896results about How to "Reduce hardware costs" patented technology

The invention discloses a deep integrated high-voltage system of an extended-range vehicle and a power-on and power-off control method, and relates to the field of new energy high-voltage control. The system comprises a power domain controller, a range extender system, a driving system, a power battery system, a charging system, a thermal management system and a high-voltage auxiliary driving system; the power domain controller is used for integrating control functions of a whole vehicle controller, a battery management controller, a range extending controller and a motor controller on a vehicle so as to realize control management of the whole vehicle; the range extender system is used for executing a control instruction of the power domain controller so as to charge the power battery; the driving system is used for driving the vehicle and recovering braking energy; the power battery system is used for providing electric energy input for the whole vehicle; and the charging system is used for charging the power battery system. The hardware cost of the whole vehicle can be reduced, and later maintenance and upgrading are facilitated.

The invention relates to the technical field of charging management, and discloses an intelligent charging management method, device and equipment for an electric scooter and a storage medium, and the method comprises the steps: carrying out the parameter collection and initialization setting of a single-phase bidirectional converter and a permanent magnet synchronous motor in the electric scooter, obtaining system basic parameters and communication data, and obtaining a network authentication identifier and a charging station operation parameter. And dynamic planning of a charging strategy is executed according to the parameters to generate a segmented charging control curve. And the data are input into the self-adaptive prediction controller to realize real-time adjustment in the charging process, and charging state data and power compensation parameters are obtained. And adopting a double-neural network model structure to obtain a battery safety feature matrix and a fault probability vector. And constructing a three-dimensional charging state space model to carry out optimization calculation on the power compensation parameters, inputting a deep reinforcement learning network to carry out iterative optimization, and generating charging control parameters of the next period, thereby improving the charging efficiency and the electric energy quality of the electric scooter.

The invention discloses a large language model distillation method, device and equipment and a storage medium, and relates to the technical field of computers, and the large language model distillation method comprises the steps: determining the teacher occupancy measure of a teacher model and the student occupancy measure of a student model based on a Markov decision process and a teacher model generation process; performing loss function definition through the teacher occupation measure, the student occupation measure and the JS divergence, and determining a training loss function of the student model; and training the student model according to an expert data set, a target discriminator and the training loss function to obtain a target distillation model, and performing a language processing task through the target distillation model. Through the mode, the distillation process combines the advantages of black box knowledge distillation and white box knowledge distillation, the time cost and hardware cost of model distillation are reduced, and the model performance and universality of the small distillation model are ensured.

The invention discloses a method and a device for detecting and positioning surface defects of a PCB (Printed Circuit Board). The method comprises the following steps: acquiring a reference image and a sample image of the PCB; processing the light and shade distribution of the sample image to obtain a sample image with uniform illumination; correcting a pixel difference between the sample image and the reference image to enable the reference image and the sample image to achieve pixel-level alignment; carrying out defect detection index calculation on the reference image and the sample image according to regions, and detecting and positioning defects. According to the method, the detection accuracy and efficiency are improved, the hardware cost and the calculation overhead are reduced, and deployment and implementation in the actual production and manufacturing process are easy; meanwhile, due to the high automation attribute, the detection reliability and the flexible adjustment capability, the method has wide application prospects and practical value.

The invention belongs to the technical field of underwater positioning, and discloses an underwater single beacon virtual long baseline positioning method and system capable of estimating an unknown sound velocity. The method comprises the following steps: constructing a virtual long baseline array by combining stored dead reckoning data; constructing a virtual long baseline positioning problem by combining the relationship between the underwater acoustic signal propagation time and the position and sound velocity of the AUV; auxiliary variables are introduced, a pseudo-linear positioning equation is obtained based on algebraic conversion, and an AUV position initial solution is obtained based on a weighted least square method; and constructing a second-step linear estimation problem by exploring the relationship between the auxiliary variable and the AUV position, and solving by using weighted least squares to obtain a more accurate AUV position estimation solution. Compared with a single-beacon positioning method which does not consider sound velocity unknown, the method provided by the invention can obtain more ideal position estimation performance, and the practical application capability of an underwater single-beacon positioning system is enhanced.

