A method and system for detecting an induction type flusher

Abstract

The application provides an inductive flusher detection method and system. The method comprises the following steps: monitoring a preset monitoring area in real time; when a target human body appears in the monitoring area, determining whether the current posture of the target human body is a first posture; if yes, performing motion analysis on the target human body; based on the analysis result, determining whether the target human body has a first motion; obtaining the sensing result of a capacitive sensing device, combining the sensing result with the determination result, and controlling the inductive flusher to perform corresponding operation; and solving the problems of the existing inductive flusher, such as the detection insensitivity leading to the failure to flush water in time, the detection error leading to the false flushing of water at a time when the flush water is not needed, and the waste of water resources.

Description

Technical Field

[0001] This invention relates to the field of detection technology, and in particular to a detection method and system for sensor-operated flushers. Background Technology

[0002] Currently, sensor-operated flushers typically use infrared reflection or capacitive sensing technology. When a person's hand is placed within the infrared area of ​​the toilet's sensor, the infrared light emitted by the infrared emitter is reflected back to the infrared receiver due to the hand's obstruction. The signal is then processed by a microcomputer within the integrated circuit and sent to a pulse solenoid valve. Upon receiving the signal, the solenoid valve opens its valve core according to a specified instruction to control the flushing. When the person's hand leaves the infrared sensing range, the solenoid valve no longer receives a signal, and the valve core is reset by an internal spring to shut off the flush. However, there are drawbacks such as insensitive detection leading to delayed flushing or incorrect detection causing accidental flushing when flushing is not needed, thus wasting water resources.

[0003] Therefore, the present invention provides a detection method and system for sensor-operated flushers, so as to at least partially solve the problems existing in the prior art. Summary of the Invention

[0004] This invention provides a detection method and system for sensor-operated flushers, which solves the problems of existing sensor-operated flushers, such as insensitive detection leading to failure to flush in time or detection errors causing accidental flushing when flushing is not needed, thus wasting water resources.

[0005] This invention provides a method for detecting an induction flushing device, comprising:

[0006] S1: Real-time monitoring of the preset monitoring area;

[0007] S2: When a target human body appears in the monitoring area, it is determined whether the current posture of the target human body is the first posture. If so, motion analysis is performed on the target human body.

[0008] S3: Based on the analysis results, determine whether the target human body has performed a first action;

[0009] S4: Obtain the sensing result from the capacitive sensing device, and combine it with the judgment result to control the sensor-type flusher to perform the corresponding operation.

[0010] Preferably, a sensor-operated flushing device detection method, S1: real-time monitoring of a preset monitoring area, including:

[0011] When the sensor-activated flushing device detection system is connected to a power source, it begins to acquire the first infrared image of the monitoring area in real time.

[0012] Determine whether there is a suspected human body in the first infrared image. If so, remove the background from the first infrared image and determine whether the remaining image to be identified after removal is the target human body.

[0013] Preferably, in a sensor-activated flushing device detection method, S2: when a target human body appears within the monitoring area, it is determined whether the current posture of the target human body is a first posture; if so, motion analysis is performed on the target human body, including:

[0014] When a target human body appears within the monitoring area, the current posture feature data of the target human body is acquired;

[0015] The current posture feature data is input into a preset posture judgment neural network to determine whether the current posture of the target human body is the first posture;

[0016] If so, the first posture change data of the target human body is calculated according to a preset period, and the first posture change data is input into a preset motion analysis neural network to determine whether the target human body has performed a first action.

[0017] Preferably, a method for detecting an induction flushing device includes, in step S4: acquiring the sensing result from the capacitive sensing device and, in conjunction with the judgment result, controlling the induction flushing device to perform corresponding operations, including:

[0018] When the sensor flushing detection system is connected to power, it begins to sense whether a target is detected within the preset sensing area.

[0019] If present, the sensor-operated flushing device will be controlled to start flushing, based on the activation signal generated when the first action occurs.

[0020] Otherwise, output the first alarm message.

[0021] Preferably, a sensor-activated flushing device detection method involves determining whether a suspected human body exists within the first infrared image; if so, background removal is performed on the first infrared image, and determining whether the remaining image to be identified after removal is the target human body, including:

[0022] The difference region between the first infrared image and the preset image is determined, and it is determined whether the area of ​​the difference region is greater than a first preset threshold. If so, it is determined that there is a suspected human body in the first infrared image, and a preset number of first images corresponding to the first infrared image are obtained based on the preset spectral band gradient.

