High-rise concealed honeycomb intelligent injection system

Abstract

The application discloses a high-rise concealed honeycomb intelligent spraying system, comprising: a parameter storage module; an image acquisition calibration module, which acquires field image data of a target area; a concealed honeycomb detection module, which determines a target detection area according to the image center coordinates and the spraying center pixel coordinates; and a target positioning module, which generates a target honeycomb area, honeycomb center coordinates and a target effective state according to the honeycomb candidate area data and the confidence parameter. The application acquires image center coordinates, spraying center pixel coordinates and preprocessed image data by collecting, calibrating and preprocessing the field image data of the target area. Further, the application can reduce the problem of unstable target positioning caused by complex background, local shielding or light changes by image recognition of the concealed honeycomb on the outer facade, window edge, eaves or device backside of the high-rise building through color features, contour features, hole texture features and shielding boundary features.

Description

Technical Field

[0001] This invention relates to the field of image recognition and target positioning technology, and in particular to a high-rise concealed honeycomb intelligent spraying system. Background Technology

[0002] With the increasing number of high-rise buildings in cities, bees easily build nests on building facades, eaves, air conditioner units, window sills, pipe angles, and the backs of decorative components. Because these nests are high above the ground, it is usually necessary to use video equipment to acquire on-site images of the target area and identify and locate the nests based on these images. However, the facades of high-rise buildings contain a large number of background image features, including wall textures, stains, holes, pipes, shadows, glass reflections, and the edges of equipment structures. Some background areas share similarities with beehives in color, outline, or local texture. Furthermore, beehives may be partially obscured by air conditioner units, window sills, pipes, or building components, causing them to appear incomplete in terms of color, with interrupted external outlines, or with only partially visible porous textures in the on-site images.

[0003] Existing image recognition methods, when applied to concealed honeycomb scenarios in high-rise buildings, typically employ color threshold matching, edge detection, contour matching, or single texture feature analysis to screen for suspected honeycomb areas from on-site images. However, when relying solely on color features for image recognition, wall stains, shadows, and building material colors can easily form color areas similar to honeycombs; when relying solely on contour features, building corners, equipment edges, and pipeline contours can easily cause interference, and the complete outline of a honeycomb is difficult to obtain directly when it is occluded; when relying solely on texture features for image recognition, the perforated texture features are prone to being missing or discontinuous in cases of long-distance shooting, changes in lighting, or partial occlusion of the honeycomb. Therefore, existing technologies struggle to comprehensively identify concealed honeycombs in complex backgrounds by simultaneously utilizing multiple image features of the honeycomb in terms of color, contour, and perforated texture. They also struggle to reliably obtain candidate honeycomb areas when partial occlusion exists, easily leading to problems such as misidentifying building backgrounds as honeycombs, missing parts of the honeycomb area, or having unstable boundaries of the identified honeycomb area, thus affecting the subsequent determination of the honeycomb center location. Summary of the Invention

[0004] This invention aims to solve the technical problem of insufficient accuracy and stability in identifying concealed honeycomb images under complex architectural backgrounds and partial occlusion conditions.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A high-rise concealed honeycomb intelligent spraying system includes: a parameter storage module for storing acquisition parameters, calibration parameters, confidence parameters, alignment parameters, permission parameters, and limit parameters.

[0007] The image acquisition and calibration module acquires on-site image data of the target area and generates preprocessed image data, image center coordinates, spray center pixel coordinates, and image timestamps based on the acquisition parameters and calibration parameters.

[0008] The concealed honeycomb detection module includes a detection region generation unit, which generates the target detection region in the preprocessed image data based on the image center coordinates, the jet center pixel coordinates, and a preset field of view offset range.

[0009] The color feature extraction unit performs color interval matching on the target detection area to obtain the color matching area.

[0010] The contour feature extraction unit performs edge extraction and connectivity analysis on the target detection region to obtain the contour connectivity region.

[0011] The texture feature extraction unit performs pore texture density feature calculation on the target detection area to obtain the texture density region.

[0012] The candidate region fusion unit fuses the color matching region, the contour connected region, and the texture density region to obtain the honeycomb candidate region data.

[0013] The target localization module generates the target honeycomb region, honeycomb center coordinates, and target validity status based on the candidate honeycomb region data and confidence parameters.

[0014] Furthermore, the image acquisition and calibration module includes: an image frame acquisition unit, which calls a preset acquisition period in the acquisition parameters to acquire the on-site image data, and generates an image timestamp for each frame of the on-site image data.

[0015] The image correction unit performs distortion correction on the on-site image data according to the lens calibration parameters in the calibration parameters to obtain corrected image data.

[0016] The center calibration unit generates the image center coordinates based on the pixel boundaries of the corrected image data, and generates the jet center pixel coordinates based on the jet center calibration parameters in the calibration parameters.

[0017] The image preprocessing unit performs brightness normalization, noise suppression, and scale normalization on the corrected image data to obtain the preprocessed image data.

[0018] Furthermore, the concealed honeycomb detection module also includes an occlusion relationship analysis unit, which performs local occlusion detection on the preprocessed image data of consecutive frames to obtain occlusion boundary data and visible area data.

[0019] The hidden feature extraction unit extracts pore texture points, arc contour segments, and building adjacency boundaries within the visible area data. The building adjacency boundaries are the set of boundary pixels adjacent to the corresponding area of ​​the honeycomb candidate area data and the building edge, wall corner, eaves edge, or equipment shell edge.

[0020] The boundary completion unit completes the honeycomb candidate region data based on the occlusion boundary data, the arc contour segment, and the building adjacent boundary, and sends the completed honeycomb candidate region data to the target positioning module.

[0021] Furthermore, the target localization module includes: a candidate region scoring unit, which receives the completed honeycomb candidate region data and generates candidate confidence based on the area, contour closure, texture density, building adjacency boundary, continuous frame boundary stability, and positional relationship with the pixel coordinates of the jet center in the honeycomb candidate region data.

[0022] The target region determination unit determines the candidate honeycomb regions whose confidence levels satisfy the confidence level parameters as the target honeycomb regions.

[0023] The center coordinate generation unit generates the center coordinates of the honeycomb based on the outer boundary, centroid position, pore texture point distribution, and pixel distribution centroid of the target honeycomb region.

[0024] The target stability determination unit generates the target valid state based on the honeycomb center coordinates in consecutive frames.

[0025] Furthermore, the high-rise concealed honeycomb intelligent injection system proposed in this invention further includes a state synchronization module configured to generate synchronized attitude data, synchronized length data, and synchronized channel status data based on the image timestamp. The state synchronization module includes: an attitude data acquisition unit that acquires pitch position detection signals and horizontal position detection signals from the injection end, generates pitch angle data based on the pitch position detection signals, and generates horizontal angle data based on the horizontal position detection signals.

[0026] The length data acquisition unit acquires the displacement detection signal of the support adjustment object connected to the injection end, and generates support length data based on the displacement detection signal.

[0027] The channel status acquisition unit acquires the pressure detection signal and valve status signal of the medium channel object connected to the injection end, and generates medium channel status data.

[0028] The timing synchronization unit synchronizes the pitch angle data, horizontal angle data, support length data, and medium channel status data according to the image timestamp to obtain the synchronized attitude data, synchronized length data, and synchronized channel status data.

