111 results about "Learning Recognition" patented technology

The invention relates to the technical field of data analysis, provides an urban meteorological disaster data identification method and system based on deep reinforcement learning, and is used for effectively improving the model performance so as to enhance the accuracy and real-time performance of meteorological disaster monitoring. The method comprises the steps of obtaining a meteorological monitoring data set of a target city area, executing meteorological data preprocessing operation on the meteorological monitoring data set to obtain a preprocessed meteorological spatial-temporal feature set, calling a trained deep reinforcement learning recognition model, and performing dynamic disaster mode matching processing on the meteorological spatial-temporal feature set to obtain a dynamic disaster mode recognition model. And generating a meteorological disaster recognition result set of the target city region, generating a disaster coping strategy set according to the meteorological disaster recognition result set, and performing dynamic strategy optimization processing on the deep reinforcement learning recognition model based on the disaster coping strategy set to obtain an optimized deep reinforcement learning recognition model. And deploying the optimized deep reinforcement learning recognition model to a meteorological disaster monitoring system.

The invention discloses an immersed tunnel leakage voiceprint recognition system based on deep learning. The system comprises an acoustic signal acquisition module, a signal preprocessing module, a voiceprint feature extraction module, a deep learning recognition module and a leakage positioning and early warning module. According to the invention, the leakage water flow sound and the structural damage elastic wave signal of the immersed tunnel are comprehensively considered, the multi-dimensional voiceprint feature extraction and deep learning model are fused, and the adaptive signal noise reduction and sound source positioning technology is combined, so that the high-precision monitoring and real-time early warning of the leakage of the immersed tunnel are realized; the accuracy and real-time performance of immersed tunnel leakage nondestructive monitoring are improved, and powerful support is provided for follow-up risk prediction and early repair.

The invention discloses a real model consistency review system and method based on a large language model, according to the scheme, through integration of multi-source heterogeneous data, standardized preprocessing of real scene cloud or images is completed, and through combination of deep learning recognition and multi-scale feature analysis, accurate correspondence and intelligent comparison of a real model and a BIM component are achieved; performing semantic reasoning and logic verification by using a large language model, automatically detecting differences such as dimensional deviation, positioning deviation, missing and redundant components and the like, and generating structured difference data; furthermore, difference information is visually presented through graphical annotation and natural language description, and batch screening and multi-scale interaction are supported. Based on a difference analysis result, a standardized review report is automatically compiled, and statistical analysis, change trend tracking and multi-format export functions are provided. According to the scheme, the consistency review efficiency and reliability in a large-scale data environment can be guaranteed, and the quality management level of building construction and operation and maintenance stages can be effectively improved.

The invention discloses an intelligent parking space guiding system for a hospital parking lot, relates to the technical field of intelligent traffic, and is used for solving the problems that the hospital parking lot is seriously jammed in peak hours and special vehicles lack exclusive passing authority. The system comprises a parking space monitoring module, a vehicle identification module, an intelligent guiding module and a user interaction module. The parking space state is monitored in real time through the multi-mode sensor array; recognizing entering and exiting vehicles through a deep learning recognition algorithm based on cameras deployed at entrances and exits, and obtaining corresponding vehicle information; based on parking space state information and vehicle identification information, a route optimal value is dynamically calculated, and a whole-course navigation service from an entrance to a target parking space is provided for a driver, so that accurate detection of the parking space state, special vehicle priority distribution, high-low peak dynamic route planning and whole-process intelligent guidance are realized, the hospital parking efficiency is remarkably improved, and the hospital parking experience is improved. And the congestion condition in the peak period is relieved.

The invention discloses a radar deception jamming identification method based on adversarial game optimization. The method comprises the following steps: S1, acquiring radar echo signal data; s2, obtaining preprocessed radar echo signal data; s3, constructing an adaptive risk game model between the radar and deception jamming based on the preprocessed radar echo signal data; s4, training a deep learning recognition network by using the preprocessed radar echo signal data; s5, introducing a reinforcement learning module, and integrating the deep learning recognition network and the adaptive risk game model; and S6, the optimized radar identification strategy is applied to processing of real-time radar echo signal data, deception jamming signals are identified and classified in real time, and corresponding identification results are output. According to the method, the radar system can effectively avoid a high-risk strategy, so that the uncertainty in combat decisions is reduced.