[0023] The first image is denoised, and the contrast of the denoised first image is set to a preset contrast.

[0024] Preset morphological operations are performed on the difference regions in the processed first image, and the corresponding spatial domain morphological features are extracted to generate a first feature image. At the same time, time-frequency analysis is performed on the difference regions in the processed first image, and the corresponding time-frequency features are extracted to generate a second feature image.

[0025] All first feature images and second feature images are defined according to a preset standard histogram and unified under the same preset coordinate system;

[0026] The first feature image and the second feature image are subjected to contour wave transform to obtain the first high-frequency coefficient and the first low-frequency coefficient corresponding to the first feature image and the second high-frequency coefficient and the second low-frequency coefficient corresponding to the second feature image;

[0027] Based on the row frequency and column frequency of the first feature image and the gray value of each pixel, the corresponding spatial frequency value is calculated.

[0028] Based on the preset window weighting coefficients, directional filtering operators, and the gray value of each pixel in the second feature image, the corresponding edge energy value is calculated.

[0029] The first fusion coefficient is calculated based on the spatial frequency value, the edge energy value, the first low-frequency coefficient, and the second low-frequency coefficient.

[0030] Based on the first fusion coefficient and the first preset fusion algorithm, the first feature image and the second feature image corresponding to each first image are fused using low-frequency contour waves. Based on the preset convolutional neural network model and the first high-frequency coefficient and the second high-frequency coefficient, the first feature image and the second feature image corresponding to each first image are fused using high-frequency contour waves to obtain the first fused image corresponding to each first image.

[0031] Based on the second preset fusion algorithm and the preset spectral band gradient, all first fusion images are fused. The region of interest with the highest overlap in all first fusion images during the image fusion process is marked as the background region. The background regions in the first feature image and the second feature image are removed, and the remaining region is used as the image to be identified.

[0032] Extract the first feature data and the second feature data corresponding to the image to be identified in the first feature image and the second feature image, and input them into a classifier trained on samples to obtain the recognition result;

[0033] The recognition result includes: the image to be recognized is a target human body or the image to be recognized is not a target human body.

[0034] Preferably, the sensor-based flushing device detection method, after obtaining the identification result, includes:

[0035] When a target human body appears in the monitoring area, the second infrared image corresponding to the image to be identified is acquired, and the compressed first infrared image is slid through a window according to a preset gradient window size and movement step size to obtain a preset number of first candidate windows.

[0036] Multiple local limb images are obtained from the standard pose image. All first candidate windows and local limb images are binary encoded to obtain the corresponding binary encoding matrix.

[0037] Based on the binary encoding matrix corresponding to the first candidate window and the local limb image, the first Hamming distance between each first candidate window and the local limb image is calculated, and the first candidate window with a first Hamming distance greater than a first preset threshold is determined as the second candidate window.

[0038] The second candidate window is mapped onto the first infrared image based on the mapping matrix to obtain the third candidate window. The third candidate window is binary encoded to determine a new binary encoding matrix. The second Hamming distance between the three candidate windows and the local limb image is calculated based on the new binary encoding matrix. The third candidate window with a second Hamming distance greater than a second preset threshold is determined as the fourth candidate window.

[0039] The region composed of all the fourth candidate windows is used to form the local limb image to be judged.

[0040] The algorithm detects the local limb image to be judged to obtain multiple corresponding first calibration points, and obtains the coordinates of the corresponding first calibration points based on the preset coordinate system;

[0041] Based on the image of the local limb to be judged, the corresponding calibration feature points in the preset calibration library are retrieved, and the coordinates of the first calibration point are corrected according to the offset distance between the coordinates of the first calibration point and the calibration feature points.

[0042] A first pose frame corresponding to the current pose of the target human body is constructed based on a preset construction algorithm and the coordinates of the first calibration point.

[0043] The first pose frame is extracted based on a preset feature data extraction neural network to obtain the current pose feature data of the target human body.

[0044] Preferably, the sensor-based flushing device detection method involves determining a new binary encoding matrix by binary encoding the third candidate window, including:

[0045] The binary encoding matrix is ​​reduced in dimensionality, and the mapping matrix is ​​initialized based on the number of bits of the reduced binary encoding matrix. The projection matrix is ​​obtained based on the reduced binary encoding matrix and the initialized mapping matrix.