[0029] Furthermore, the high-rise concealed honeycomb intelligent jetting system proposed in this invention further includes a visual pointing mapping module configured to generate jet pointing deviation data based on the target's effective state, honeycomb center coordinates, jetting center pixel coordinates, synchronization attitude data, and synchronization length data. The visual pointing mapping module includes an effective state determination unit that receives the target's effective state and allows the generation of the jet pointing deviation data when the target's effective state is valid.

[0030] The image deviation generation unit generates image deviation data based on the position of the honeycomb center coordinates relative to the pixel coordinates of the jet center when the generation of the jet pointing deviation data is allowed.

[0031] The injection coordinate establishment unit establishes an injection coordinate system based on the injection centerline according to the calibration parameters, the synchronization attitude data, and the synchronization length data.

[0032] The pointing deviation generation unit maps the honeycomb center coordinates to the jet coordinate system to obtain the target jet coordinates, and generates the jet pointing deviation data based on the image deviation data and the target jet coordinates.

[0033] Furthermore, the high-rise concealed honeycomb intelligent injection system proposed in this invention further includes a visual closed-loop control module configured to generate attitude adjustment amounts based on the injection pointing deviation data and limit parameters, and to control the pitch adjustment and horizontal deflection of the injection end. The visual closed-loop control module includes an adjustment amount generation unit that reads the limit parameters from the parameter storage module and generates attitude adjustment amounts including pitch and horizontal adjustment amounts based on the injection pointing deviation data and the limit parameters.

[0034] The bidirectional adjustment unit controls the spray end to rotate in the pitch direction according to the pitch adjustment amount, and controls the spray end to rotate in the horizontal direction according to the horizontal adjustment amount.

[0035] The deviation update unit triggers the image acquisition and calibration module to acquire the adjusted on-site image data after the bidirectional adjustment unit completes its rotation, and updates the injection pointing deviation data according to the adjusted honeycomb center coordinates.

[0036] Furthermore, the high-rise concealed honeycomb intelligent spraying system proposed in this invention also includes a spraying permission output module, which includes a visual alignment determination unit that generates a visual alignment state based on the adjusted spraying direction deviation data and the alignment parameters.

[0037] The spray area generation unit generates a spray-allowed image area based on the target honeycomb area and a spray center image area based on the spray center pixel coordinates.

[0038] The region overlap determination unit performs overlap determination on the spray allowable image region and the spray center image region to obtain the region overlap state.

[0039] The permission instruction generation unit generates a media spraying permission instruction based on the visual alignment state, the area overlap state, the target validity state, the synchronization channel status data, and the permission parameters. The media spraying permission instruction includes a spraying start permission state and a spraying end permission state.

[0040] Furthermore, the high-rise concealed honeycomb intelligent spraying system proposed in this invention also includes a post-spraying image verification module, which includes a post-spraying image acquisition unit that triggers the image acquisition and calibration module to acquire post-spraying on-site image data after the spraying end permission state is generated.

[0041] The post-spray detection unit enables the concealed honeycomb detection module to generate post-spray honeycomb candidate area data based on the post-spray field image data.

[0042] The residual area generation unit enables the target positioning module to generate a target honeycomb region after spraying based on the candidate honeycomb region data after spraying, and performs regional difference analysis between the target honeycomb region after spraying and the target honeycomb region before spraying to obtain residual area data.

[0043] The residual target generation unit generates residual center coordinates based on the residual area data when the residual area data meets the preset residual conditions in the permission parameters.

[0044] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0045] (1): This invention acquires, calibrates, and preprocesses on-site image data of the target area to obtain the image center coordinates, the pixel coordinates of the spray center, and preprocessed image data, thus providing a clear image coordinate basis for the honeycomb identification process. Furthermore, by using color features, contour features, pore texture features, and occlusion boundary features to identify hidden honeycombs on the facades of high-rise buildings, window sills, eaves, or the back of equipment, it can reduce the problem of unstable target positioning caused by complex backgrounds, local occlusion, or changes in lighting.

[0046] (2): This invention generates the target honeycomb region, honeycomb center coordinates, and target effective state based on the honeycomb candidate region data, and combines the image timestamp with the synchronous jet end attitude data, support length data, and medium channel state data, so that the honeycomb position identified in the image can correspond to the current attitude and operating state of the jet end. Thus, the system can generate jet pointing deviation data based on the deviation between the honeycomb center coordinates and the jet center pixel coordinates, and use this data for jet end attitude adjustment, so that the image recognition result can directly participate in the jet direction control.

[0047] (3): The present invention also generates attitude adjustment amount based on the spray direction deviation data through a visual closed-loop control module, and controls the spray end to perform pitch adjustment and horizontal deflection, so that the spray end can perform directional correction around the identified target honeycomb area. This effect corresponds to the mechanical basis of the spray head being able to adjust up and down, deflect left and right, and the medium channel passing through the support structure for transportation. It can combine mechanical adjustment capability with image recognition results to realize directional spray control for high-rise concealed honeycomb.

[0048] (4): In the injection permission output stage, the present invention generates a medium injection permission command by comprehensively adjusting the injection pointing deviation data, target validity status, area overlap status, and synchronization channel status data, and re-acquires the post-injection field image data for residual area analysis after injection. Thus, the injection process can form a continuous processing flow of "image identification before injection, pointing permission during injection, and image verification after injection", reducing the occurrence of injection when the target is misaligned, the target is invalid, or the channel status is abnormal. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a connection diagram of the high-rise concealed honeycomb intelligent spraying system proposed in this invention;

[0051] Figure 2 This is a flowchart of the image acquisition and calibration module proposed in this invention;

[0052] Figure 3 This is a flowchart of the post-spraying image verification and residual target generation proposed in this invention;

[0053] Figure 4 This is a three-dimensional schematic diagram of the high-rise concealed honeycomb intelligent spray gun proposed in this invention;

[0054] Figure 5 This invention is proposed. Figure 4 An enlarged diagram of A in the diagram.

[0055] The components include: 1. Spray gun head; 2. Adjustment pipe; 3. First connecting sleeve; 4. Second connecting sleeve; 5. Spray end visual perception module; 6. Telescopic rod; 7. Attitude drive control module; 8. Fixed rod; 9. Adjustment block; 10. Main control interaction module; 11. Delivery pipe; 12. Adjustment knob; and 13. Spray end medium control module. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] The high-rise concealed honeycomb intelligent spraying system of the present invention will be described below with reference to the accompanying drawings.

[0058] like Figure 4 and Figure 5 As shown, the high-rise concealed honeycomb intelligent spraying system proposed in this invention includes a spray gun, wherein the spray gun includes a spray gun head 1, an adjusting pipe 2, a first connecting sleeve 3, a second connecting sleeve 4, a spray end visual perception module 5, a telescopic rod 6, an attitude drive control module 7, a fixed rod 8, an adjusting block 9, a main control interaction module 10, a delivery pipe 11, an adjusting knob 12, and a spray end medium control module 13.