The embodiment of the invention provides a road disease inspection method and device based on AI recognition, equipment and a storage medium, and is used for road maintenance. The method comprises the steps of performing multi-source data synchronous acquisition on a road surface to obtain an original image-positioning data set, performing dynamic interference suppression and image enhancement on the original image-positioning data set to obtain a stable enhanced image sequence, and performing disease target detection and classification in combination with a deep learning recognition model to obtain a disease target set, and performing multi-target space-time correlation and satellite positioning data fusion on the disease target set to obtain a stable target trajectory set, performing geometric coordinate conversion on the stable target trajectory set to obtain a road disease data set, and performing clustering analysis through a spatial clustering analysis algorithm to generate a road disease maintenance strategy report. Through the technical means of multi-source cooperation, deep learning, space-time fusion and the like, high-precision and intelligent detection of road diseases is realized, the road maintenance efficiency is improved, and the manpower and material resource cost is reduced.

The invention discloses a manipulator grabbing control method and equipment based on position and posture recognition. The manipulator grabbing control method comprises the following steps that S1, initial image information of a target object is acquired through image acquisition equipment; s2, performing preprocessing and feature extraction on the initial image information, and obtaining three-dimensional position coordinates and attitude parameters of the target object through a preset position and attitude recognition algorithm; and S3, according to the three-dimensional position coordinates and the posture parameters, a grabbing path of the manipulator is planned by combining a kinematic model of the manipulator. By introducing technical means such as a deep learning recognition algorithm, forward / inverse kinematics model collaborative planning, real-time posture dynamic adjustment and force sensing feedback control, the problems that a traditional mechanical arm is low in grabbing precision, poor in adaptability and insufficient in operation stability are solved, intelligent and high-precision grabbing control in a complex scene is achieved, and the grabbing precision of the mechanical arm is improved. And the requirements of the modern industry on high efficiency, reliability and flexibility of automatic equipment are met.

The invention discloses a high-precision liquid phase chip sample positioning and identification system, which belongs to the field of high-precision liquid phase chip sample positioning and identification, and comprises an active liquid phase sample positioning control module used for collecting a liquid flow state signal and adjusting the speed and position of an input sample in real time through an adjustable microfluid structure, fixed-point residence and directional conveying of samples in a chip channel are realized; the image recognition and feature position fine adjustment module is used for carrying out multi-modal image collection on an area where the sample is located, extracting spatial feature information of the target sample based on a deep learning recognition model, and carrying out fine adjustment correction on the position of the sample according to a feedback signal; by constructing an active liquid phase sample positioning and recognition system, high-precision directional control and spatial residence of a liquid sample in a chip channel are realized, the problems of sample drift and unstable recognition in a traditional passive conveying mode are effectively avoided, and the timeliness and accuracy of sample recognition are improved.

The invention discloses a teaching object interaction method, system and device. The teaching object interaction method comprises the steps of obtaining a plurality of interaction behavior characteristics of a real-time user at a current knowledge point; constructing a plurality of interaction behavior characteristics of the real-time user at the current knowledge point into a real-time interaction vector; inputting the real-time interaction vector into a learning recognition model, performing forward propagation after pre-training in the learning recognition model, and generating a mastering state label of a real-time user at the current knowledge point; according to the mastering state label of the real-time user at the current knowledge point, constructing a teaching interaction strategy of a plurality of to-be-learned knowledge points in a future time axis; wherein the mastering state labels of the plurality of knowledge points to be learned are inherited in a training sample of the learning recognition model; according to the method, the accuracy of learning path deduction and the individuation of learning strategies are improved, and the user learning experience of a teaching system is remarkably improved by combining real-time mastering state prediction and deduction based on a historical behavior mode.

The invention discloses a road damage identification and positioning method fused with multi-source information, and relates to the technical field of road and bridge facility health monitoring. The method comprises the following steps: S1, collecting multi-source image data of a road bridge through an unmanned aerial vehicle, a wall-climbing robot and / or an industrial camera; and S2, preprocessing the multi-source image data, respectively inputting the multi-source image data into a pre-trained deep learning recognition model, and outputting the type and initial position information of a disease. According to the method, the global view angle of the unmanned aerial vehicle and the local high-definition data of the wall-climbing robot / industrial camera are deeply fused, and the deep learning-based semantic segmentation model is utilized to perform pixel-level recognition on tiny diseases such as cracks, so that the recognition accuracy and reliability are greatly improved, and meanwhile, the recognition efficiency is improved. Image coordinates are accurately mapped to the three-dimensional live-action model through a coordinate transformation formula, centimeter-level space positioning of the disease is achieved, and detection accuracy and operation safety are remarkably improved.