The invention discloses a roadside parking space management method and related equipment, and the method comprises the steps: continuously carrying out the real-time laser radar sampling of a to-be-detected parking space according to a first sampling period, and obtaining a sampling result; determining a parking space state at a corresponding moment based on a sampling result, and storing the parking space state and / or corresponding license plate information to a cloud; obtaining state transition information of the to-be-detected parking space based on a plurality of parking space states stored in the cloud; and obtaining charging data of the corresponding vehicle based on the state transition information and the license plate information. According to the invention, real-time laser radar sampling is continuously carried out on the to-be-detected parking space to determine the state transition information of the to-be-detected parking space, so that the vehicle entering and exiting condition in the parking space is identified, charging management is carried out on the vehicle, and unmanned management is realized.

The invention provides a multi-camera hard synchronization method based on a microcontroller, and relates to the technical field of camera synchronization, and the method comprises the steps: enabling a plurality of cameras to be sequentially connected through a synchronization interface, and enabling each camera to be connected with a switch; defining a synchronization source; when each pixel clock runs, the microcontroller of each camera is taken as a core, the microcontroller detects whether a synchronization pulse exists or not, if yes, synchronization is carried out, and when an asynchronous phenomenon occurs, the next camera can receive a synchronization signal sync in of the previous camera, and the microcontroller can judge whether the current synchronization time point is a proper synchronization time point or not; if the synchronization time point is not suitable, waiting for the current camera to capture the current frame and waiting for the next synchronization time point to perform synchronization, if the camera does not receive the synchronization signal sync in of the previous camera, performing internal synchronization after capturing one frame by the microcontroller, and after the internal synchronization is completed, retaining the synchronization state information of this time; and checking in the next execution period.

The invention provides a multi-path street lamp data acquisition and analysis method based on a centralized controller, relates to the field of intelligent measurement, and realizes intelligent monitoring and fault detection of running states of street lamps through group control, centralized data acquisition and dynamic analysis of the street lamps. According to the method, grouping is carried out based on a distribution network connection relation of street lamps, switching operation of each street lamp group is controlled through a centralized controller, and operation data such as current, voltage and load are collected. And for the street lamp which operates abnormally, through multiple times of data acquisition and trend analysis, an accurate fault detection result is generated. The invention further provides a method for dynamically adjusting the grouping strategy and the delay time, and the accuracy and the detection efficiency of data acquisition are improved. According to the invention, comprehensive monitoring of the operation state of the street lamp in a complex environment can be realized under the condition that the hardware cost is limited, and the method is especially suitable for management and maintenance of an urban intelligent lighting system.

The invention discloses a control method and a control system of an ultrasonic therapeutic apparatus and the ultrasonic therapeutic apparatus, and the method comprises the following steps: S1, transmitting ultrasonic waves to skins at a plurality of target positions one by one for sampling, and receiving an ultrasonic echo signal of each target position, the plurality of target positions are obtained by a transducer of the ultrasonic therapeutic apparatus moving along a skin surface area directly facing the end face of a probe of the ultrasonic therapeutic apparatus; s2, converting the received ultrasonic echo signal into an ultrasonic image; s3, skin features of each target position in the ultrasonic image are extracted, the distance between the skin of each target position and the end face of the probe is calculated, the calculated maximum distance is compared with a target threshold value, and if the maximum distance is smaller than or equal to the target threshold value, it is judged that the end face of the probe is effectively attached to the surface of the skin; otherwise, judging that the end face of the probe is not effectively attached to the skin surface. Positioning is carried out through ultrasonic imaging, it is ensured that operation is carried out in the state that the end face of the probe is effectively attached to the skin surface area, and the using effect and safety are ensured.

The invention relates to an image ranging method and system, computer equipment and an endoscope system. The method comprises the following steps: acquiring a to-be-processed image which is shot by a camera module under the irradiation of a light source module and comprises a shot object, so as to calculate difference data between brightness data of a first area and brightness data of a second area in the to-be-processed image, and the distance between the camera module and the shot object is determined according to the difference data of the to-be-processed image. Wherein the second area is arranged around the first area. According to the method, distance measurement is carried out based on the image acquired by the camera module without depending on additional hardware equipment, so that the hardware cost of distance measurement is reduced.

