. In order to solve the problem of accurate classification of similar categories and category boundaries in land use classification method, the present application comprises the following steps: constructing a multi-scale normalized spatial attention mechanism layer; constructing a multi-scale normalized direction position attention mechanism layer; constructing a multi-scale normalized channel attention mechanism layer; constructing a land use classification network model, including a normalized multi-attention feature extraction network and a multi-deep and shallow feature fusion dilated convolution network; collecting data in the city-rural area adaptive land surface cover dataset; performing data preprocessing to obtain land data set; inputting the land data set into the constructed land use classification network model; performing land classification training based on deep and shallow feature and multi-attention mechanism fusion; and obtaining the land classification model based on deep and shallow feature and multi-attention mechanism fusion. The present application is accurate in classification.

The invention relates to the technical field of railway line selection, and discloses a terrain-driven intelligent site selection and line trend search method for a long and large mountain railway tunnel. The search method integrates three parts of topographic data preprocessing, topographic classification and tunnel portal search and route trend search, and specifically comprises the following steps: converting three-dimensional topographic data into spatial waves, and processing the spatial waves through Fourier transform to obtain topographic relief features; a low-pass filter is designed based on the representative tunnel length for filtering preprocessing; carrying out block classification on a topographic region by utilizing the processed topographic data, and searching a feasible tunnel portal position; grouping the potential tunnel portal positions; the construction cost is used as a main consideration factor, a bidirectional A algorithm is used for connecting tunnel portals, a plurality of feasible schemes are generated, and an optimal path is obtained through comparison and selection; and generating a specific three-dimensional railway line scheme by using the optimal path. The searching method is high in searching efficiency and suitable for railway line design in complex mountainous areas.

The application discloses a remote sensing image-based paleontological fossil producing area identification method and system, and belongs to the field of paleontological fossil exploration, comprising the following steps: acquiring GF-2 multispectral images, digital elevation data and geological map data, pre-processing the GF-2 multispectral images to obtain a land classification map, performing Gram-Schmidt fusion processing on a panchromatic band and a multispectral band, performing vectorization processing on the geological map data and extracting Qiongzhusi group stratum boundaries, and obtaining a slope factor map based on the digital elevation data; superimposing the Qiongzhusi group stratum boundary vector data on the fused GF-2 multispectral images, and extracting an image candidate area of the Qiongzhusi group stratum through spatial analysis. The application realizes efficient and accurate fossil exploration through multi-source data fusion and intelligent analysis technology, and effectively solves the problem of fossil feature identification in mixed pixels by combining a dynamic spectral decoupling algorithm with a professional fossil spectral feature library and solving component abundance through constraint energy minimization.

A land use classification method based on remote sensing images, comprising the following steps: step 1: obtaining the original image of remote sensing images of urban and township land, marking the pictures after screening, and a total of four categories: cultivated land, orchard, forest and building; step 2: cutting and data enhancement are carried out on the marked pictures to make a data set; step 3: the training set in the prepared data set is put into a feature extraction network and a multi-scale fusion up-sampling structure for training to obtain a segmentation model; step 4: the remote sensing image to be classified is put into the trained segmentation model to obtain the result; the purpose of the present application is to solve the problems of low small target precision and poor edge segmentation effect when the land categories are many and the features are complex in the land classification method based on remote sensing images, and a method capable of realizing more accurate identification result of various land types in remote sensing images is proposed.

This invention discloses a spatial downscaling method for weather forecast models based on a terrain classification super-resolution model. The method first obtains nested data at each layer through dynamic downscaling of the weather forecast model, then trains a super-resolution model using data at different resolutions of the target area. Following the above process, super-resolution models are established for different terrains, and a terrain classification model is trained using terrain data. The super-resolution model and the terrain classification model are combined to obtain a fused prediction result. This fused model is the SRBTC model, which considers spatial correlation. The model outputs the similarity probability between the area to be predicted and various different terrains. The prediction results are weighted and summed to obtain the fused prediction result of the super-resolution model. During model training, a specified value scaling method is proposed to consider the differences between high- and low-resolution meteorological simulation data. When applying the model, a splitting and merging method is proposed to consider the differences between multi-scale simulation data, and the applicability of the model is specified.