The invention relates to the technical field of building disease detection, in particular to an urban building disease detection method and device, electronic equipment and a storage medium. Multi-modal image data formed by original visible light and thermal infrared image data is obtained, and an original thermal infrared image is subjected to geometric correction; calculating a mapping relation with an original visible light image so as to complete pixel-level registration, obtaining target multi-modal image data, inputting the target multi-modal image data into a hierarchical deep learning recognition model, recognizing building disease information, then performing three-dimensional space mapping, generating a building three-dimensional mesh model containing disease three-dimensional space setting coordinates, and finally performing three-dimensional mesh modeling. And then calculating a relationship between a model surface grid vertex and a disease point cloud density, generating a disease distribution thermodynamic diagram, analyzing disease aggregation characteristics in multiple dimensions according to the thermodynamic diagram, and quantitatively analyzing spatial correlation between the disease and a building construction node in combination with building component information. According to the invention, the urban building disease detection efficiency and precision are improved.

The invention relates to the technical field of geological informatization, in particular to an artificial intelligence modeling analysis method for a hydrate pilot production data set, which comprises the following steps of: acquiring logging data, lithology data, stratum physical property parameters and natural gas hydrate production dynamic data; screening, cleaning, complementing, de-noising and standardizing are carried out in sequence to obtain an artificial intelligence modeling data set; and establishing a stratum lithology machine learning recognition model, a stratum physical property machine learning recognition model and a natural gas hydrate artificial intelligence historical fitting model through a support vector machine SVM, a random forest RF and a neural network DNN. According to the method, a serial modeling architecture of lithology identification, physical property prediction and production history fitting is created, and the prediction output of the upstream model is used as the optimization input of the downstream model, so that the downstream production prediction model can learn physical property parameters which are recalculated based on machine learning and have higher precision; and the accuracy of final production prediction is improved from the data source.

The invention discloses a file anti-desensitization self-learning recognition system and method based on information entropy, belongs to the technical field of intersection of natural language processing and content security recognition, and is applied to document screening and risk recognition in a multi-task scene. The implementation method comprises the following steps of: 1, performing character recognition and noise reduction processing on an original file to form a data set; 2, training labeled sample data through small samples, respectively adopting probability distribution of a data sliding window and information entropy to carry out maximum and minimum normalization screening, and further utilizing a fitted linear regression model to form an anti-desensitization word list; 3, screening the anti-desensitization degrees of the sentence segments of the data set by adopting a dictionary tree Trie structure to form an anti-desensitization sentence segment table; 4, marking the chapter-level anti-desensitization degree data set text fragments by using the large model; 5, generating an anti-desensitization report according to the anti-desensitization word and the anti-desensitization degree of the marked anti-desensitization file; compared with the prior art, the anti-desensitization file screening method and device have the advantage that the anti-desensitization file screening accuracy is improved.

The invention discloses a multi-mode induction display screen response method fusing voice interaction, and relates to the technical field of man-machine interaction, and the method comprises the following steps: in a user interaction process, calling a built-in sensor through an interface component real-time state monitoring mechanism, continuously monitoring a rendering process of a target interface component at a millisecond-level time granularity, and displaying the rendering process of the target interface component; collecting multi-dimensional availability state parameter information of the interface component; the method comprises the following steps: preprocessing collected interface component multi-dimensional availability state parameter information, and extracting key indexes reflecting potential asynchronous triggering risks from preprocessed data through a feature engineering method; by sensing the state of the interface component in real time, extracting the asynchronous risk index and combining machine learning recognition and TCN prediction modeling, dynamic regulation and control of the induction triggering window and the animation rhythm are achieved, the problem of mispointing or dislocation of the induction component is avoided, the accuracy and safety of interaction response are improved, and the user experience is improved. The method is suitable for high-precision scenes such as vehicle-mounted, medical and industrial control.