[0059] One side of the spray gun head 1 is connected to the regulating pipe 2. The regulating pipe 2 has a cavity inside, and a delivery pipe 11 is located inside the cavity. One end of the delivery pipe 11 extends into the spray gun head 1, and the other end passes through the cavity inside the regulating pipe 2. A first connecting sleeve 3 is connected to the end of the regulating pipe 2. The outside of the first connecting sleeve 3 is movably connected to the second connecting sleeve 4 through an adjusting knob 12. That is, the regulating pipe 2 can be adjusted up and down relative to the second connecting sleeve 4 around the adjusting knob 12. A spray end visual perception module 5 is set on the outside of the adjusting knob 12. The spray end visual perception module 5 is used to acquire on-site image data of the target area and generate the target honeycomb area, the honeycomb center coordinates, and the target valid status. The attitude drive control module 7 drives the adjusting knob 12 to move according to the attitude adjustment amount generated by the visual closed-loop control module, so that the regulating pipe 2 drives the spray gun head 1 to rotate to the specified position. A telescopic rod 6 is rotatably connected to the outside of the second connecting sleeve 4. The second connecting sleeve 4 can rotate around the axis of the telescopic rod 6.

[0060] The attitude drive control module 7 integrates a pitch drive motor, a horizontal drive motor, and corresponding motor drive circuits. The output shaft of the pitch drive motor is linked to the pitch rotation shaft (adjustment knob 12) between the first connecting sleeve 3 and the second connecting sleeve 4 via a pitch transmission mechanism (such as a gear set, worm gear, or flexible shaft).

[0061] A rotating bearing is provided between the sleeve end of the second connecting sleeve 4 and the top end of the telescopic rod 6. The output shaft of the horizontal drive motor is connected to the outer wall of the gear ring of the second connecting sleeve 4 through a horizontal transmission gear. When the attitude drive control module 7 receives the horizontal adjustment amount in the attitude adjustment amount, it drives the horizontal drive motor to rotate. The horizontal transmission gear drives the second connecting sleeve 4 to rotate left and right around the axis of the telescopic rod 6, thereby driving the first connecting sleeve 3, the adjusting pipe 2 and the spray gun head 1 to rotate together around the axis of the telescopic rod 6, thus completing the left and right horizontal deflection adjustment.

[0062] A fixed rod 8 is connected to the outside of the telescopic rod 6, and the telescopic rod 6 extends into the interior of the fixed rod 8. An adjusting block 9 is provided at the top of the fixed rod 8. By adjusting the tightness of the adjusting block 9, the telescopic rod 6 can be controlled to extend or retract. A main control interaction module 10 is provided at the top of the fixed rod 8. The delivery pipe 11 passes through the interior of the adjusting pipe 2 and through the interior of the telescopic rod 6 and the fixed rod 8. The delivery pipe 11 is connected to a gas tank, which can be filled with combustible gas or insecticide.

[0063] The spray end visual perception module 5 includes an image acquisition and calibration module, a concealed honeycomb detection module, a target positioning module, and a post-spray image verification module.

[0064] The attitude drive control module 7 includes a vision closed-loop control module.

[0065] The main control interaction module 10 includes a parameter storage module, a status synchronization module, and a visual pointing mapping module.

[0066] The injection end medium control module 13 includes an injection permission output module.

[0067] Furthermore, a visual feedback control relationship is established between the spray end visual perception module 5 and the attitude drive control module 7. The spray end visual perception module 5 is used to acquire on-site image data of the target area, and sequentially generates preprocessed image data, honeycomb candidate area data, target honeycomb area, honeycomb center coordinates, and target valid status through the image acquisition and calibration module, the concealed honeycomb detection module, and the target positioning module; the visual pointing mapping module generates spray pointing deviation data based on the honeycomb center coordinates, the spray center pixel coordinates, and the synchronous attitude data; the visual closed-loop control module in the attitude drive control module 7 generates attitude adjustment amount based on the spray pointing deviation data, and drives the spray end (spray gun head 1) to perform pitch adjustment and horizontal deflection.

[0068] After the spray nozzle (spray head 1) completes an attitude adjustment, the visual closed-loop control module sends a re-acquisition trigger signal to the spray nozzle visual perception module 5, causing the image acquisition and calibration module to reacquire the adjusted on-site image data. The concealed honeycomb detection module regenerates adjusted honeycomb candidate area data based on the adjusted on-site image data, the target positioning module regenerates adjusted honeycomb center coordinates based on the adjusted honeycomb candidate area data, and the visual pointing mapping module updates the spray pointing deviation data based on the adjusted honeycomb center coordinates. Thus, a closed-loop control relationship of "image acquisition—honeycomb detection—target positioning—attitude adjustment—re-image acquisition" is formed between the spray nozzle visual perception module 5 and the visual closed-loop control module.

[0069] like Figure 1 As shown, the present invention proposes a high-rise concealed honeycomb intelligent spraying system, comprising the following modules:

[0070] The parameter storage module stores the acquisition parameters, calibration parameters, confidence parameters, alignment parameters, permission parameters, and limit parameters.

[0071] In practical applications, the parameter storage module stores the basic parameters needed during system operation, providing a unified basis for subsequent image acquisition, honeycomb recognition, target localization, attitude adjustment, and spray permission determination. Acquisition parameters may include image acquisition period, image resolution, and exposure parameters; calibration parameters may include lens calibration parameters and spray center calibration parameters; confidence parameters are used to determine whether a candidate honeycomb region can be used as a target honeycomb region; alignment parameters are used to determine whether the spray nozzle is aligned with the target honeycomb region; permission parameters are used to determine whether media spraying is permitted; and limit parameters are used to limit the pitch adjustment range and horizontal deflection range of the spray nozzle.

[0072] For example, in the scenario of honeycomb processing behind the outdoor unit of an air conditioner in a high-rise building, the parameter storage module can store the following parameters: image acquisition period of 100ms, image resolution of 1280×720 pixels, confidence parameter of 0.70, alignment parameter of pixel deviation between the pixel coordinates of the injection center and the honeycomb center of no more than 25 pixels, permissible pressure range of medium channel of 0.15MPa to 0.80MPa, pitch limit range of -45° to 60°, and horizontal deflection limit range of -90° to 90°.

[0073] The image acquisition and calibration module acquires on-site image data of the target area and generates preprocessed image data, image center coordinates, spray center pixel coordinates, and image timestamps based on the acquisition and calibration parameters.

[0074] In practical applications, such as Figure 2 As shown, the image acquisition and calibration module is used to acquire on-site image data of the target area and convert the raw images into data that can be used for subsequent honeycomb recognition and jet pointing calculation. The on-site image data is the raw image acquired by the camera; the preprocessed image data is the image obtained after distortion correction, brightness normalization, noise suppression, and scale normalization of the on-site image data; the image center coordinates are the center position of the image pixel plane; the jet center pixel coordinates are the projection position of the current jet centerline of the jet end in the image; the image timestamp is used to record the acquisition time of each frame of on-site image data.

[0075] For example, in a 1280×720 pixel frame of on-site image data, the image acquisition and calibration module can generate image center coordinates of (640, 360). If, after injection center calibration, the corresponding position of the injection centerline at the injection end in the image is (620, 350), then the pixel coordinates of the injection center are (620, 350). Simultaneously, this image frame is timestamped, for example, 14:30:15.120, for subsequent synchronization with injection end attitude data, support length data, and media channel status data.