The invention relates to the technical field of heating temperature prediction, and relates to a heating room temperature prediction method, a medium, equipment and a heating room temperature prediction system, and the method comprises the steps: obtaining target information corresponding to a target house; inputting the target information into a predetermined heating room temperature prediction model to obtain a predicted room temperature corresponding to the target house; wherein the heating room temperature prediction model is a model obtained by training according to first historical heating data, and the first historical heating data comprises house information and actual room temperature of a plurality of houses, historical water supply temperature and historical water return temperature of a heating station for supplying heat to the houses, and historical meteorological information when the actual room temperature is collected; the actual room temperature is acquired by temperature sensors deployed in the houses, and the plurality of houses are part of houses in the building. According to the method, the prediction precision is higher, the hardware cost and the installation cost are remarkably reduced, the power consumption of the heating system is reduced, and the external temperature change can be responded in real time.

The invention discloses a cross-platform file incremental synchronization method, a computer device, a medium and a product. The method comprises the following steps: monitoring a determined to-be-monitored file in real time, obtaining file change information, dynamically obtaining a target coding algorithm and a target compression ratio, processing the file change information, obtaining intermediate file information, distributing a unique identifier and unique verification information for the intermediate file information, and obtaining target file data; detecting and acquiring a plurality of target network paths; and according to the network delay information, the bandwidth information and the reliability information, determining an optimal transmission path in the plurality of target network paths, and then transmitting the target file data on the path. The method is suitable for different operating systems and platforms, the optimal network path can be detected and selected for file transmission, and it is ensured that cross-platform file synchronization can be efficiently, stably and safely carried out under different network conditions.

A control model acquisition method and system, a controller and a control method for compactness online monitoring in an asphalt mixture paving construction process, comprising the steps of: step 100: establishing a paving system dynamics model, step 200: establishing an acceleration signal detection system, step 300: establishing signal filtering noise reduction, step 400: establishing field calibration and online monitoring, and step 400: establishing an acceleration signal detection system, step 400: establishing an acceleration signal detection system, step 400: establishing an acceleration signal detection system, and step 400: establishing an acceleration signal detection system. Modeling analysis is carried out on corresponding dynamic characteristics of a vibrating mechanism and a vibrating mechanism by constructing a dynamic model of the compaction system, and a screed acceleration signal measurement and acquisition system is constructed based on the modeling analysis, so that a vibration signal filtering processing method is provided, and an effective acceleration analysis signal is provided; a numerical model is established through measurement and calibration to realize on-line monitoring of the paving compactness of the asphalt mixture, and operation parameters of a vibrating mechanism are guided based on a kinetic model.

The invention belongs to the technical field of hydraulic engineering, and provides a sediment particle size distribution measuring system and method.The method comprises the steps that images of sediment particles in a water body are collected, and a first sediment particle image is obtained; processing the first sediment particle image to obtain a sediment grayscale image; a K-means clustering algorithm is adopted to carry out clustering analysis on the sediment gray level image, and an image with a first-level clustering result is obtained; carrying out clustering analysis on sediment particles with particle sizes smaller than a preset threshold value in the image with the first-level clustering result by adopting a DBSCN clustering algorithm to obtain an image with a second-level clustering result; and carrying out edge detection on the sediment particles in the image with the secondary clustering result, and then sequentially carrying out binarization processing, equivalent circle diameter transformation and statistics on the particle size distribution of the sediment particles to obtain a particle size distribution result. The problem that an existing sediment particle size measurement system is low in measurement precision in a complex underwater environment with dense sediment and more fine particles is solved.

The invention discloses an adaptive power control method for a duplex-winding induction generator system, and belongs to the technical field of generator control. The double-winding induction generator system is composed of a double-winding induction generator, an energy storage unit, a static excitation converter and a diesel engine. According to the self-adaptive power control method, control winding current is decoupled into excitation control and electromagnetic torque control through coordinate transformation based on rotor flux linkage orientation, independent adjustment of output power of the generator is achieved, passive regulation and control of power of an energy storage unit are achieved through control of electromagnetic torque of the generator, and the self-adaptive power control method is achieved. The problem that the control freedom degree of a single converter system is insufficient is solved. Besides, according to the control method adopted by the invention, a dynamic planning-based generator / energy storage unit power distribution strategy with minimum fuel loss as a target is adopted, so that the generator is ensured to always work in a high-efficiency interval of the diesel engine within a wide load change range, and the power generation cost is remarkably reduced.