The invention belongs to the technical field of remote sensing, and provides a surface feature classification method based on combination of a hyperspectral image and a laser radar. The method comprises the following steps: acquiring a hyperspectral image and laser radar data; labeling ground feature categories in the processed data, and making a label file; constructing a ground object target detection model and performing training according to the data and the label file; the model comprises a multi-scale spectrum-space sparse coding module used for extracting hyperspectral sparse features, a multi-scale geometric structure sparse coding module used for extracting laser radar geometric sparse features, and a competitive sparse selection module used for multi-scale feature dynamic fusion. And finally, utilizing the trained model to generate a ground feature classification result. According to the terrain classification method, the problem of neglecting modal sparsity difference and feature redundancy is solved through a'modal specificity sparse modeling + multi-scale competitive selection 'mechanism, and the accuracy and robustness of multi-source remote sensing data terrain classification in a complex scene are effectively improved.

The invention discloses an ancient biological fossil producing area identification method and system based on a remote sensing image, and belongs to the field of ancient biological fossil exploration, and the method comprises the steps: obtaining a GF-2 multi-spectral image, digital elevation data and geological map data, carrying out the preprocessing of the GF-2 multi-spectral image, obtaining a land classification map, carrying out the Gram-Schmi dt fusion processing of a panchromatic wave band and a multi-spectral wave band, and obtaining a land classification map; vectorizing the geological map data, extracting the stratum boundary of the qiongzhuea temple group, and obtaining a slope factor graph based on the digital elevation data; and superposing the boundary vector data of the qiongzhuea tumidinoda temple formation to the fused GF-2 multispectral image, and extracting an image candidate region of the qiongzhuea tumidinoda temple formation through spatial analysis. According to the method, efficient and accurate fossil exploration is achieved through the multi-source data fusion and intelligent analysis technology, the dynamic spectrum decoupling algorithm is combined with a professional fossil spectrum feature library, the component abundance is solved through constraint energy minimization, and the problem of fossil feature recognition in mixed pixels is effectively solved.

The invention provides a terrain recognition method, device and equipment for a foot-type robot and a medium, and the method comprises the steps: obtaining pose data collected by a body sensor of a target foot-type robot, and calculating state parameters of the target foot-type robot; inputting the pose data and the state parameters into a foot-ground contact force model, and determining a three-dimensional foot-ground contact force; based on the three-dimensional foot-ground contact force, generating a contact mask sequence representing the ground contact state of the foot end of the target foot type robot, and segmenting the pose data to obtain a segmented pose data set; and inputting the pose data set into a pre-trained terrain classification model, determining confidence vectors of the current ground under various terrain attributes, and determining the terrain category of the current ground according to the multiple confidence vectors. By means of the method and device, the self-adaptive capacity and environment sensing reliability of the foot type robot under the complex terrain are remarkably improved.

A method for terrain classification and adaptive gait switching of a turtle-like amphibious robot based on visual-haptic perception, comprising: establishing a robot model equipped with a pressure sensor and a waterproof camera; collecting haptic time series data and visual images on various terrains such as gravel and sand, and constructing a training data set; establishing a multi-modal convolutional neural network, fusing features through double-channel feature extraction and adaptive attention mechanism; collecting motion energy consumption and speed on each terrain to determine the optimal gait; combining terrain physical property classification and optimal gait to construct a gait switching finite state machine. The method forms a "perception-decision-driving" closed loop, which perceives the terrain through visual-haptic perception, decides the gait through the finite state machine, and executes the gait driving motion. Experimental verification shows that the adaptive gait can improve the mobility efficiency and speed, and enhance the adaptability of the robot in amphibious environment.