[0076] The image frame acquisition unit is used to acquire on-site image data at fixed intervals. The preset acquisition interval can be 33ms to 200ms, corresponding to approximately 5 frames / second to 30 frames / second. The image timestamp can be recorded at the millisecond level to mark the acquisition time of each image frame.

[0077] The image correction unit uses lens calibration parameters to correct distortion in the on-site image data. Lens calibration parameters can include focal length, principal point coordinates, and radial distortion parameters. Radial distortion parameters can be obtained using checkerboard calibration, dot array calibration, or factory calibration. The purpose of distortion correction is to correct positional deviations caused by stretching or bending at the lens edges, ensuring a calculable consistency between the honeycomb center coordinates, the jet center pixel coordinates, and subsequent image deviation data.

[0078] The center calibration unit calculates the image center coordinates based on the pixel boundaries of the corrected image data. For example, when the image width is... Image height is At that time, the coordinates of the image center can be In the formula, Image width, The height is the image height. The pixel coordinates of the spray center can be obtained through spray end calibration. For example, in a safe calibration environment, the spray end is oriented towards the center of the calibration board, the pixel position of the spray center line in the image is recorded, and saved as the spray center calibration parameter.

[0079] The image preprocessing unit performs brightness normalization, noise suppression, and scale normalization on the corrected image data. Brightness normalization reduces brightness differences caused by backlighting, shadows, or nighttime illumination; noise suppression can employ median filtering, Gaussian filtering, or bilateral filtering; scale normalization unifies images of different resolutions to the size required for model or algorithm processing. The preprocessed image data obtained through these processes can be used as a unified input for subsequent honeycomb image detection.

[0080] The concealed honeycomb detection module determines the target detection area based on the image center coordinates and the pixel coordinates of the spray center, performs honeycomb image detection on the target detection area, and obtains honeycomb candidate area data.

[0081] In practical applications, the concealed honeycomb detection module is used to identify suspected honeycomb regions in preprocessed image data and generate candidate honeycomb regions. This module does not directly treat the entire image as the object of recognition; instead, it determines the target detection region based on the image center coordinates and the jet center pixel coordinates, and then performs color feature, contour feature, and pore texture feature analysis within the target detection region. The candidate honeycomb region data can include the candidate region's boundary, area, centroid position, contour closure, texture density, and its positional relationship with the jet center pixel coordinates.

[0082] The target detection region is a local image region generated around the pixel coordinates of the spray center and combined with the image center coordinates. The preset field of view offset range can be 20% to 70% of the image width and 20% to 70% of the image height. Its purpose is to retain the area that the spray end may currently cover and reduce the amount of processing for irrelevant building backgrounds.

[0083] The color matching region is obtained by matching color ranges. A honeycomb pattern typically appears yellowish-brown, grayish-brown, or dark brown in visible light images; therefore, hue, saturation, and brightness ranges can be set in the HSV color space, for example, a hue range of 10° to 50°, a saturation range of 20% to 80%, and a brightness range of 15% to 90%. These ranges can be adjusted according to the actual camera and ambient lighting conditions as described in the embodiments of the specification.

[0084] The contour connected region is a candidate contour region obtained through edge extraction and connected component analysis. Edge extraction can employ Canny edge detection, the Sobel operator, or the Laplacian operator; connected component analysis is used to merge adjacent edge pixels into the same candidate region. The texture density region is a region calculated using pore texture density features, which can be expressed as:

[0085]

[0086] In the formula, The density of the porous texture, This represents the number of points identified as porous texture points within the target detection area. The area of ​​the target detection region.

[0087] Furthermore, to ensure that the density features of the pore texture have a repeatable basis for calculation, the texture feature extraction unit converts the target detection region into a grayscale image, performs edge extraction and connected component search on the grayscale image, and selects local regions enclosed by closed edges or approximately closed edges as candidate regions for the pore texture. For the first... Each candidate region with a porous texture has its area calculated. Perimeter of the boundary Average gray value within the region and the average gray value within a preset width neighborhood surrounding the candidate region. .

[0088] The circularity of the candidate regions for the pore texture is calculated based on the area and the perimeter of the boundary. ,in:

[0089]

[0090] When the candidate region of the pore texture simultaneously satisfies , as well as When the candidate region of the pore-like texture is determined, it is identified as a pore-like texture point; where, and These are the lower and upper limits of the area of ​​the aperture region, determined based on the image scale. The circularity threshold. It is the threshold for the grayscale difference between the inside and the surrounding neighborhood of the perforated region.

[0091] Count the number of pore-like texture points within the target detection area that meet the above conditions. And based on the area of ​​the target detection region Calculate the density of the porous texture :

[0092]

[0093] When the target detection region has undergone scale normalization processing, the , , and Preset values ​​under the corresponding normalized scale are used to make the density of pore texture comparable under different acquisition distances or different image resolutions.

[0094] The candidate region fusion unit merges color-matching regions, contour-connected regions, and texture density regions by overlapping or adjacency. Fusion methods can include intersection, union followed by filtering, or weighted scoring fusion. The resulting honeycomb candidate region data contains not only color features but also contour and texture features, which can be used in subsequent target localization modules to calculate candidate confidence.

[0095] For example, the system uses the center pixel coordinates (620, 350) of the spray center as the center and extracts a target detection area with a width of 600 pixels and a height of 420 pixels from the preprocessed image data. The color feature extraction unit extracts the yellowish-brown area, the contour feature extraction unit extracts the arc-shaped connected area on the back edge of the outer machine, and the texture feature extraction unit identifies the pore-shaped texture points. Finally, the two honeycomb candidate regions are fused together.

[0096] A concealed honeycomb refers to a honeycomb structure that is partially obscured by air conditioner units, pipes, window sills, tree branches, or building edges, resulting in only a portion of the honeycomb being visible in the on-site image data. The occlusion relationship analysis unit uses continuous frame analysis to determine which image regions are stable occlusion areas and which are visible honeycomb areas. Occlusion boundary data can include the set of pixels at the edge of the occluding object, the occlusion direction, and the occlusion ratio; visible area data represents the unoccluded pixel areas that contain the honeycomb texture or outline.

[0097] Building adjacency boundaries are now defined as a set of boundary pixels, a definition that gives the term a clear meaning. Building adjacency boundaries are used to determine whether a honeycomb is attached to a wall corner, eaves edge, or equipment casing edge. Since high-rise honeycombs often form at building edges or on the back of equipment, using building adjacency boundaries can reduce the possibility of misidentifying ordinary stains, shadows, or equipment parts as honeycomb candidate areas.

[0098] The occlusion analysis unit performs position mapping on edge pixels in consecutive frames, and will then perform position mapping on consecutive frames. The positional change within the frame does not exceed a preset boundary offset threshold. Furthermore, continuous edge segments located outside the honeycomb candidate region are defined as stable external boundaries. For candidate contours that are interrupted within the honeycomb candidate region, when the two interrupted endpoints of the candidate contour are located on either side of the stable external boundary or the shortest distance between both endpoints and the stable external boundary is not greater than [value missing], [the boundary is defined as follows]. When the corresponding stable external boundary is defined as the occlusion boundary, the area separated by the occlusion boundary and not directly imaged is defined as the area to be completed.