The invention provides a method and device for discharging residual charges of an air conditioner and the air conditioner. The air conditioner at least comprises a capacitor charging control circuit, a bus capacitor circuit, a fan driving circuit, a main control chip and a fan load. The discharging method comprises the steps that the main control chip responds to a bus capacitor discharging instruction and outputs a control signal used for controlling a capacitor charging control circuit to be disconnected; and obtaining a capacitor voltage value of a bus capacitor in the bus capacitor circuit, and controlling the fan driving circuit to drive the fan load to work through the obtained capacitor voltage value. According to the invention, rapid and efficient discharge of the bus capacitor can be realized, and the hardware cost and the system complexity are reduced on the premise of not influencing the working efficiency of the air conditioner.

The invention relates to the technical field of image processing and machine vision, in particular to an anti-occlusion target tracking method integrating Kalman filtering and particle filtering, which is low in cost, low in complexity, high in real-time performance and high in stability, and compared with the prior art, the robustness and the accuracy of a target tracking algorithm in an occlusion scene can be remarkably improved, and the target tracking method has the advantages of high robustness and high robustness. Target tracking loss caused by shielding is effectively avoided; according to the invention, the anti-shielding function is realized only depending on the algorithm, the hardware cost of the system is reduced, and the target tracking algorithm is simpler, more convenient and lighter in engineering application, so that the anti-shielding target tracking method provided by the invention has the characteristics of low cost, low complexity, strong real-time performance and high stability; the target can be continuously tracked in a complex scene, and the method has important application value and wide application prospect.

The invention discloses a quadruped robot obstacle avoidance method and system based on cooperation of a laser radar and a depth camera, and relates to the technical field of robot obstacle avoidance. The system comprises a quadruped robot, and the quadruped robot is provided with an industrial personal computer, a laser radar, a laser radar protection device and a depth camera. When the robot walks, the laser radar obstacle avoidance method is executed firstly, when a laser radar obstacle avoidance program fails due to the conditions of large blind areas and the like, the robot continues to walk until the laser radar obstacle avoidance program appears in the detection range of the depth camera, the depth camera obstacle avoidance method is executed for obstacle avoidance judgment again, and selection is made. The laser radar and the depth camera are combined for use to realize obstacle avoidance, and the depth camera is used as a blind compensation means and supplementation of laser obstacle avoidance, so that the safety is further improved. Processing depth data acquired by the depth camera by using a random forest; the operation amount is small and the processing speed is high.

The invention relates to the technical field of batteries, in particular to a battery life prediction method and device, a vehicle and a storage medium, and the method comprises the steps: obtaining the adjacent historical data of a target battery at the current moment under the condition that the target battery is detected to be in a rapid attenuation stage based on the current capacity attenuation slope, and extracting health features of the target battery, obtaining an input feature vector to train a preset neural network, inputting the time sequence feature vector of the target battery into the trained neural network to carry out rolling prediction until a preset capacity limit value is reached, and obtaining a life prediction result. According to the embodiment of the invention, the rapid attenuation stage of the battery can be accurately identified, the attenuation stage approaching historical data is used for training the long-short-term memory neural network, the influence of early attenuation slow historical data of the battery on the capacity prediction precision is effectively avoided, and the residual life prediction accuracy is improved; the operation time and hardware cost of the model are reduced, and the method is safer and more efficient.

The invention provides an engine test verification system and method based on a virtual rack, and belongs to the field of engine testing. A control upper computer is used for acquiring a control instruction, controlling the virtual rack to start, setting the input quantity and initial test conditions of the virtual rack, and receiving data fed back by the virtual rack in the running state of the virtual rack; the test upper computer is used for calibrating test parameters of the ECU and receiving, displaying and storing real-time data of the ECU when a test is executed; the ECU receives the control quantity of the actuator fed back by the virtual rack and the test parameters sensed by the sensor, compares the control quantity and the test parameters with a target value required by the engine under the current test working condition, and controls the control quantity of the actuator of the virtual rack through a closed loop; according to the invention, the simulation of the engine is carried out by using the virtual rack, the test of the engine on the real rack is effectively replaced, and the flexibility and convenience of the test are improved.

The invention discloses a database AI question answering system and method based on a large language model and retrieval enhancement, and relates to the technical field of database retrieval. The system comprises a front-end interaction module, a distributed cache management module, a task scheduling center, a large language model processing sub-module, a retrieval enhancement sub-module, an answer optimization sub-module and a database management sub-module. All the sub-modules are mutually independent and work cooperatively, through an efficient task scheduling mechanism, the system can monitor the load and resource states of all the sub-modules in real time, accurate allocation of tasks is conducted according to an intelligent algorithm, in a large-scale concurrent request scene, single-point faults and resource waste are effectively avoided, stable operation of the system is ensured, and meanwhile, the system can be widely applied to large-scale concurrent requests. The distributed cache management module further optimizes the storage and access of the data, reduces repeated data query and transmission, and improves the overall performance.