The invention discloses a ship cable identification method based on an improved VGG16 network and bidirectional feature fusion. The method comprises an image acquisition module, a preprocessing module, a deep learning recognition module and an output module, and comprises the following steps: firstly, carrying out gray enhancement and anti-interference filtering on a construction drawing through the image preprocessing module, and constructing a ship cable marking database; then, an improved VGG16 network deep convolutional layer is adopted to extract cable multi-scale features, a full connection layer module of an original network is deleted, a lightweight channel self-attention module is added, and BiFPN is adopted to carry out bidirectional feature fusion on shallow texture features and deep semantic features of the improved VGG16 network, so that the features of the ship cable are captured more effectively; and finally, outputting structured data of cable numbers, specifications and path coordinates. The identification precision mAP, the processing speed and the generalization ability of the method are superior to those of a traditional method, the method can be effectively applied to the field of ship cable identification and the like, and the industrial detection efficiency and the automation level are greatly improved.

The invention discloses a myopia image deep learning recognition model training method, particularly relates to the technical field of medical image processing and deep learning, and is used for solving the problem that an existing deep learning model lacks anatomical structure priori knowledge guidance in myopia eye bottom image analysis. The method comprises the following steps: acquiring a myopia eye bottom image and anatomical structure priori knowledge data, extracting a multi-scale feature map by using a deep learning model, analyzing the geometric morphology of a key anatomical component based on standard spatial relationship information, and generating a spatial constraint loss item; according to the method, key anatomical path topology coherence is evaluated based on topology connection information, topology constraint loss items are generated, a loss item fusion strategy is dynamically adjusted according to a training stage, finally, a model is iteratively trained to convergence through a gradient back propagation algorithm, and organic combination of medical priori knowledge and a deep learning model is realized. And the clinical rationality and reliability of model output are improved.

The invention provides a signal optimization preprocessing-based adolescent schizophrenia deep learning recognition system and method. The system comprises a data preprocessing module, a data fusion module and a schizophrenia recognition module based on a deep learning model which are arranged in sequence, the data preprocessing module is used for carrying out filtering processing on the screened electroencephalogram signals, retaining frequency bands which are most obviously represented by schizophrenia, extracting power characteristics of each frequency band through a time-frequency analysis method, and carrying out compression and normalization processing; the data fusion module is used for disrupting and fusing a plurality of electroencephalogram signal data sets; and the schizophrenia identification module based on the deep learning model is used for realizing accurate identification of the schizophrenia patient in combination with the deep learning model. By optimizing the data processing flow and the model training method, the schizophrenia identification accuracy and stability are improved.

The invention relates to the technical field of coal mine data processing, and discloses a coal mine multi-source data fusion visualization system based on deep learning. The system comprises seven modules including a multi-source data acquisition module, a preprocessing module, a three-dimensional geologic model construction module, a fusion strategy generation module, a deep learning recognition module and a data optimization and evaluation module. The multi-source data acquisition module acquires geological sonar signals, environment monitoring signals and equipment operation image data of a coal mine target area; the preprocessing module performs noise reduction on the data and extracts features; the three-dimensional geologic model building module builds a regional three-dimensional geologic model according to the preprocessed data; the fusion strategy generation module combines the model and the preprocessed data to generate a fusion strategy; the deep learning recognition module recognizes the equipment image according to a strategy to obtain equipment state data; the data optimization module optimizes the data; and the evaluation module evaluates and analyzes the optimized data and generates a result, thereby providing support for coal mine production management.

The invention relates to the field of tube push bench operation, and discloses a device for controlling tube push bench operation through visual tracking, which comprises a visual perception module comprising a plurality of groups of industrial cameras arranged in a triangular array at intervals of 2-3m and equipped with 850-1200nm infrared light supplement and a self-adaptive exposure algorithm; the dynamic calibration and coordinate conversion module is used for carrying out SIFT feature matching based on a workshop fixed bracket and establishing a conversion matrix from an image coordinate system to a world coordinate system; and the deep learning identification module adopts a YOLOv5 and ResNet-50 double-model collaborative architecture and has static and dynamic shielding processing algorithms. According to the invention, multiple groups of industrial cameras are arranged in a triangular array, and an image splicing technology is combined, so that a single-visual-angle blind area is eliminated, and a detection area is fully covered; continuous and stable tracking of a moving target is ensured, tracking interruption and position deviation caused by view angle limitation or time asynchronization are avoided, infrared light supplement and a self-adaptive exposure algorithm are matched, ambient light interference is inhibited, strong light is coped, a clear image can still be obtained under complex illumination, and the positioning precision is improved.