[0046] Set the values ​​corresponding to the non-negative positions in the projection matrix to 1, and set the values ​​corresponding to the negative positions in the projection matrix to 0 to obtain the rotation matrix;

[0047] The transition matrix is ​​obtained based on the dimension-reduced binary encoding matrix, the initialized mapping matrix, and the transpose of the rotation matrix.

[0048] Calculate the first product of the transition matrix and the transpose of the transition matrix, the second product of the transpose of the transition matrix and the transition matrix, and set the third product of the singular value decomposition result of the first product and the singular value decomposition result of the second product as the new binary encoding matrix.

[0049] Preferably, a sensor-activated flushing device detection method includes calculating the first posture change data of the target human body according to a preset cycle, comprising:

[0050] The infrared image of the target human body is detected by an algorithm to obtain multiple second calibration points. Based on the coordinates of the second calibration points in a preset coordinate system and a preset construction algorithm, a corresponding second pose frame is generated. Based on a preset point tracking algorithm, the second calibration points in the preset coordinate system are tracked.

[0051] Based on the tracking results, the movement trajectory of each second calibration point within a preset period is obtained. Based on the movement trajectory and the second posture frame, the first posture change data of the target human body in the preset coordinate system is obtained.

[0052] Preferably, a method for detecting an induction flushing device, S4: acquiring the sensing result of the capacitive sensing device, and combining it with the judgment result to control the induction flushing device to perform corresponding operations, further includes:

[0053] When a target is detected within the preset sensing area, if the determination result is that the start signal generated when the first action was performed was not received, and the real-time monitoring did not detect the target human body, then the first alarm message is issued.

[0054] When a target is detected within a preset sensing area, if no activation signal generated when the first action occurs is received and the target human body is detected in real time, the second posture change data corresponding to the target human body within the period of the first moment when the target is detected within the preset sensing area is obtained, and the preset action analysis neural network is corrected based on the second posture change data.

[0055] Preferably, an induction-type flusher detection system includes:

[0056] Monitoring module: Real-time monitoring of preset monitoring areas;

[0057] First judgment module: When a target human body appears in the monitoring area, it determines whether the current posture of the target human body is the first posture. If so, it performs motion analysis on the target human body.

[0058] The second judgment module: Based on the analysis results, determines whether the target human body has performed the first action;

[0059] Control module: Acquires the sensing results from the capacitive sensing device and, based on the judgment results, controls the sensor-operated flushing device to perform corresponding operations.

[0060] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0061] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0062] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0063] Figure 1 This is a flowchart of a sensor-type flushing device detection method according to an embodiment of the present invention;

[0064] Figure 2 This is a structural diagram of an induction flushing device detection system according to an embodiment of the present invention;

[0065] Figure 3 This is a schematic diagram of a capacitive sensing device in an embodiment of the present invention;

[0066] Figure 4 This is a schematic diagram of a flushing control device according to an embodiment of the present invention. Detailed Implementation

[0067] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0068] Example 1:

[0069] This invention provides a method for detecting induction-type flushers, such as... Figure 1 , 3 As shown in Figure 4, it includes:

[0070] S1: Real-time monitoring of the preset monitoring area;

[0071] S2: When a target human body appears in the monitoring area, it is determined whether the current posture of the target human body is the first posture. If so, motion analysis is performed on the target human body.

[0072] S3: Based on the analysis results, determine whether the target human body has performed a first action;

[0073] S4: Obtain the sensing result from the capacitive sensing device, and combine it with the judgment result to control the sensor-type flusher to perform the corresponding operation.

[0074] In this embodiment, the sensor-activated flusher can be installed on the smart cover of a toilet seat or squat toilet, enabling functions such as flushing upon leaving the seat and one-button flushing, and does not interfere with the manual flushing function after installation.

[0075] In this embodiment, the preset monitoring area is monitored by infrared surveillance and includes the toilet.

[0076] In this embodiment, the first posture is the posture of starting to use a toilet or a squat toilet, such as: sitting posture, standing posture, squatting posture, etc.

[0077] In this embodiment, the first action is the action that the target human body should take when the flushing should be turned on, such as reaching out to sense the movement or getting up from the seat.