Systems and methods for real-time terrain classification via an autonomous robot are provided. An example method may obtain environmental data indicating one or more characteristics of a physical environment including terrain, and classify the terrain based upon the environmental data. Based upon a first classification of the terrain, the method may determine a terrain assessment task associated with reclassifying at least the portion of the terrain and determine whether performing the terrain assessment task will exceed a performance threshold indicating an adverse effect on performing a mission task. Based upon determining whether performing the terrain assessment task will exceed the performance threshold, the method may generate terrain assessment task configuration data for configuring the autonomous robot and transmit the terrain assessment task configuration data to the autonomous robot causing configuration of the autonomous robot associated with the terrain assessment task.

The invention belongs to the technical field of weather forecast, and discloses a rainfall downscaling method and device and a medium, and the method comprises the steps: obtaining a low-resolution rainfall forecast field of a target region, and carrying out the upsampling to obtain a basic high-resolution rainfall forecast field; obtaining topographic features of the target area, dividing topographic types for the target area according to the topographic features, and generating topographic classification masks corresponding to the topographic types; fusing the topographic features and the topographic classification mask to obtain a topographic feature map; fusing the topographic feature map and the basic high-resolution rainfall forecast field, and inputting the fused topographic feature map and basic high-resolution rainfall forecast field into a topographic perception variational auto-encoder to obtain a hidden space vector; denoising the hidden space vector by using a diffusion model under the condition of the topographic feature map; and reconstructing the denoised hidden space vector into a high-resolution rainfall forecast field by using a variational self-decoder. The topographic features are fused into the downscaling process, depth utilization of topographic information is achieved in the downscaling process, and the problem that a traditional method is insufficient in rainfall detail depicting capacity of a complex topographic area is solved.

The invention discloses a complex-terrain-oriented low-gravity simulation equipment parameter layered adjustment system and method. The system comprises a test scene layer, a sensing layer, a control layer and an execution layer, the method comprises the following steps: 1) simulating a lunar surface terrain in a ground surface base, and constructing a low-gravity simulation environment and low-gravity simulation equipment; 2) perceiving a terrain structure of the low-gravity simulation environment, and performing preliminary terrain classification on the perceived terrain structure; 3) generating initial simulation equipment parameters matched with the terrain based on the preliminarily classified terrain categories; 4) operating the low-gravity simulation equipment; 5) sensing the running state of the low-gravity simulation equipment in real time; 6) dynamically correcting simulation equipment parameters based on the running state sensed in real time; according to the low-gravity simulation equipment real-time adjustment method based on the layered architecture, through the layered architecture and the dynamic adjustment strategy, the complex terrain adaptability, the parameter adjustment real-time performance and the gravity compensation control precision of the low-gravity simulation equipment are comprehensively improved, and an efficient solution is provided for testing and training of the low-gravity complex environment.

The present disclosure provides a system for terrain classification and vehicle suspension sub-system health status indication. The system includes run-time terrain classification circuitry to determine a first terrain classification, for a vehicle travelling on a terrain segment, based on comparing a trained model to suspension sub-system sensor data and vehicle speed data; model accuracy determination circuitry to determine a second terrain classification, for the vehicle travelling on the terrain segment, based on the suspension sub-system sensor data and vehicle speed data and independent of the trained model; and sub-system health determination circuitry to determine a health indication of the suspension sub-system by comparing a difference between the first and second terrain classifications to at least one health threshold.

Disclosed is a robot perception based terrain recognition and gait control method, system, medium, and device. The method includes: acquiring pitch angle data and foot force data of a bipedal robot based on optimal foot forces; inputting the pitch angle data and the foot force data into a trained k-nearest neighbor (KNN) model to recognize a terrain on which the bipedal robot is currently walking, and obtaining a terrain recognition result; a training process of the KNN model including: driving the bipedal robot to pass through a rough terrain and a flat terrain with a fixed step frequency gait, inputting collected pitch angle sample data and foot force sample data from different terrains into the KNN model, and using the KNN model to perform terrain classification; receiving the terrain recognition result, and adjusting step frequency and gait based on the terrain recognition result according to a preset gait control strategy.

