366 results about "Direction detection" patented technology

The invention discloses a directional detection method and system for a power supply shell based on a multi-source heterogeneous sensor, belongs to the field of material characteristic detection, and aims to solve the problems of single dimension, weak anti-interference performance, data isomerism and poor adaptability in traditional industrial detection. Through systematic innovation of multi-source sensor collaboration, dynamic mode control, cross-modal data fusion and self-learning optimization, a high-precision, high-efficiency and high-reliability solution is provided for the precision manufacturing field, and an optical, infrared and ultrasonic multi-modal collaborative detection framework is constructed; the optical unit realizes accurate capture of surface texture and morphology through polarized light imaging and 3D structured light scanning, the infrared thermal imaging unit analyzes the heat conduction characteristic of a material, and the ultrasonic array analyzes internal structure defects to form a surface-material-internal full-dimension detection capability; and nine types of defects such as scratches, pits, weld marks, bubbles, sink marks, material layering, stress cracking, thermal stress abnormity and material pollution are covered.

The invention provides an engineering drawing intelligent processing system based on a ComfyUI, and the system comprises an algorithm service module which integrates a plurality of drawing processing algorithm assemblies, such as drawing layout analysis, OCR recognition, drawing direction detection, semantic understanding, drawing frame disassembly, and axis net recognition, and provides services for the outside through a standardized interface; the workflow arrangement module is used for packaging the algorithm components into parameterized drawing processing nodes based on a visual node configuration framework of a ComfyUI, and supporting flexible combination, real-time preview and batch execution of drawing processing flows; and the workflow management module is used for importing, analyzing and publishing the workflow configuration file in the ComfyUI format, and realizing metadata management, publishing verification, API document generation, interface registration and call authority control of the workflow. The method supports multi-type input of PDF and CAD drawings, is suitable for engineering drawing processing tasks such as drawing label extraction, sub-drawing recognition, drawing frame segmentation and drawing filing, and is high in intelligent degree, flexible in process configuration, easy to integrate and high in expansibility.

The invention discloses size detection equipment for casting perforation processing, and relates to the technical field of size measurement. The device comprises a base, a supporting plate and a displacement assembly. The detection part goes deep into a workpiece hole through the contact ball, and the laser emitter and the light sensor are used for detecting the contact position, thereby realizing high-precision size measurement. The displacement assembly adopts the design of a negative pressure rotating cylinder and a resistance sliding rod, adjusts the displacement of a contact ball according to the rotating speed, and flexibly adapts to different hole requirements. Through the control of the planetary gear and the electromagnet, the contact ball can rotate to realize multi-angle detection. The detection table is driven by a central electric cylinder and an edge inclination control electric cylinder to stably adjust the posture of a workpiece. The device is high in automation degree and convenient to operate, and overcomes the defects of low precision, poor efficiency and single-direction detection of traditional manual measurement.

The invention relates to the technical field of sensors, in particular to a bionic hair tactile sensor and application, and the sensor comprises a substrate and a bionic cilia array arranged on the substrate; the bionic cilia array comprises a plurality of flexible bionic cilia arranged in an array, and a marking layer is arranged at the bottom of the flexible bionic cilia; an optical system is integrated in the substrate, the optical system comprises an imaging module and a micro prism, and the micro prism reflects the deformation of the marking layer at the bottom of the flexible bionic cilia to the imaging module; and the imaging module is used for capturing deformation data of the marking layer and sending the deformation data to the deep learning algorithm module to realize decoupling and resolution of three-dimensional force. According to the invention, high-resolution multi-dimensional force detection and direction detection are realized through mark point change caused by cilium deformation, and the array design is expanded to a sliding sensing function and a multi-modal object identification function.

The invention discloses a clar steady-state filtering flow velocity and flow direction detection method, and belongs to the technical field of target detection, and the method comprises the steps: collecting and processing the real-time data of a fluid radar signal, and determining the target data of the fluid radar signal; analyzing and identifying the fluid radar signal target data, and determining a clar steady-state filtering flow velocity and flow direction detection result; and visually displaying the clar steady-state filtering flow velocity and flow direction detection result, and timely managing and controlling the fluid according to the clar steady-state filtering flow velocity and flow direction detection result. According to the invention, the problems of low flow velocity and flow direction detection precision and poor effect caused by the fact that the flow velocity and flow direction of the fluid cannot be effectively detected according to cfu steady-state filtering in the prior art are solved. The flow velocity and the flow direction of the fluid can be effectively detected according to cfu steady-state filtering, so that reliable flow velocity and flow direction detection results can be provided under various environmental conditions, and the flow velocity and flow direction detection precision and the flow velocity and flow direction detection effect can be improved.