The invention relates to a cleaning system, the cleaning system comprises a cleaning device and a base station, the cleaning device comprises a dust box, a first dust collection unit communicating with the dust box and a battery electrically connected with the first dust collection unit, and the first dust collection unit is used for absorbing garbage outside the cleaning device into the dust box; the base station comprises a dust collection cavity and a second dust collection unit communicated with the dust collection cavity; the cleaning equipment further comprises a controller, and when the current voltage of the battery is smaller than or equal to a first voltage threshold value, the controller controls the cleaning equipment to return to the base station; the cleaning system has a dust collection mode, and when the cleaning system is in the dust collection mode, the second dust collection unit is started to absorb garbage in the dust box into the dust collection cavity. While the dust collection efficiency is improved, the base station can be equipped with the fan with smaller power, and meanwhile, the fan noise during working is also reduced.

The invention discloses a high-speed direct digital frequency synthesizer based on an optimized CORDIC algorithm. The high-speed direct digital frequency synthesizer comprises a phase accumulator, a CORDIC converter and a digital-to-analog converter (DAC). Wherein the phase accumulator is responsible for accumulating input phases and properly intercepting the input phases. The CORDIC converter is based on an optimized CORDIC algorithm. The CORDIC converter is composed of a small-angle extraction module, a lookup table module, a rotation calculation module and a domain folding reduction module. The small angle extraction module converts a phase angle into a small angle within pi / 12, and the lookup table module only stores 11 groups of sine and cosine values divided for pi / 12. And the data after meter reading is firstly subjected to scaling-free processing, then is subjected to parallel three-group rotation calculation processing of a remainder four algorithm, finally, a signal is restored in the domain folding and restoring module through a sum angle formula, and finally, a digital signal is converted into an analog signal by a DAC (Digital-to-Analog Converter). By optimizing the CORDIC algorithm, the precision and speed of the frequency synthesizer are remarkably improved, and the storage requirement is greatly reduced, so that the production cost in large-scale chip manufacturing is greatly reduced.

The utility model discloses an MCU general test platform with high test coverage. The MCU general test platform comprises an excitation board, a mother board, a daughter board, a tested MCU chip, an upper computer and peripheral test resources, the tested MCU chip is connected with the daughter board through a chip test seat on the daughter board, the daughter board is connected with the mother board and used for sending a test command received by the mother board to the daughter board, and the daughter board transmits test data to the mother board; the mother board is connected with the excitation board and is used for sending a test instruction to the mother board by the excitation board, and the mother board transmits test data to the excitation board; the excitation board is respectively connected with the upper computer and the peripheral test resource and is used for sending a test command to the excitation board by the upper computer, the excitation board uploads test data to the upper computer, and the peripheral test resource assists the excitation board in function test; the device has the advantages that full-function testing can be carried out on MCU chips of different models and different packages, and the testing efficiency is improved; functional tests under different environment temperatures are completed through the incubator, the hardware cost of the test device is saved, and the scheme is simple to disassemble and easy to transport and store the test device.

The invention discloses a parameter design method, device and system for a hybrid electromagnetic interference filter, and belongs to the technical field of power electronic control. The target passive device parameter corresponding to the minimum common mode inductance meeting the preset condition is found; the preset condition is that under the control of the target passive device parameter and the corresponding target active control parameter, the corresponding attenuation degree difference value of the hybrid electromagnetic interference filter in the concerned frequency band is smaller than a threshold value; and outputting the target passive device parameter and the corresponding target active control parameter. According to the invention, with the attention frequency band attenuation degree as a target, the attenuation degree of the HEF can approach to the difference between the actual interference and the standard limit value to the maximum extent, and unnecessary redundancy design is effectively avoided. The self-adaption process of the attenuation degree can fully develop the interference attenuation performance of the AEF, and the performance requirement for the PEF is reduced.