[0099] The boundary completion unit first obtains the visible arc-shaped contour segments located on both sides of the occlusion boundary, and then obtains the contour pixels on the arc-shaped contour segments. Using the contour pixels as fitting input, the following is minimized:

[0100]

[0101] Obtain the center of the fitted circular arc and radius .

[0102] When the fitting error When the error is not greater than the preset fitting error threshold, the fitting arc is extended from the first contour interruption endpoint along the contour continuity direction to the occluded area, and the extension is stopped when the extended arc reaches the second contour interruption endpoint. The extended arc is used as the completion boundary of the occluded part.

[0103] When only one contour interruption endpoint is detected, the contour tangent direction is calculated based on the continuous contour pixels adjacent to the interruption endpoint, and the extended arc is extended in accordance with the tangent direction. When the extended arc intersects with the adjacent boundary of the building, the extension length reaches the preset maximum completion length, or the area of ​​the completed area reaches the preset ratio upper limit of the area of ​​the original visible candidate area, the extension stops to avoid over-completion caused by insufficient visible contour on one side.

[0104] The boundary completion unit combines the visible outline, the completed boundary, and the adjacent building boundary connected to it to form a closed candidate boundary, and uses the area enclosed by the closed candidate boundary as the completed honeycomb candidate area; further, it obtains the area of ​​the completed area, the position of the centroid after completion, and the pixel set of the completed boundary, and sends the above data as the completed honeycomb candidate area data to the target positioning module.

[0105] For example, the outdoor unit of an air conditioner obscures the lower left part of the honeycomb structure. The occlusion relationship analysis unit detects that the edge of the outdoor unit casing forms a stable occlusion boundary. The hidden feature extraction unit extracts pore-like texture points and arc-shaped contour segments within the visible area. The boundary completion unit completes the occluded side boundary along the arc-shaped contour segments, thus obtaining the completed honeycomb candidate region data.

[0106] The target localization module generates the target honeycomb region, honeycomb center coordinates, and target validity status based on the candidate honeycomb region data and confidence parameters.

[0107] In practical applications, the target localization module is used to determine the final target honeycomb region to be sprayed from the honeycomb candidate region data, and to generate the honeycomb center coordinates and target validity status. The target honeycomb region is a honeycomb candidate region that meets the confidence parameters; the honeycomb center coordinates are the center position of the target honeycomb region in the image, which can be determined based on the outer boundary, centroid position, pore texture point distribution, and pixel distribution centroid of the target honeycomb region; the target validity status is used to indicate whether the current target recognition result is stable and whether it can participate in the subsequent spray direction calculation.

[0108] Candidate confidence is used to characterize the degree of agreement between candidate honeycomb regions and preset honeycomb image features. For the first... Each of the candidate honeycomb regions is scored for its area. Contour closure score Texture density score Building Adjacency Boundary Scoring Continuous frame boundary stability score and positional relationship score .

[0109] Area scoring is based on the area of ​​the candidate honeycomb region. With respect to the preset effective area range The relationship between them is determined; when When the area is within the effective area range, the area score is increased. The area score is reduced when the area deviates from the defined effective area range.

[0110] Contour closure score is based on the actual continuous boundary length of the candidate region. Its complete outer contour estimated length Sure:

[0111]

[0112] Texture density scoring is based on pore texture density. Compared with the preset honeycomb texture density benchmark value Sure:

[0113]

[0114] Building adjacency boundary scoring is based on the boundary length between the candidate honeycomb region and the adjacent building adjacency boundary. and the total boundary length of the candidate region Sure:

[0115]

[0116] The continuous frame boundary stability score is based on the centroid of the current frame's honeycomb candidate region. Centroid of the honeycomb candidate region in the previous corresponding frame The pixel displacement between them is determined:

[0117]

[0118] in, The maximum allowable reference displacement for consecutive frames.

[0119] Positional relationship scoring is based on the distance between the centroid of the honeycomb candidate region and the pixel coordinates of the jet center. Sure:

[0120]

[0121] in, This is the preset maximum reference distance within the target detection area.

[0122] Candidate confidence levels are generated based on the above scores. :

[0123]

[0124] in, , , , , and These are the corresponding scoring weights, and they satisfy the following:

[0125] .

[0126] Candidate confidence The value range is from 0 to 1. When When the confidence level is not lower than the preset confidence threshold corresponding to the confidence level parameter, the corresponding candidate honeycomb region is determined as the target honeycomb region.

[0127] The honeycomb center coordinates are not simply taken from the circumscribed boundary center, but are generated by combining the centroid position, the distribution of pore texture points, and the centroid of pixel distribution. This allows the honeycomb center coordinates to be closer to the sprayable coverage center when the honeycomb is partially occluded. The target's valid state can be generated based on whether the honeycomb center coordinates are stable over 3 to 10 consecutive frames and whether the candidate confidence score meets the confidence parameter; for example, a valid state is generated when the honeycomb center coordinates change by no more than 20 pixels over 5 consecutive frames and the candidate confidence score is not lower than 0.70.

[0128] For example, if the candidate confidence scores of two honeycomb candidate regions are 0.78 and 0.42 respectively, and the confidence score parameter is set to 0.70, then the target region determination unit will determine the honeycomb candidate region with a candidate confidence score of 0.78 as the target honeycomb region. The center coordinate generation unit combines the completion of the outer boundary and the distribution of pore texture points to generate the honeycomb center coordinates, for example (690, 380).

[0129] The status synchronization module generates synchronization posture data, synchronization length data, and synchronization channel status data based on the image timestamp.

[0130] In practical applications, the state synchronization module is used to time-match image data with the current state of the injection end based on the image timestamp. Synchronized attitude data can include the pitch and horizontal angle data of the injection end; synchronized length data can include the extension length data of the retractable support assembly; synchronized channel status data can include medium channel pressure data, valve status data, and injection end medium status data. Through the state synchronization module, it is possible to avoid using image data from one moment to match the injection end attitude at another moment, reducing time mismatch problems in injection pointing calculations.

[0131] Pitch position detection signals can be acquired by angle sensors, encoders, potentiometer-type position sensors, or Hall effect angle sensors; horizontal position detection signals can be acquired by similar sensors. Pitch angle data represents the vertical angle of the injection end relative to the support direction, and horizontal angle data represents the horizontal yaw angle of the injection end relative to the support direction. Pitch angle data and horizontal angle data together constitute the basic attitude data of the injection end.

[0132] The adjustable support is a structure connected to the injection end, used to change the distance between the injection end and the operating position. It can correspond to a telescopic rod or an adjustable support structure. Displacement detection signals can be acquired by a wire displacement sensor, a magnetostrictive displacement sensor, an coded slide rail, or a segmented locking position switch. The support length data can range from 0.5m to 8m and can be adjusted according to the actual rod structure.

[0133] The media channel is a channel structure connected to the injection end and used to transport the injected media. Pressure detection signals can be acquired by pressure sensors, and valve status signals can be acquired by valve opening / closing feedback switches, valve drive current, or solenoid valve status signals. Media channel status data can include pressure values, valve opening / closing status, and abnormal status flags.