The invention discloses a strength detection tool and method for a composite cabin for an unmanned aerial vehicle. The detection tool comprises a cabin connecting plate, an end fixing device, a force application adjusting device and a deformation measuring structure. The force application adjusting device comprises a one-way stress application structure, a torsional stress application structure and a stress transmission structure. The stress transmission structure is provided with a special-shaped stress plate, a one-way stress transmission piece and a torsional stress transmission assembly. Particularly, a multi-linkage test structure is arranged, and conversion from torsional moment to radial one-way pressure is realized through geometric constraint and contact mechanics principles. According to the detection method, a step-by-step loading and multi-direction detection mode is adopted, initial detection, surface change detection and multi-direction comprehensive detection are achieved, and the preset deformation quantity is rapidly achieved through a displacement control mode by means of the qualified deformation quantity value detected last time. According to the invention, bending moment and torque load detection in four directions of 0 degree, 90 degrees, 180 degrees and 270 degrees can be completed through one-time clamping.

A mobile device sends to a radio communicatively coupled to the mobile device, an instruction to monitor a frequency band. It is determined that the frequency band is jammed by a jamming source. The mobile device receives, from the radio, signal waveform information corresponding to the frequency band that is received by the radio from a 360 span about the radio. Based on the waveform information a direction of the jamming source is determined. The mobile device presents on a display device information that indicates that the frequency band is being jammed, and a direction of the jamming source with respect to a current location of the radio.

The present application belongs to the field of machine vision defect detection, and in particular to a thermocell electrode sheet vision defect detection method and a vision defect detection device. Product size information of a thermocell electrode sheet is obtained, and the field of view of a camera is configured. A thermocell electrode sheet to be detected is placed on a detection platform, front surface images, side surface images, back surface images, and inner hole images of the thermocell electrode sheet are captured by multiple direction detection cameras, and are processed to obtain the to-be-detected images of the front surface images, the side surface images, the back surface images, and the inner hole images. According to the to-be-detected images of the front surface images, the side surface images, the back surface images, and the inner hole images, a trained YOLO8 recognition model is used to describe and determine defects.

The invention provides a conference environment intelligent sound field control system. A main sound source module, a plurality of sound field node intensity adjusting modules, a human body recognition positioning module and a processing control module work cooperatively. The main sound source module integrates the functions of main sound production, direction adjustment, sound source reception and direction detection, and can automatically track the position of a speaker and adjust the direction of a sound source. And the sound field node intensity adjustment module optimizes the sound field effect at each part of the conference space through node sound reception, compensation sound production, intensity adjustment and node processing. The human body recognition positioning module tracks the positions of participants in real time and provides the positions to the processing control module for intelligent decision making. The system can automatically adjust the direction and intensity of the sound field according to the distribution of participants, guarantees the precise transmission of sound, and reduces the interference of environmental noise. According to the intelligent sound field control scheme, conference voice definition and communication efficiency are remarkably improved, conference scene requirements of various scales and layouts are met, and an intelligent and efficient acoustic solution is provided for a conference environment.

The invention discloses a multi-wavelength laser collimator based on radial shearing. The multi-wavelength laser collimator comprises a near-field reference, a telescope objective combination, a bifocal objective and an image detector which are sequentially arranged along a light path. Aberration-free imaging of near-field light beams is achieved through a bifocal objective lens, and meanwhile far-field pointing is detected through coaxial shearing interference generated by means of amplitude light splitting. A single non-mobile sampling detection light path and a single detection sensor are used to realize simultaneous acquisition of near-field position information and far-field pointing information of a laser beam; the light spot position deviation and the fringe translation amount are analyzed through a computer, and high-precision collimation is achieved through a feedback control system. And the detection of the angle direction of the laser beams with various wavelengths can be realized by using a radial shearing interference technology. The device breaks through the limitation of traditional double-optical-path collimation, has the characteristics of simple structure, strong anti-interference performance, multi-wavelength applicability and the like, remarkably improves the collimation precision, reduces the cost, and has important application value in the field of laser precision collimation.