This application provides a semantic segmentation method and system for medium- and low-resolution time-series satellite imagery. The method includes determining the target time series of each pixel in a target area; and extracting temporal features based on the target time series of each pixel to obtain the temporal feature h of each pixel. T Based on the target time series of each pixel, spatial features are extracted to obtain the spatial feature x of each pixel. T For each pixel, based on the corresponding temporal feature h T and spatial features x T Perform feature fusion memorization to obtain the corresponding feature fusion sequence r. T ; Feature fusion sequence r based on each pixel T Feature stitching is performed to obtain a target feature map; based on the target feature map, the corresponding segmentation result is output. This method can improve the utilization rate of medium- and low-resolution satellite imagery in land classification.

The application provides a kind of surface deformation monitoring and classification method based on InSAR and deep learning, comprising the following steps: (1) collecting the multi-scene time series SAR image and DEM data of monitoring area, (2) registering, interfering and DEM difference to the multi-scene time series SAR image, (3) selecting PS pixel by amplitude dispersion and coherence coefficient threshold, (4) linking PS pixel to construct spatial network, (5) solving the model parameters of spatial network arc segment, (6) restoring the deformation information of PS monitoring point, (7) using random forest to classify the registered SAR image once, (8) using primary classification sample and ViT network to classify SAR image twice, (9) classifying, counting and cause analysis on deformation based on multi-classification result.The application improves the classification accuracy of SAR image classification method, and the land classification result can effectively classify and analyze InSAR deformation.

A computerised system for environment-based accident, incident, or risk detection for human subjects in outdoor environments comprises at least one user device configured to be carried or worn by a human subject. The at least one user device comprises a location determination module; a movement detection module comprising at least one of an accelerometer, gyroscope, or other motion sensor; and a communication module configured to transmit data via non- cellular-dependent protocols. The system further comprises a central server comprising at least one processor and memory. The central server is configured to receive, from the at least one user device, subject-specific data comprising: current location data indicating a current location of the human subject in an outdoor environment; and movement data relating to the human subject, including at least one of: speed, direction, acceleration, or stationary duration. The central server is configured to receive or access environment-specific data indicating environmental conditions for the outdoor environment. Said environment-specific data comprises terrain characteristic data indicating terrain characteristics, said terrain characteristic data comprising one or more of: geospatial data including topographical and geographical features; terrain classification data. Said environment-specific data further comprises meteorological data indicating localised weather conditions for the outdoor environment including at least one of: temperature, precipitation, wind, or visibility. The central server is configured to generate a digital geospatial representation of the outdoor environment, comprising: a plurality of geospatial polygons defining one or more active monitoring zones and one or more non-active monitoring zones; wherein active monitoring zones designate outdoor areas where accident, incident or risk detection monitoring is enabled; and wherein non-active monitoring zones designate areas where accident, incident or risk detection monitoring is suspended or not enabled. The central server is configured to determine, for the current location of the human subject, whether the location falls within an active monitoring zone or a non-active monitoring zone by performing geospatial polygon intersection analysis. If the location falls within an active monitoring zone, the central server is configured to perform behavioral anomaly analysis comprising: retrieving historical movement data for the human subject from a historical movement database, said historical movement data comprising prior movement data of the human subject under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data; calculating expected movement data for the human subject at the current location based on: (i) the historical movement data, (ii) the terrain characteristics indicated by the environment- specific data, and (iii) the weather conditions indicated by the environment-specific data; and comparing the movement data of the human subject against the expected movement data to generate an anomaly score. If the anomaly score exceeds a predetermined threshold, the central server is configured to generate an output alert signal to at least one recipient device.