[0078] In this embodiment, the sensing result is obtained through... Figure 3 It is achieved through a capacitive sensing device, which has a faster capacitive sensing response and stronger anti-interference performance.

[0079] In this embodiment, controlling the corresponding operation of the flusher is achieved by... Figure 4 The flushing control device is installed on the switch control device of the water tank to control the flusher.

[0080] The beneficial effects of the above technical solution are: by adopting a sensing method that combines capacitive sensing and infrared sensing, it retains the advantages of fast response and strong anti-interference performance of capacitive sensing technology, while solving the problems of existing inductive flushers, such as insensitive detection leading to failure to flush in time or detection errors causing accidental flushing when flushing is not needed, thus wasting water resources.

[0081] Example 2:

[0082] Based on the above embodiment 1, the sensor-type flushing device detection method, S1: real-time monitoring of a preset monitoring area, includes:

[0083] When the sensor-activated flushing device detection system is connected to a power source, it begins to acquire the first infrared image of the monitoring area in real time.

[0084] Determine whether there is a suspected human body in the first infrared image. If so, remove the background from the first infrared image and determine whether the remaining image to be identified after removal is the target human body.

[0085] In this embodiment, the sensor-operated flushing device uses a lithium battery contact power source, and the sensor-operated flushing device detection system is connected to the power source when the sensing device is installed.

[0086] In this embodiment, the first infrared image is the infrared image corresponding to the scene within the monitoring area.

[0087] In this embodiment, the image to be identified is the image remaining after removing the background from the first infrared image.

[0088] In this embodiment, the target human body is the user (the human body to be sensed).

[0089] The beneficial effects of the above technical solution are: real-time acquisition of infrared images within the monitoring area, and monitoring to determine whether there is a suspected human body. When a suspected human body is initially determined to exist, the infrared image is then subjected to background removal for detailed judgment to determine whether it is the target human body. This achieves rapid judgment when there is no target human body in the monitoring area, and ensures accurate judgment when there is a target human body in the monitoring area. In this way, the efficiency of infrared detection can be guaranteed.

[0090] Example 3:

[0091] Based on the above embodiment 1, the sensor-activated flushing device detection method, S2: when a target human body appears in the monitoring area, it is determined whether the current posture of the target human body is a first posture; if so, motion analysis is performed on the target human body, including:

[0092] When a target human body appears within the monitoring area, the current posture feature data of the target human body is acquired;

[0093] The current posture feature data is input into a preset posture judgment neural network to determine whether the current posture of the target human body is the first posture;

[0094] If so, the first posture change data of the target human body is calculated according to a preset period, and the first posture change data is input into a preset motion analysis neural network to determine whether the target human body has performed a first action.

[0095] In this embodiment, motion analysis is to determine whether the last action performed by the target human body is the first action.

[0096] In this embodiment, the current posture feature data is the feature data that characterizes the current posture of the target human body.

[0097] In this embodiment, a preset pose determination neural network is used to input the current pose feature data and obtain a determination result on whether the current pose is the first pose.

[0098] In this embodiment, the first posture change data is the posture change data within the last cycle corresponding to the current time.

[0099] In this embodiment, a preset motion analysis neural network is used to input posture change data and obtain a judgment result on whether the current action of the target human body is the first action.

[0100] The beneficial effects of the above technical solution are as follows: when a target human body is detected in the monitoring area, it is determined whether the target human body is in a posture of using a toilet seat or squat toilet. If so, the posture change data of the target human body is collected periodically and the action analysis is performed to determine whether the user has finished using the toilet seat or squat toilet. This realizes the monitoring and posture and action analysis of the entire process from the start of the target human body using the toilet seat or squat toilet to the end of use, providing a basis for subsequent judgment on whether to control the flushing device to turn on the flush. It is more accurate than the judgment of traditional sensor devices, avoiding accidental flushing and water waste.

[0101] Example 4:

[0102] Based on the above embodiment 1, the sensor-type flushing device detection method, S4: acquiring the sensing result of the capacitive sensing device, and combining the judgment result to control the sensor-type flushing device to perform corresponding operations, including:

[0103] When the sensor flushing detection system is connected to power, it begins to sense whether a target is detected within the preset sensing area.

[0104] If present, the sensor-operated flushing device will be controlled to start flushing, based on the activation signal generated when the first action occurs.