The invention discloses an anti-interference mounting structure of an inductive sensor, and aims to solve the technical problems of insufficient mounting space and limited application scene caused by high requirement on interference-free distance of peripheral metal of the conventional TURCKNI100U-K90SR-VP4X2 type inductive sensor, the structure comprises a metal protection box, an electromagnetic shielding assembly and a sensor body; the electromagnetic shielding assembly comprises two electromagnetic shielding films with the size of 25 * 10 * 0.5 mm, and the two electromagnetic shielding films are attached to the two sides of the upper portion of the sensor body respectively. The top of the metal protection box is provided with a non-transparent anti-aging plastic cover plate, one end is provided with a through hole for wire connection, the inner wall of the bottom is provided with a sensor mounting blind hole, and the outer wall is provided with an equipment fixing guide hole. The metal interference-free distance around the sensor is shortened to 18.5 mm through local electromagnetic shielding, meanwhile, it is guaranteed that the detection distance of the sensor in the vertical direction is not attenuated, sealing protection of the sensor is achieved through the protection box, the application scene of the high-detection-distance inductive sensor is remarkably expanded, and the high-detection-distance inductive sensor is particularly suitable for mechanical equipment in a narrow space.

The utility model provides an airing height self-adaptive system and a clothes airing machine, and relates to the technical field of airing. The airing height self-adaption system is applied to the clothes airing machine and comprises a main control module arranged in a main machine of the clothes airing machine and a distance detection module arranged on an airing rod of the clothes airing machine. The distance detection module is in communication connection with the main control module, and is used for detecting existence of a detection object within a preset distance from the airing rod to a preset direction and sending a detection result to the main control module; and the main control module is used for controlling the airing rod to stop lifting through a power system of the clothes airing machine according to the received detection result. According to the system, the distance detection module connected with the main control module is arranged, the airing rod of the clothes airing machine can stay at the airing height matched with the height of a detected object on the basis of the structure and line structure, airing height self-adaption of a user of the clothes airing machine is achieved, and the intelligent degree of the clothes airing machine is remarkably improved.

The invention belongs to the technical field of chemical reaction OCR (Optical Character Recognition), and particularly relates to a chemical reaction OCR method and system based on multi-model collaboration. The recognition method comprises the following steps: (1) reaction image data, (2) a chemical reaction detection model, (3) a reaction role detection model, (4) a compound recognition model, (5) a text recognition model, (6) a direction detection model, (7) result merging, (8) verification frequency judgment, (9) leak detection verification, (10) molecule legal verification, and finally outputting a standardized chemical reaction Smiles expression result. The method has the capability of independently analyzing the chemical reaction equation in the pure image form, and the applicability and robustness in complex documents and unstructured scenes are remarkably improved.

The invention relates to the technical field of meteorological remote sensing, in particular to a three-dimensional wind speed and wind direction detection method and system based on GNSS tomography inversion, and the method comprises the steps: building a three-dimensional water vapor density tomography inversion model through obtaining observation data and meteorological observation data of a plurality of GNSS observation stations, obtaining water vapor density distribution of different height layers, and obtaining a three-dimensional wind speed and wind direction detection model; further obtaining the PWV (atmospheric precipitable water) of each height layer; a three-dimensional wind field inversion algorithm is established by utilizing a coupling relation between water vapor transportation and a wind field and combining a spatial position of a GNSS observation station and spatial and temporal variation characteristics of PWV of different height layers, so that quantitative estimation of wind speeds and wind directions of the different height layers is realized, and three-dimensional wind speed and wind direction quantitative detection based on GNSS tomography inversion can be realized. And important observation data support is provided for meteorological monitoring and numerical weather forecast.

The invention relates to the technical field of sonar detection, and discloses an inspection well underwater pipeline and flow direction detection method, device, equipment and medium, and the inspection well underwater pipeline and flow direction detection method comprises the steps: obtaining sonar detection data of a plurality of target inspection wells by using target sonar scanning equipment equipped with a jacking rod rail device; the target sonar scanning equipment is used for transmitting an ultrasonic original frequency in a target inspection well and receiving a reflected target ultrasonic frequency; performing three-dimensional modeling on the target inspection wells according to the sonar detection data to obtain a target inspection well three-dimensional model of each target inspection well; according to the three-dimensional model of each target inspection well, determining a pipe orifice circle center coordinate of each target inspection well; the water flow direction of each target inspection well is determined according to the ultrasonic offset frequency data corresponding to the coordinates of the circle center of the pipe opening of each target inspection well, the water flow direction of each target inspection well is determined through sonar detection, and the accuracy of determining the water flow direction of the inspection well is improved.