The invention discloses a stock ground mining efficiency analysis method based on multi-mode excavator intelligent monitoring. The method comprises the steps that multiple single-mode original features corresponding to preprocessed excavator monitoring data are extracted; calculating correlation among modals of the plurality of single-modal original features to generate a preliminary fusion feature, and fusing a self-attention mechanism and a multi-head self-attention mechanism to perform multi-level feature fusion to obtain a multi-modal fusion feature; training a machine learning model by using the multi-modal fusion features to obtain an optimal machine learning recognition model so as to output an excavator activity state recognition result; and calculating the action time, the average cycle time and the productivity of the excavator in different activity states, and predicting the future production efficiency according to the calculated productivity. According to the invention, accurate identification and classification of the activity state of the excavator can be realized, the identification accuracy of the activity state of the excavator is improved, and powerful support is provided for real-time prediction and optimization of the mining efficiency of a stock yard.

An area on video is monitored that overlaps a radio range of an access point on the data communication network. Failure to identify a person currently within the Wi-Fi area and shown on video is detected. Parallel Wi-Fi parameters for a device associated with the person in the video, including a group identification assigned to the person is received. The parallel Wi-Fi parameters are used to identify the person. A machine learning module, or other image recognition system, can be updated with images to build a machine learning recognition model from a history of images associating the identified person. Finally, a surveillance security policy of the identified group is associated with the identified person.

The invention relates to the technical field of fire detection and early warning, in particular to an intelligent ultraviolet flame detection and early warning system and method suitable for complex industrial scenes. Comprising distributed multispectral ultraviolet detection nodes and a central processing early warning unit. The nodes carry out signal acquisition and preliminary identification through a multispectral sensor, local processing and high-precision time synchronization; and the central unit fuses multi-node data, performs three-dimensional flame positioning by using a TDOA algorithm, performs machine learning recognition and false alarm elimination, and performs trajectory tracking by using a Kalman filter so as to realize graded early warning linkage. According to the intelligent ultraviolet flame detection and early warning system and method suitable for the complex industrial scene, the problems that in the prior art, the false alarm rate is high, flame positioning is not accurate, trajectory tracking is missing, cooperative monitoring is limited and the like are solved or at least relieved, false alarms are effectively reduced, and the positioning precision, the tracking capacity and the early warning reliability are improved.

The invention relates to the technical field of streaming media detection, in particular to a complex element new communication streaming media detection method based on a multi-modal deep learning recognition technology, which comprises the following steps: S1, multi-modal time-space synchronization preprocessing: mapping video key frames, audio clips and bullet screen texts to a unified time axis through a combined time-space calibration technology, establishing spatial semantic association; s2, hierarchical multi-modal feature distillation is carried out, and discriminative multi-granularity features including local details, global semantics and cross-modal association modes are extracted from all modals; and S3, establishing a dynamic graph modal interaction network, constructing a learnable multi-modal relation graph, and dynamically modeling cross-modal semantic interaction. According to the complex element new communication streaming media detection method based on the multi-modal deep learning recognition technology, the problem that cross-modal complex semantic collaboration cannot be captured through single-modal analysis or shallow fusion, so that the detection missed judgment rate is high is solved.

The application discloses a CSI-based position-independent human activity continuous learning recognition method, which comprises the following steps: 1, collecting CSI action sample data; 2, pre-processing the CSI action sample data; 3, constructing positive samples by randomly scaling the pre-processed samples in the time dimension; 4, constructing a multivariate time graph neural network and extracting CSI action sample features; 5, calculating the similarity between the sample feature values and the positive samples and the feature values of the remaining samples, obtaining a comparison loss, and optimizing the feature extraction network; 6, freezing the feature extraction network, sending the features obtained from the input samples into a classifier for training to obtain a classification model. When the application continuously learns new action categories, the user does not need to retrain the feature extraction network, and the new and old action recognition in any position in the room can be realized by providing limited position new category samples to train the classifier, and the practicability is relatively high.

The invention relates to the technical field of immunohistochemical staining, and discloses an automatic scoring method and system for immunohistochemical staining results. The method comprises the following steps: performing spectral signal intelligent deconvolution processing on a multi-immunohistochemical staining image to obtain target image data; performing multi-scale context modeling on the target image data to obtain a sub-region segmentation mask; performing depth map feature extraction on the target image data based on the subregion segmentation mask to obtain a multi-dimensional image feature vector; inputting the multi-dimensional image feature vector into a mixed deep learning recognition model to carry out marker intelligent recognition to obtain a multi-marker expression state recognition result; collaborative scoring is carried out based on the multiple marker expression state recognition result, a personalized unified scoring value is obtained, cell heterogeneity analysis is carried out on the personalized unified scoring value, and an intelligent scoring report is generated. According to the invention, the problem of spectrum crosstalk in multiple staining is effectively solved, and high-precision intelligent identification of multiple immunohistochemical markers is realized.