This invention proposes a terrain classification method based on SVM hard sample mining and CNN weighted training, belonging to the field of robot environmental perception technology. The method includes the following steps: collecting multimodal monitoring data of a robot moving on various terrains; preprocessing the multimodal monitoring data to generate multimodal feature vectors at different times; and forming a training sample set from these multimodal feature vectors. This invention uses a support vector machine to perform prior difficulty assessment on the training samples, quantifying the distance from the sample to the classification decision boundary as a difficulty index, and generating dynamic training weights accordingly. Then, during the training of the convolutional neural network, these weights are used as weighting factors to construct a loss function, allowing the model optimization process to focus on learning the features of samples with higher classification difficulty, thereby effectively improving the system's accuracy in overall terrain classification and the model's robustness in real-world complex scenarios.

The invention discloses a meteorological data correction method fusing topographic features, which relates to the technical field of meteorological data correction, and comprises the following steps: carrying out topographic classification by adopting DBSCAN clustering according to a topographic feature parameter set, obtaining a topographic classification result, collecting historical meteorological data of a target area, carrying out regression analysis in combination with the topographic classification result, and obtaining a meteorological data correction result; establishing a deviation correction knowledge base, training a TCN architecture according to the historical meteorological data of the target area in combination with the topographic feature parameter set, generating a topographic-meteorological response mapping model, inputting the real-time meteorological data of the target area and the topographic feature parameter set into the topographic-meteorological response mapping model, and generating a meteorological predicted value; and calling the deviation correction knowledge base to dynamically correct the weather prediction value to obtain a corrected weather prediction value. Through the meteorological data correction method, prediction errors caused by terrain differences are effectively reduced, and the meteorological prediction value is further optimized.

This invention provides a land classification method and apparatus based on space-frequency coordination and cross-modal fusion. The method includes: acquiring cross-modal remote sensing image data to be classified; preprocessing and classifying the cross-modal remote sensing image data to obtain first modality data and second modality data; inputting the first modality data and second modality data into a trained land classification model, and predicting the land classification result based on the model. The land classification model includes: a first encoder, a second encoder, and a decoder; at each level of the first encoder and the second encoder, a semantic feature focusing module (SFFM) and a cross-modal feature fusion module (CCMF) are embedded, respectively. The method provided in this application effectively improves the accuracy of land classification.

The invention relates to the technical field of unmanned aerial vehicle application, in particular to a distribution network construction cost terrain classification calculation method and application, and is applied to cost auxiliary calculation of various terrains such as mountain land, plateau, plain, river valley, desert and gobi. The method is especially suitable for Tibetan area construction cost. Through the unmanned aerial vehicle and AI identification technology, intelligent auxiliary calculation of the distribution network project cost by the unmanned aerial vehicle is realized, and the problem of field inspection is solved. A big data standard image library is established, intelligent image identification field application is deepened, comparison with background images is carried out, a field equipment material list is automatically identified and counted, and the cost accuracy and efficiency are improved. And analyzing engineering use equipment, and performing comparative analysis in combination with different terrains to realize accurate cost of intelligent power distribution network equipment.

The invention provides a multi-algorithm fused adaptive foot motion control method for a quadruped robot. The method comprises the following steps: S1, constructing a real-time terrain elevation map through multi-sensor fusion, extracting terrain feature parameters, performing real-time classification on terrain features by using a pulse neural network, and outputting terrain types and confidence coefficients; s2, dynamically adjusting CPG network parameters based on a terrain classification result, generating an adaptive rhythm signal, and solving a foot end force optimization problem according to a fuselage dynamic state and stability constraints; s3, based on terrain elevation information and obstacle distribution, a self-adaptive foot end movement track is planned through a self-adaptive foot end mechanism; s4, solving a joint angle through inverse kinematics, controlling an actuator to complete foot movement, adjusting control parameters in real time according to plantar pressure feedback, and inhibiting foot end slippage; according to the invention, the complex terrain environment adaptability can be improved, the movement stability is enhanced, the energy consumption is reduced, and the obstacle avoidance capability is improved.