The invention discloses a hangar unmanned aerial vehicle wind disturbance resisting landing system based on multi-mode perception, which comprises an unmanned aerial vehicle, a hangar, a wind speed and direction detection unit and a visual identification unit, and is characterized in that the wind speed and direction detection unit is arranged on the hangar, and the visual identification unit is arranged on the unmanned aerial vehicle; the hangar comprises a base and a parking apron, the parking apron is rotationally connected to the base, and the landing area is arranged on the parking apron; a centering driving device and two bin covers are arranged on the parking apron, and the centering driving device is used for driving the unmanned aerial vehicle to center; the two bin covers are rotationally connected to the two longitudinal sides of the parking apron correspondingly, the bin covers are configured to be switched among a closed state, a windproof state and an open state, and when the bin covers are in the windproof state, the bin covers rotate to the highest position; the centering driving device comprises two transverse push rods and two longitudinal push rods, and the landing area is located between the two transverse push rods and the two longitudinal push rods; during landing, when the wind speed and wind direction detection unit detects that the wind speed is larger than a preset threshold value, the bin cover is controlled to be switched to the windproof state.

The invention discloses an ultrasonic guidance navigation system based on multi-modal detection and a CNB ejection mechanism, and belongs to the technical field of ultrasonic auxiliary diagnosis, multi-modal detection loads a directional bounding box model and a segmentation model when the system is initialized, and the directional bounding box model performs directional detection of a puncture needle on an input frame; in the result extraction stage, puncture needle information is extracted from a directional bounding box model detection result, a focus mask is extracted from a segmentation result, and the focus mask is adjusted to be the same as an input frame in size; when the focus mask is extracted, the validity of a segmentation result is verified, then first effective mask data are extracted from multiple types of segmentation results, the mask data are transmitted to a CPU from a GPU memory, a floating point probability value is converted into an integer binary mask, and puncture needle direction detection and segmentation of multiple types of anatomical structures are achieved. According to the CNB ejection mechanism, a mathematical model of the ejection process is established through an ejection trajectory calculation function, and a reliable mathematical basis is provided for CNB ejection operation diagnosis.

The utility model provides a pipe network flow direction detection device, which comprises a fixed part fixed at the top of a pipe network, the top of a support is hinged with the fixed part, the bottom of the support is respectively and rotatably connected with a floater assembly and a plectrum assembly, and the floater assembly is heavier than the plectrum assembly and is positioned at the top of the plectrum assembly. The floater assembly can synchronously move along with water flow so as to drive the support to rotate relative to the fixing part, a trigger ball and a trigger switch are arranged in the shifting piece assembly, the trigger ball can move relative to the trigger switch, and the shifting piece assembly can synchronously move along with the water flow so as to drive the trigger ball to move relative to the trigger switch. The installation requirement precision is low, the manufacturing and maintenance cost is reduced, and assembly and disassembly are convenient; the support can be driven to rotate through the floater assembly, so that the support is driven to rotate when water flow impact exists, the support is driven to reset when no water flow impact exists, and the automatic reset effect can be effectively achieved.

The invention discloses a geologic body boundary extraction method based on gravity total horizontal derivative anomaly, and belongs to the technical field of geophysical data processing and interpretation. Calculating gravity total horizontal derivative anomaly based on actually measured gravity anomaly data; in the sliding window, detecting whether the gravity total horizontal derivative abnormity meets a preset extreme value structure condition in multiple directions, and screening out boundary points based on an amplitude threshold value and a slope threshold value; according to the invention, a geologic body boundary extraction method based on gravity total horizontal derivative anomaly is established by using directional extremum features of gravity total horizontal derivative anomaly; compared with a traditional boundary recognition method, the method does not depend on prior information, is high in anti-noise performance, can quickly and accurately recognize the boundary position of the geologic body under a complex background, and has higher boundary resolution.

The invention relates to the technical field of electronic design automation, and provides a PCB schematic diagram component information extraction method based on text recognition and a graph algorithm. The method aims at solving the problem that component models in a complex PCB schematic diagram are extracted inaccurately, and the method mainly comprises the steps that preprocessing of binaryzation, denoising and text area enhancement is conducted on the PCB schematic diagram, and an optimized image is generated; performing character direction detection and rotation processing on the obtained optimized image, inputting the fine-tuned OCR model to perform multi-scale character recognition, and outputting a recognition result containing character content, coordinates and confidence; constructing a graph model containing MCU position nodes and character block nodes, and establishing associated edges through a weight calculation module; performing global optimal matching on the graph model based on a Hungary algorithm, and screening associated pairs of the MCU and the character blocks; and performing editing distance calculation on a matching result and a preset MCU model template library, correcting character confusion errors through a multi-stage verification mechanism, and outputting final component information.