The present invention relates to the technical field of intelligent voice customer service, and a method and system for accurately identifying the intention of a voice customer service based on deep learning, including: determining whether there is only one intersection tree node in the subsequence of intersection tree nodes. If not, identifying multiple groups of associated customer voice word sets containing the subsequence of intersection tree nodes, and constructing multiple groups of branched tree node sequences on the last intersection tree node. If so, constructing multiple groups of branched tree node sequences on the last intersection tree node according to multiple groups of iterative customer voice word sets to obtain a customer voice word tree, collecting the customer voice word trees corresponding to each voice customer service intention to obtain a customer voice word tree forest, and performing deep learning recognition of the voice customer service intention according to the set of overlapping tree nodes to obtain the current customer service intention. The present invention can improve the recognition accuracy and recognition speed of the current intelligent customer service in the aspect of voice customer service intention recognition.

The invention discloses a power grid fault data processing method and device and a medium, and the method comprises the steps: carrying out the multi-dimensional detection data collection of a target power grid, obtaining a power grid image and sensing parameters, and collecting the environment parameters in a target power grid environment; acquiring historical detection data of a target power grid, extracting a fault sample proportion coefficient, performing combination division on the historical detection data, and performing training of an integrated image fault identification branch and an integrated sensing fault identification branch; performing fault rate influence analysis and detection data influence analysis according to the environmental parameters to obtain a fault rate influence coefficient, an image influence coefficient and a sensing influence coefficient; according to the fault rate, the image and the sensing influence coefficient, the number of image recognition branches and the number of sensing recognition branches are calculated and obtained, fault recognition branch calling and fault recognition are carried out, a power grid fault recognition result is obtained, and the technical problem that the machine learning recognition fault perception rate and accuracy are low due to the fact that power grid line inspection fault data samples are few is solved.

The invention relates to the technical field of fusion gene detection, and discloses a high-throughput detection platform of a thyroid cancer specific fusion gene CCDC6. According to the detection platform disclosed by the invention, high-throughput and high-sensitivity detection on the thyroid cancer CCDC6 fusion gene is realized through a technical scheme of combining sample enrichment and deep learning recognition. Through combination of the sample enrichment module and the deep learning recognition module, the sample enrichment module promotes high-sensitivity recognition of the deep learning recognition module, an end-to-end integrated solution is provided, a large number of samples can be processed in one-time operation, the result can be automatically interpreted, and the detection efficiency and accuracy of the CCDC6 fusion gene are remarkably improved.

The invention is suitable for the technical field of computers, and provides a large language model training data generation method based on flow playback and implicit feedback, and the method comprises the steps: obtaining interaction session data, including user questions, model answers and user subsequent behaviors, of a user and a large language model in online service; based on the interactive session data, identifying a user preference signal through a hierarchical implicit feedback judgment algorithm; the hierarchical implicit feedback judgment algorithm preferentially processes high-confidence implicit signals, including collaborative learning recognition based on user editing behaviors, recognition based on user query reconstruction and automatic judgment based on confidence scoring; a structured preference training data set is generated according to the user preference signal and comprises a plurality of preference pairs, and each preference pair comprises user questions, correct answers and wrong answers; the authenticity and quality of the data are effectively improved, the problems of reward model cracking and model catering are relieved, and continuous optimization and rapid iteration of model performance are achieved.

The embodiment of the invention provides an automatic parking method, device and equipment and a vehicle. The method comprises the steps that parking environment information and configuration information of a to-be-parked vehicle are acquired, and the parking environment information comprises the distance between an obstacle and the to-be-parked vehicle and azimuth information of the obstacle; the configuration information of the to-be-parked vehicle comprises the vehicle head terrain clearance, the vehicle tail terrain clearance and the vehicle size. Based on the deep learning recognition model, target parameters of the target obstacle are determined according to the parking environment information, and the target parameters of the target obstacle comprise the height of the target obstacle. And determining a parking mode of the to-be-parked vehicle based on the target parameter of the target obstacle and the configuration information of the to-be-parked vehicle. And based on the parking mode of the to-be-parked vehicle, the parking environment information and the to-be-parked vehicle configuration information, a parking path is determined to realize parking. The invention aims to improve the safety and success rate of automatic parking in a complex scene.