[0105] Otherwise, output the first alarm message.

[0106] In this embodiment, the sensing target is the sensing target of the capacitive sensing device, which is determined according to a preset setting;

[0107] For example: if the sensing range of the capacitive sensor is 30cm, and the sensing target is the user's hand gesture, then the capacitive sensor needs to be triggered by the user's hand gesture.

[0108] If the sensing range of the capacitive sensing device is 1m, and the sensing target is the human body, then the capacitive sensing device needs to be triggered by the user's standing or sitting posture.

[0109] In this embodiment, the activation signal is an indication signal generated when the judgment result is the first action of sound emission. The activation signal combined with the sensing result can control the opening of the flushing device.

[0110] In this embodiment, the first alarm message is used to remind the user that the capacitive sensing device has malfunctioned.

[0111] The beneficial effects of the above technical solution are: combining the results of capacitive sensing and infrared sensing to determine whether flushing needs to be started. Flushing can only be started when both sense the flushing trigger action or the sensing target. This solves the problems of existing sensor-type flushers where detection errors lead to accidental flushing when flushing is not needed, wasting water resources. Moreover, the first alarm message can remind the user that the capacitive sensing device has malfunctioned, thus avoiding the occurrence of capacitive sensing failure.

[0112] Example 5:

[0113] Based on the above embodiment 2, the sensor-type flushing device detection method, which determines whether there is a suspected human body in the first infrared image, and if so, performs background removal on the first infrared image, and determines whether the remaining image to be identified after removal is the target human body, includes:

[0114] The difference region between the first infrared image and the preset image is determined, and it is determined whether the area of ​​the difference region is greater than a first preset threshold. If so, it is determined that there is a suspected human body in the first infrared image, and a preset number of first images corresponding to the first infrared image are obtained based on the preset spectral band gradient.

[0115] The first image is denoised, and the contrast of the denoised first image is set to a preset contrast.

[0116] Preset morphological operations are performed on the difference regions in the processed first image, and the corresponding spatial domain morphological features are extracted to generate a first feature image. At the same time, time-frequency analysis is performed on the difference regions in the processed first image, and the corresponding time-frequency features are extracted to generate a second feature image.

[0117] All first feature images and second feature images are defined according to a preset standard histogram and unified under the same preset coordinate system;

[0118] The first feature image and the second feature image are subjected to contour wave transform to obtain the first high-frequency coefficient and the first low-frequency coefficient corresponding to the first feature image and the second high-frequency coefficient and the second low-frequency coefficient corresponding to the second feature image;

[0119] Based on the row frequency and column frequency of the first feature image and the gray value of each pixel, the corresponding spatial frequency value is calculated.

[0120] Based on the preset window weighting coefficients, directional filtering operators, and the gray value of each pixel in the second feature image, the corresponding edge energy value is calculated.

[0121] The first fusion coefficient is calculated based on the spatial frequency value, the edge energy value, the first low-frequency coefficient, and the second low-frequency coefficient.

[0122] Based on the first fusion coefficient and the first preset fusion algorithm, the first feature image and the second feature image corresponding to each first image are fused using low-frequency contour waves. Based on the preset convolutional neural network model and the first high-frequency coefficient and the second high-frequency coefficient, the first feature image and the second feature image corresponding to each first image are fused using high-frequency contour waves to obtain the first fused image corresponding to each first image.

[0123] Based on the second preset fusion algorithm and the preset spectral band gradient, all first fusion images are fused. The region of interest with the highest overlap in all first fusion images during the image fusion process is marked as the background region. The background regions in the first feature image and the second feature image are removed, and the remaining region is used as the image to be identified.

[0124] Extract the first feature data and the second feature data corresponding to the image to be identified in the first feature image and the second feature image, and input them into a classifier trained on samples to obtain the recognition result;

[0125] The recognition result includes: the image to be recognized is a target human body or the image to be recognized is not a target human body.

[0126] In this embodiment, the preset image is a pre-stored infrared image of the monitoring area when no one is present.

[0127] In this embodiment, the difference region is the region that is different between the first infrared image and the preset image.

[0128] In this embodiment, the first preset threshold is the minimum size of the target human body in the first infrared image.

[0129] In this embodiment, the suspected human body is the physical object corresponding to the difference area in the first infrared image that is initially judged to be a target human body.