The embodiment of the invention discloses a bus voltage sampling circuit and a power supply system, and the circuit comprises a high-voltage sampling unit which is used for sampling a DC bus voltage and outputting a first voltage signal, and is connected with a first controller; the signal conversion unit is used for converting the first voltage signal into a current signal and outputting the current signal, and is connected with the output end of the high-voltage sampling unit; and the signal restoration unit responds to the current signal to output a second voltage signal, is connected with the output end of the voltage-controlled current source, and is also connected with the second controller. According to the embodiment of the invention, the direct current bus voltage of the first controller is converted into the current signal, then the current signal is restored into the voltage signal, and the voltage signal is input to the second controller, so that many mismatching problems of front and rear controllers are solved, and sampling of the direct current bus voltage is completed by one sampling circuit; the power consumption, the hardware cost and the PCB area caused by an extra sampling circuit can be avoided.

The invention provides a current signal iteration Vold-Kalman filtering detection method for a rotating machinery fault, and belongs to the technical field of rotating machinery, and the method comprises the steps: carrying out the coordinate transformation of a three-phase stator current during the operation of the rotating machinery, and obtaining Id; obtaining a time-frequency spectrum of the original current, constructing a cost function, and extracting an instantaneous frequency ridge line of the fundamental wave of the original current; according to the instantaneous frequency ridge line and the component design parameters, calculating the characteristic frequency when each component breaks down; an iterative Vold-Kalman filter is constructed, the calculated fault frequency and the frequency multiplication of the fault frequency are taken as cut-off frequencies, and fault harmonics in the current signal Id are iteratively checked and extracted through a substitution check method; and carrying out time-frequency analysis on the extracted fault harmonic waves, and identifying the fault type. According to the method, a non-intrusive method is adopted, the non-stationary characteristic of the signal is considered, the calculation complexity is reduced, meanwhile, the fault harmonic extraction precision is improved, and the rotating machine fault can be effectively detected through the current signal under the variable rotating speed working condition.

The invention discloses an integrated wafer pre-adjusting part and a wafer processing device, the integrated wafer pre-adjusting part comprises a first lifting platform, an adjusting mechanism, a detector and a code scanner, and the adjusting mechanism comprises a hook claw, a turntable, a first driving assembly and a second driving assembly; the first driving assembly is matched with the hook claw to move the wafer at the preparation station to the adjusting station, so that the wafer is positioned; the second driving assembly is matched with the turntable to rotate the wafer, so that the detector and the code scanner can confirm the notch position and coding information of the wafer, and edge searching and code scanning of the wafer are realized; the left and right sides of the wafer returning to the preparation station are limited by the slots, and the front and back positions are limited by the output stroke of the first driving assembly, thereby achieving the centering of the wafer. According to the wafer processing device, the wafers are pre-adjusted quickly through the integrated wafer pre-adjusting part, each wafer in the wafer processing device can have a unified and accurate position and posture before processing, and accurate data and a positioning basis are provided for subsequent wafer processing procedures.

The invention provides a data backup method and device, electronic equipment and a storage medium, and relates to the technical field of data storage. The method comprises the following steps: in response to a data writing request for target data, acquiring node information corresponding to each node set in a data backup system; each node set comprises a first node for data sharing between a working main node and a plurality of standby main nodes in the data backup system, and a second node for determining a new working main node from the plurality of standby main nodes when the working main node fails; based on the information of each node, screening out two target slave node sets meeting a first node quantity requirement and a second node quantity requirement from the plurality of candidate slave node sets; the target data are backed up through the two target slave node sets, and storage positions corresponding to the multiple pieces of backup data are recorded. Therefore, the data management cost required by data backup is reduced while complete data backup is realized.

The invention discloses an image recognition method for confrontation domain self-adaption and a conjugate gradient kernel extreme learning machine. The method is characterized by comprising the following steps: firstly, constructing a feature extractor to obtain initial features of a source domain and a target domain; and then constructing an adversarial domain adaptive network to optimize two domain features, enabling a source domain feature extractor and a target domain feature extractor to play a game with a discriminator so as to reduce inter-domain differences between a target domain and a source domain, projecting the inter-domain differences to a common feature space, and finding an optimal target domain feature. And finally, constructing a classifier to recognize the image, and constructing a conjugate gradient kernel extreme learning machine as a classifier to classify the optimal target domain features to obtain a classification result. The invention aims to provide a lightweight network, namely an adversarial domain adaptive and conjugate gradient kernel extreme learning machine, which solves the problem of dependence of existing deep learning on a large amount of annotation data, reduces sample annotation cost in image recognition, reduces hardware implementation cost, realizes cross-domain knowledge migration, and completes unsupervised and high-precision image recognition of a target domain.