The invention provides an unmanned aerial vehicle-mounted vehicle converse running detection method based on lightweight key point detection. The method comprises the following steps: step 1, constructing a data set; five-point labeling of a central point, vehicle width and height and a rotation angle of a UAV-ROD data set needs to be converted into four-point labeling of a bounding box central point, bounding box width and height and a normalized key point; 2, designing a model; the network architecture of the YOLO11 is optimized, and a light detection head LADH is adopted to replace an original detection head to form a YOLO11-LADH model; 3, vehicle direction detection and retrograde driving judgment are carried out; and training the YOLO-LADH model on the UAV-ROD key point data set, inputting the pictures of the view angle of the unmanned aerial vehicle into the model to obtain four key points of the vehicle, and judging the driving direction of the vehicle through the obtained four key points. Through the lightweight detection head LADH, the detection precision and the reasoning speed are improved while the model parameter quantity and the calculation quantity are reduced, the anti-jamming capability is high, judgment is accurate, multi-frame association is not needed, and an efficient solution is provided for real-time traffic management.

To reliably determine the direction of the subject's gaze. [Solution] The gaze direction detection device 1 comprises a camera 10 that acquires a face image by capturing the face of subject A, and an image processing device 20 that processes the face image acquired by the camera 10 to determine the gaze direction of subject A. The image processing device 20 detects the three-dimensional positions of the right pupil, left pupil, and nose based on the face image, identifies the three-dimensional position of the intersection of a second straight line that passes through the three-dimensional position of the nose and intersects perpendicularly with a first straight line connecting the three-dimensional positions of the right pupil and the three-dimensional positions of the left pupil, and the first straight line, and determines the gaze direction of subject A using the first distance between the three-dimensional position of the intersection and the three-dimensional position of the right pupil, the second distance between the three-dimensional position of the intersection and the three-dimensional position of the left pupil, and the third distance between the three-dimensional position of the intersection and the three-dimensional position of the nose.

The utility model provides a storage battery pack direction detection tool, which comprises a mounting plate and at least one row of direction detection components, the direction detection components are mounted on the mounting plate, a plurality of direction detection components in each row are distributed at intervals along the length direction of the mounting plate, and the plurality of direction detection components are arranged corresponding to a plurality of storage battery packs in a storage battery box. The direction detection assembly comprises a circuit board which is provided with a first light emitting diode and a second light emitting diode which are arranged in parallel and are opposite in installation direction, and the light emitting colors of the first light emitting diode and the second light emitting diode are different; and the two conductive probe assemblies are installed on the two sides of the circuit board, electrically connected with the circuit board and used for correspondingly abutting against the positive electrode and the negative electrode of the storage battery pack. Therefore, whether the positive electrode and the negative electrode of each storage battery pack are wrongly installed or not can be visually judged at a time through the light emitting color of the diodes, the detection efficiency is greatly improved, and the production cost is reduced.

The invention relates to a large-aperture spherical optical element bus on-line detection device which comprises a bottom supporting structure, an X-direction detection assembly and a Z-direction detection assembly. Wherein the supporting mechanism is mainly formed by welding a heavy-load type Foma wheel, a top transverse square pipe, a top mounting plate, a supporting stand column square pipe, a reinforcing rib square pipe and a bottom transverse square pipe; the X-direction detection assembly is mainly composed of a grating ruler, an A servo motor, a speed reducer, a large coupler, a lead screw and a linear guide rail. The detection assembly in the Z direction is mainly composed of a contact type length gauge, a servo motor B, a small coupler and a lead screw module. The device is used for online precise detection of generatrix data of a large-aperture spherical optical element by detection personnel, is especially suitable for large-aperture, heavy-weight and difficult-to-overturn spherical optical elements, and can accurately detect the generatrix data of the large-aperture spherical optical element by adjusting the displacement of the precise detection element in the X and Z directions and the measuring range of the contact type length gauge. And on-line detection is carried out on detected optical element buses with different calibers, thicknesses and bus depths.