[0134] The timing synchronization unit synchronizes the above data based on the image timestamp. For example, when the image timestamp of a frame of on-site image data is... At that time, the system selects the sampling time and Pitch angle data, horizontal angle data, support length data, and media channel status data with a difference of no more than 20ms to 100ms are used as synchronization data for the corresponding image frames.

[0135] For example, the image timestamp is 18:01:15.120. The attitude data acquisition unit obtains pitch angle data of 12° and horizontal angle data of -8°; the length data acquisition unit obtains support length data of 4.2m; the channel status acquisition unit obtains pressure value of 0.35MPa and valve closed. The timing synchronization unit binds these data with the field image data of this frame to obtain synchronized attitude data, synchronized length data, and synchronized channel status data.

[0136] The visual pointing mapping module generates jet pointing deviation data based on the target's effective state, the honeycomb center coordinates, the jet center pixel coordinates, the synchronization attitude data, and the synchronization length data.

[0137] In practical applications, the visual pointing mapping module is used to convert the honeycomb position in the image into the directional deviation that the jetting end needs to adjust. This module receives the target's valid state, honeycomb center coordinates, jetting center pixel coordinates, synchronization attitude data, and synchronization length data. When the target's valid state is active, it calculates the deviation of the honeycomb center coordinates relative to the jetting center pixel coordinates, and combines this with the jetting end's current pitch angle, horizontal angle, and support length to generate jetting pointing deviation data. The jetting pointing deviation data can include image deviation, pitch deviation, and horizontal deviation.

[0138] The effective state determination unit first determines the effective state of the target. Only when the effective state of the target is effective will subsequent injection pointing deviation data be generated. If the effective state of the target is invalid, no new injection pointing deviation data will be generated, and the system can either keep the medium injection permission command closed or wait for the detection result of the next frame.

[0139] Image offset data represents the offset between the honeycomb center coordinates and the jet center pixel coordinates in the image coordinate system. It can include lateral pixel offset and vertical pixel offset, as shown in the formula:

[0140] ,

[0141] In the formula, This refers to the lateral pixel deviation. For vertical pixel deviation, and The coordinates of the center of the honeycomb and These are the pixel coordinates of the injection center. In this article... and It is used only to represent the two components of image deviation data.

[0142] The jet pointing mapping is calculated using the correspondence between image pixel deviation and jet direction angle deviation, rather than directly calculating the spatial distance between the honeycomb and the jet end based on the monocular image.

[0143] Let the coordinates of the honeycomb center be... The pixel coordinates of the spray center are Then the horizontal pixel deviation and vertical pixel deviation They are respectively:

[0144]

[0145]

[0146] Lens calibration parameters in the calibration parameters include the equivalent focal length in the horizontal direction of the image. Equivalent focal length in the vertical direction of the image The image deviation generation unit calculates the horizontal angular deviation based on the horizontal pixel deviation and the vertical pixel deviation, respectively. and pitch direction angle deviation :

[0147]

[0148]

[0149] in, The positive or negative sign indicates the direction the injection end needs to be adjusted to, either left or right. The sign indicates the direction in which the injection end needs to be adjusted upwards or downwards. The specific sign relationship is determined based on the calibration results of the image coordinate system and the mechanical coordinate system of the injection end.

[0150] The jet coordinate establishment unit uses the jet center position at the jet end as the origin, the current jet centerline as the forward axis, and the current horizontal angle in the synchronized attitude data. and current pitch angle Determine the current attitude of the injection end. The synchronization length data is used to determine the current position of the injection end relative to the support reference position, and is not used to deduce the cellular distance from the monocular image.

[0151] Target horizontal pointing angle and target pitch angle They are respectively:

[0152]

[0153]

[0154] Therefore, the injection pointing deviation data should include at least the horizontal angular deviation. and pitch direction angle deviation The vision closed-loop control module generates corresponding attitude adjustment amounts based on the horizontal and pitch angle deviations, and reacquires the field image after completing one attitude adjustment, then recalculates. , , and Continue until the jet pointing deviation meets the alignment parameters.

[0155] For example, when the center coordinates of the honeycomb are (690, 380) and the pixel coordinates of the jet center are (620, 350), the following is obtained: Pixels Pixels. The image acquisition and calibration module reads the pre-calibrated pixels. and Then, based on the above angle mapping relationship, the horizontal direction angle deviation and the pitch direction angle deviation are calculated and sent to the vision closed-loop control module, instead of directly generating the target distance that cannot be uniquely determined by the monocular image based on the two-dimensional honeycomb center coordinates.

[0156] The vision closed-loop control module generates attitude adjustment amounts based on injection direction deviation data and limit parameters, and controls the pitch adjustment and horizontal deflection of the injection end.

[0157] In practical applications, the vision-based closed-loop control module generates attitude adjustment amounts based on the spray pointing deviation data and controls the spray end (spray nozzle position 1) to perform pitch and horizontal adjustments. The attitude adjustment amounts include pitch and horizontal adjustments; the pitch adjustment controls the vertical rotation of the spray end, while the horizontal adjustment controls its horizontal rotation. After each attitude adjustment at the spray end, the module can also trigger the image acquisition and calibration module to reacquire the adjusted field image data and update the spray pointing deviation data based on the adjusted honeycomb center coordinates, thus forming a closed-loop control.

[0158] The adjustment generation unit converts the injection pointing deviation data into attitude adjustment values. These attitude adjustment values ​​include pitch and horizontal adjustment values, corresponding to the vertical and horizontal adjustments of the injection end, respectively. Limit parameters are used to limit the adjustment range, preventing the output from exceeding the mechanically permissible range. Limit parameters may include a single pitch adjustment range of -10° to 10°, a single horizontal adjustment range of -15° to 15°, a cumulative pitch range of -45° to 60°, and a cumulative horizontal range of -90° to 90°.

[0159] The bidirectional adjustment unit controls the jetting end's movement according to the attitude adjustment amount. If the jetting direction deviation data indicates that the honeycomb center is to the right of the jetting center pixel coordinate, the bidirectional adjustment unit can output a horizontal adjustment amount; if the jetting direction deviation data indicates that the honeycomb center is below or above the jetting center pixel coordinate, the bidirectional adjustment unit can output a pitch adjustment amount. The adjustment amount can be obtained through proportional control, using the following formula:

[0160]

[0161]

[0162] In the formula, For pitch adjustment, This is the horizontal adjustment amount. This is the pitch scaling factor. This is the horizontal proportionality coefficient. For vertical pixel deviation, This represents the horizontal pixel deviation. and It can be taken as 0.005° / pixel to 0.05° / pixel, and determined in conjunction with the lens field of view calibration.

[0163] After the injection end rotates, the deviation update unit re-triggers image acquisition to obtain adjusted on-site image data and re-acquires the adjusted honeycomb center coordinates. This process feeds the adjusted visual results back to the system, gradually reducing the injection pointing deviation data and forming a visual closed loop.

[0164] For example, the system generates horizontal and vertical adjustment amounts based on a lateral pixel deviation of 70 pixels and a vertical pixel deviation of 30 pixels, such as a horizontal adjustment of 2.1° and a vertical adjustment of 0.9°. After the injection end completes the horizontal yaw and vertical adjustment, the deviation update unit triggers the re-acquisition of the adjusted on-site image data and updates the injection pointing deviation data.