The utility model discloses a wireless stethoscope which comprises a stethoscope shell and a handheld shell, one end of the handheld shell is fixedly connected with a connecting block, the other end of the connecting block is fixedly connected with the outer wall of the stethoscope shell, and the stethoscope shell is fixedly installed at the end of the handheld shell through the connecting block. A trumpet-shaped opening is formed in one side of the auscultation shell, a mounting ring is fixedly arranged at the position, located on the inner wall of the trumpet-shaped opening, of the auscultation shell, and a sound filtering diaphragm is arranged in the middle of the interior of the mounting ring. According to the utility model, the plurality of microphones are uniformly arranged on the mounting ring in a surrounding manner, so that a microphone matrix is formed on the mounting ring, the sound direction detection sensor on the PLC judges the sound emitting direction according to the sound receiving sequence of the different microphones, and the sound direction is displayed on the display; therefore, a doctor can accurately find the sounding position according to the direction displayed on the display, and the doctor can judge the illness state of a patient.

The utility model provides a moving device with a direction detection function, which comprises a first track, a second track and a moving table, the second track is connected to the first track through a first driving device, and the moving table is connected to the second track through a second driving device; the moving table is provided with a rotating mechanism, the rotating mechanism is provided with an angle detector used for detecting the rotating angle of the rotating mechanism, and the angle detector is installed at the rotating end of the rotating mechanism. The rotating direction is accurately measured through the angle detector, then the control handle is used for controlling movement, and therefore the situation that errors possibly occur through manual judgment is avoided, and the accuracy of the device in the moving process is improved.

To provide a reflector and a reference point surveying method having a simple structure and capable of detecting the incident direction of a light beam to the reflector without changing a measuring reference point regardless of the incident direction of the light beam.SOLUTION: The direction detector includes an image-forming lens (4) and a light-receiving sensor (5), and is configured such that principal rays 11a, 11b, 11c, and 11d that are incident so as to pass through the center of the ball lens and are transmitted through the ball lens are reflected by an outer face of the spherical crown and an inner face of the spherical crown or a face of the ball lens and are then incident on the direction detector. The direction detector is configured to detect the incident direction of the main light beam with respect to the main optical axis based on a light receiving position 5d of the light receiving sensor when the main light beam is incident on the imaging lens and is imaged on the light receiving sensor by the imaging lens.SELECTED DRAWING: Figure 1

An apparatus, method and computer program is described comprising: detecting, at a second user device, one or more display patterns displayed at a display of a first user device, wherein the one or more display patterns comprise features with spatial variations configured to be detected by a radar module of the second user device; determining at least one of an orientation, position, or size of the first user device based, at least in part, on the one or more display patterns; and determining at least one of an orientation or position of the second user device relative to the first user device within a coordinate system of the first user device.

The embodiment of the application provides a window-cleaning robot motor rotation direction detection method and related device, and relates to the cleaning technical field.The method comprises the following steps: obtaining a pulse signal of a detection sensor; the detection sensor comprises a single-phase Hall sensor and / or a single-phase photoelectric sensor; and the rotation direction of the motor is determined according to the characteristics of the pulse signal.The above method replaces the traditional dual-phase Hall sensor or dual-phase photoelectric sensor with the single-phase Hall sensor and / or the single-phase photoelectric sensor, uses the time sequence characteristics of the pulse signal output by the single-phase Hall sensor and / or the single-phase photoelectric sensor to realize high-precision and high-reliability detection of the motor rotation direction, and effectively reduces the complexity of the cost system of the window-cleaning robot.

The invention belongs to the field of Internet of Things, and particularly relates to a pedestrian sight line direction detection method and device based on a top-mounted camera and a medium, and the method comprises the steps: obtaining an indoor scene video stream; inputting the indoor scene video stream into the key point detection model to generate enhanced basic features fused with diversified gradient information; performing dynamic sparse attention calculation on the enhanced basic features to obtain an optimized feature map; performing pooling and convolution processing on the optimized feature map twice to obtain a spatial weight corresponding to the optimized feature map; a pedestrian upper body key point composed of a left shoulder point, a right shoulder point, an afterbrain midpoint and a forehead midpoint is obtained; calculating an initial sight direction according to the afterbrain midpoint and the forehead midpoint; and based on the position and the line-of-sight direction of the same pedestrian in the continuous multiple frames, compensating the initial line-of-sight direction through a filtering algorithm. Accurate, stable and small-calculated-amount sight line direction calculation is realized.