[0165] The injection permission output module generates a medium injection permission command based on the adjusted injection direction deviation data, synchronization channel status data, alignment parameters, and permission parameters.

[0166] In practical applications, the injection permission output module is used to determine whether the current media injection conditions are met and to generate a media injection permission command. This module does not solely rely on whether the injection end is close to the honeycomb structure to determine whether to inject; instead, it comprehensively considers the adjusted injection pointing deviation data, synchronization channel status data, alignment parameters, and permission parameters. The media injection permission command can include a injection start permission state and a injection end permission state. A injection start permission state is generated when the target is valid, the injection pointing deviation meets the alignment parameters, the media channel pressure meets the permission parameters, and the valve status is normal. A injection end permission state is generated when injection is completed, the target fails, the deviation exceeds the range, or the channel status is abnormal.

[0167] The visual alignment status indicates whether the visual pointing of the jet end meets the alignment requirements. Alignment parameters may include maximum pixel deviation, maximum angle deviation, or continuous frame stabilization conditions; for example, the maximum pixel deviation can be 10 to 30 pixels, the maximum angle deviation can be 1° to 3°, and the continuous frame stabilization condition can be that the above deviation range is met for 3 to 8 consecutive frames.

[0168] The permissible spray image region is the allowable spray range generated based on the target honeycomb region, which can be an area extended outward by 5% to 30% from the outer boundary of the target honeycomb region. The spray center image region is the spray coverage area generated based on the spray center pixel coordinates, which can be a circular or elliptical region with a radius of 10 to 50 pixels centered at the spray center pixel coordinates. The region overlap status can be determined by the region overlap rate, using the formula:

[0169]

[0170] In the formula, For regional overlap rate, The overlapping area between the spray center image region and the spray allowable image region. This represents the area of ​​the image region at the center of the spray. When the region overlap rate is between 0.60 and 0.95, a region overlap state that meets the preset overlap conditions can be generated.

[0171] The permission command generation unit generates a media injection permission command by integrating visual alignment status, area overlap status, target validity status, synchronization channel status data, and permission parameters. Permission parameters may include a target validity status of 0.60 to 0.95, an area overlap rate of 0.15 MPa to 0.80 MPa, and valve feedback being controllable. The injection start permission status allows the media channel object to open, while the injection end permission status indicates that a preset injection duration has been reached, the target validity status has become invalid, the area overlap status condition has not been met, or the synchronization channel status data is abnormal.

[0172] For example: after adjustment, the corresponding pixel deviation of the injection pointing deviation data is 18 pixels, and the alignment parameter is set to no more than 25 pixels; the overlap rate between the injection allowable image area and the injection center image area is 0.82; the synchronous channel status data shows that the pressure is 0.35 MPa and the valve is controllable. The permission command generation unit generates a injection start permission state accordingly; when the set injection duration is reached or a decrease in the area overlap rate is detected, a injection end permission state is generated.

[0173] like Figure 3 As shown, the high-rise concealed honeycomb intelligent spraying system proposed in this invention also includes a post-spraying image verification module.

[0174] Specifically, the post-spraying image acquisition unit reacquires on-site image data after the spraying end permit is issued, enabling the system to determine whether there are uncovered areas based on the visual results after spraying. The post-spraying on-site image data still enters the aforementioned image acquisition, calibration, preprocessing, cellular detection, and target localization processes to maintain consistency in the data processing chain.

[0175] Furthermore, the post-spraying candidate region data represents the suspected honeycomb areas that are still identified after spraying. The post-spraying target honeycomb region is the target area redefined based on the post-spraying candidate region data. Residual region data is obtained through regional difference analysis between the post-spraying target honeycomb region and the pre-spraying target honeycomb region. Regional difference analysis can employ area difference, boundary difference, or overlap difference. The residual ratio can be expressed as:

[0176]

[0177] In the formula, The residual percentage, This represents the area of ​​the target honeycomb region after spraying. This refers to the area of ​​the target honeycomb region before spraying. The preset residual conditions can be set to a residual ratio greater than 0.15 to 0.40, or a residual area center distance from the spray center pixel coordinates greater than 20 to 60 pixels.

[0178] When the residual area data meets the preset residual conditions in the permitted parameters, the residual target generation unit generates residual center coordinates based on the residual area data. The residual center coordinates can be used as target coordinates for subsequent compensation alignment and input to the vision pointing mapping module, so that the system continues to perform compensation processing according to the aforementioned chain of "residual center coordinates - jet pointing deviation data - attitude adjustment amount - vision closed-loop control".

[0179] For example, after the system generates a spray end permission state, it re-acquires post-spray scene image data. The post-spray detection unit still detects residual texture in the lower right region of the honeycomb, and the residual region generation unit calculates the residual ratio as 0.28. If the preset residual condition in the permission parameters is a residual ratio greater than 0.20, the residual target generation unit generates residual center coordinates, such as (710, 410), based on the residual region data for subsequent compensation alignment.

[0180] The present invention has the following advantages:

[0181] (1): This invention acquires, calibrates, and preprocesses on-site image data of the target area to obtain the image center coordinates, the pixel coordinates of the spray center, and preprocessed image data, thus providing a clear image coordinate basis for the honeycomb identification process. Furthermore, by using color features, contour features, pore texture features, and occlusion boundary features to identify hidden honeycombs on the facades of high-rise buildings, window sills, eaves, or the back of equipment, it can reduce the problem of unstable target positioning caused by complex backgrounds, local occlusion, or changes in lighting.

[0182] (2): This invention generates the target honeycomb region, honeycomb center coordinates, and target effective state based on the honeycomb candidate region data, and combines the image timestamp with the synchronous jet end attitude data, support length data, and medium channel state data, so that the honeycomb position identified in the image can correspond to the current attitude and operating state of the jet end. Thus, the system can generate jet pointing deviation data based on the deviation between the honeycomb center coordinates and the jet center pixel coordinates, and use this data for jet end attitude adjustment, so that the image recognition result can directly participate in the jet direction control.

[0183] (3): The present invention also generates attitude adjustment amount based on the spray direction deviation data through a visual closed-loop control module, and controls the spray end to perform pitch adjustment and horizontal deflection, so that the spray end can perform directional correction around the identified target honeycomb area. This effect corresponds to the mechanical basis of the spray head being able to adjust up and down, deflect left and right, and the medium channel passing through the support structure for transportation. It can combine mechanical adjustment capability with image recognition results to realize directional spray control for high-rise concealed honeycomb.

[0184] (4): In the injection permission output stage, the present invention generates a medium injection permission command by comprehensively adjusting the injection pointing deviation data, target validity status, area overlap status, and synchronization channel status data, and re-acquires the post-injection field image data for residual area analysis after injection. Thus, the injection process can form a continuous processing flow of "image identification before injection, pointing permission during injection, and image verification after injection", reducing the occurrence of injection when the target is misaligned, the target is invalid, or the channel status is abnormal.

[0185] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0186] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0187] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A high-rise concealed honeycomb intelligent spray system, characterized in that, include: The parameter storage module stores the acquisition parameters, calibration parameters, confidence parameters, alignment parameters, permission parameters, and limit parameters. The image acquisition and calibration module acquires on-site image data of the target area and generates preprocessed image data, image center coordinates, spray center pixel coordinates, and image timestamps based on the acquisition parameters and calibration parameters. The concealed cellular detection module includes: The detection region generation unit generates the target detection region in the preprocessed image data based on the image center coordinates, the jet center pixel coordinates, and a preset field of view offset range. The color feature extraction unit performs color interval matching on the target detection region to obtain the color matching region; The contour feature extraction unit performs edge extraction and connectivity analysis on the target detection region to obtain the contour connectivity region. The texture feature extraction unit performs pore texture density feature calculation on the target detection area to obtain the texture density area; The candidate region fusion unit fuses the color matching region, the contour connected region, and the texture density region to obtain the honeycomb candidate region data; The target localization module generates the target honeycomb region, honeycomb center coordinates, and target validity status based on the candidate honeycomb region data and confidence parameters.

2. The high-rise concealed honeycomb intelligent spray system according to claim 1, characterized in that, The image acquisition and calibration module includes: The image frame acquisition unit calls the preset acquisition period in the acquisition parameters to acquire the on-site image data, and generates the image timestamp for each frame of the on-site image data; The image correction unit performs distortion correction on the on-site image data according to the lens calibration parameters in the calibration parameters to obtain corrected image data. The center calibration unit generates the image center coordinates based on the pixel boundaries of the corrected image data, and generates the jet center pixel coordinates based on the jet center calibration parameters in the calibration parameters; The image preprocessing unit performs brightness normalization, noise suppression, and scale normalization on the corrected image data to obtain the preprocessed image data.

3. The high-rise concealed honeycomb intelligent spraying system according to claim 2, characterized in that, The concealed honeycomb detection module also includes: The occlusion relationship analysis unit performs local occlusion detection on the preprocessed image data of consecutive frames to obtain occlusion boundary data and visible area data; The hidden feature extraction unit extracts pore texture points, arc contour segments and building adjacency boundaries within the visible area data. The building adjacency boundaries are the set of boundary pixels adjacent to the corresponding area of ​​the honeycomb candidate area data and the building edge, wall corner, eaves edge or equipment shell edge. The boundary completion unit completes the honeycomb candidate region data based on the occlusion boundary data, the arc contour segment, and the building adjacent boundary, and sends the completed honeycomb candidate region data to the target positioning module.

4. The high-rise concealed honeycomb intelligent spraying system according to claim 3, characterized in that, The target localization module includes: The candidate region scoring unit receives the completed honeycomb candidate region data and generates candidate confidence scores based on the area, contour closure, texture density, building adjacency boundary, continuous frame boundary stability, and positional relationship with the pixel coordinates of the jet center in the honeycomb candidate region data. The target region determination unit determines the candidate honeycomb regions whose confidence levels satisfy the confidence parameters as the target honeycomb regions; The center coordinate generation unit generates the center coordinates of the honeycomb based on the outer boundary, centroid position, pore texture point distribution, and pixel distribution centroid of the target honeycomb region. The target stability determination unit generates the target valid state based on the honeycomb center coordinates in consecutive frames.

5. The high-rise concealed honeycomb intelligent spraying system according to claim 1, characterized in that, The system also includes a state synchronization module, configured to generate synchronized attitude data, synchronized length data and synchronized channel state data based on the image timestamp; The status synchronization module includes: The attitude data acquisition unit acquires pitch position detection signals and horizontal position detection signals from the injection end, generates pitch angle data based on the pitch position detection signals, and generates horizontal angle data based on the horizontal position detection signals. The length data acquisition unit acquires the displacement detection signal of the support adjustment object connected to the injection end, and generates support length data based on the displacement detection signal; The channel status acquisition unit acquires the pressure detection signal and valve status signal of the medium channel object connected to the injection end, and generates medium channel status data; The timing synchronization unit synchronizes the pitch angle data, horizontal angle data, support length data, and medium channel status data according to the image timestamp to obtain the synchronized attitude data, synchronized length data, and synchronized channel status data.

6. The high-rise concealed honeycomb intelligent spraying system according to claim 5, characterized in that, The system also includes a visual pointing mapping module, configured to generate jet pointing deviation data based on the target's effective state, the honeycomb center coordinates, the jet center pixel coordinates, the synchronization attitude data, and the synchronization length data. The visual pointing mapping module includes: The effective state determination unit receives the target effective state and allows the generation of the injection pointing deviation data when the target effective state is an effective state. The image deviation generation unit generates image deviation data based on the position of the honeycomb center coordinates relative to the pixel coordinates of the jet center when the generation of the jet pointing deviation data is allowed. The jet coordinate establishment unit establishes a jet coordinate system with the jet centerline as the reference based on the calibration parameters, the synchronization attitude data, and the synchronization length data. The pointing deviation generation unit maps the honeycomb center coordinates to the jet coordinate system to obtain the target jet coordinates, and generates the jet pointing deviation data based on the image deviation data and the target jet coordinates.

7. The high-rise concealed honeycomb intelligent spraying system according to claim 6, characterized in that, The system also includes a vision closed-loop control module, configured to generate attitude adjustment amount based on the injection pointing deviation data and limit parameters, and control the pitch adjustment and horizontal deflection of the injection end; The vision closed-loop control module includes: The adjustment amount generation unit reads the limit parameter from the parameter storage module and generates the attitude adjustment amount, including pitch adjustment amount and horizontal adjustment amount, based on the injection pointing deviation data and the limit parameter. The bidirectional adjustment unit controls the spray end to rotate in the pitch direction according to the pitch adjustment amount, and controls the spray end to rotate in the horizontal direction according to the horizontal adjustment amount; The deviation update unit triggers the image acquisition and calibration module to acquire the adjusted on-site image data after the bidirectional adjustment unit completes its rotation, and updates the injection pointing deviation data according to the adjusted honeycomb center coordinates.

8. The high-rise concealed honeycomb intelligent spraying system according to claim 7, characterized in that, It also includes a spray license output module, which comprises: The visual alignment determination unit generates a visual alignment state based on the adjusted jet pointing deviation data and the alignment parameters. The spray area generation unit generates a spray-allowed image area based on the target honeycomb area and generates a spray center image area based on the spray center pixel coordinates. The region overlap determination unit performs overlap determination on the jetting permissible image region and the jetting center image region to obtain the region overlap status. The permission instruction generation unit generates a media spraying permission instruction based on the visual alignment state, the area overlap state, the target validity state, the synchronization channel status data, and the permission parameters. The media spraying permission instruction includes a spraying start permission state and a spraying end permission state.

9. The high-rise concealed honeycomb intelligent spraying system according to claim 8, characterized in that, It also includes a post-spray image verification module, which includes: The post-spraying image acquisition unit triggers the image acquisition and calibration module to acquire post-spraying on-site image data after the spraying end permission state is generated. The post-spraying detection unit enables the concealed honeycomb detection module to generate post-spraying honeycomb candidate region data based on the post-spraying field image data. The residual area generation unit enables the target positioning module to generate a target honeycomb region after spraying based on the candidate honeycomb region data after spraying, and to perform regional difference analysis between the target honeycomb region after spraying and the target honeycomb region before spraying to obtain residual area data. The residual target generation unit generates residual center coordinates based on the residual area data when the residual area data meets the preset residual conditions in the permission parameters.

Images