[0130] In this embodiment, the preset spectral band gradients include, for example, 100nm, 200nm, 300nm, 400nm, and 500nm.

[0131] In this embodiment, the first image is the image corresponding to the first infrared image under the spectral band values ​​contained in the preset spectral band gradient.

[0132] In this embodiment, morphological operations include, for example, corrosion and expansion as well as opening and closing operations.

[0133] In this embodiment, the spatial domain morphological features are the morphological features of the target human body within the spatial domain.

[0134] In this embodiment, the first feature image is the image corresponding to the morphological features of the infrared image of the target human body in the spatial domain.

[0135] In this embodiment, time-frequency analysis refers to analyzing how various frequency components of an image change spatially, and using local frequency and other features of the signal to achieve the purpose of image analysis and processing.

[0136] In this embodiment, the time-frequency feature is the spatial variation feature of various frequency components of the image.

[0137] In this embodiment, the second feature image is the image corresponding to the spatial variation characteristics of various frequency components of the infrared image of the target human body.

[0138] In this embodiment, the standard histogram is an image in which the grayscale value of each pixel is set in advance.

[0139] In this embodiment, the contour wave transform is an effective multi-resolution image representation method that approximates the image with sub-bands in different directions at different scales, which can better capture the internal geometric structure of the image and obtain discriminative features.

[0140] In this embodiment, the first high-frequency coefficient is the high-frequency coefficient obtained after the first feature image undergoes contour wave transformation;

[0141] The first low-frequency coefficient is the low-frequency coefficient obtained after the first feature image is transformed by the contour wave transform.

[0142] The second high-frequency coefficient is the high-frequency coefficient obtained after the second feature image is transformed by the contour wave transform.

[0143] The second low-frequency coefficient is the low-frequency coefficient obtained after the second feature image is transformed by the contour wave.

[0144] In this embodiment, based on the row frequency and column frequency of the first feature image and the grayscale value of each pixel, the corresponding spatial frequency value is calculated, including:

[0145]

[0146] In the formula, ξ is the spatial frequency value, n is the total number of rows of pixels in the first feature image, m is the total number of columns of pixels in the first feature image, i is the row number of pixels in the first feature image, j is the column number of pixels in the first feature image, a1(i,j) is the gray value of the pixel in the i-th row and j-th column of the first feature image, ξ1 is the column frequency of the first feature image, ξ2 is the row frequency of the first feature image; a1(i,j-1) is the gray value of the pixel in the i-th row and (j-1)-th column of the first feature image, and a1(i-1,j) is the gray value of the pixel in the (i-1)-th row and (j-1)-th column of the first feature image.

[0147] For example: In the first feature image, the gray values ​​of the pixels in the first row are 5, 10, and 15 respectively, the gray values ​​of the pixels in the third row are 4, 9, and 14 respectively, the gray values ​​of the pixels in the third row are 3, 8, and 13 respectively, the column frequency ξ1 is 100, and the row frequency ξ2 is 10. Then the corresponding spatial frequency value ξ is calculated to be 1225.

[0148] In this embodiment, based on preset window weighting coefficients, directional filtering operators, and the grayscale value of each pixel in the second feature image, the corresponding edge energy value is calculated, including:

[0149]

[0150] In the formula, E is the edge energy value, p is the total number of rows of pixels in the second feature image, q is the total number of columns of pixels in the second feature image, a is the row number of pixels in the second feature image, b is the column number of pixels in the second feature image, a2(a,b) is the gray value of the pixel in the i-th row and j-th column of the second feature image, e(a,b) is the energy value of the pixel in the i-th row and j-th column of the second feature image, σ is the preset window weighting coefficient, θ1 is the horizontal filtering operator, and θ2 is the vertical filtering operator;

[0151] For example, in the second feature image, the gray values ​​of the pixels in the first row are 5 and 10 respectively, and the gray values ​​of the pixels in the second row are 4 and 9 respectively. The energy values ​​of the first row in the second feature image are 1 and 2 respectively, and the energy values ​​of the second row are 5 and 10 respectively. The preset window weighting coefficient σ is 0.2, the horizontal filtering operator θ1 is 0.3, and the vertical filtering operator θ2 is 0.4. Then the edge energy value is 1.35.

[0152] In this embodiment, a first fusion coefficient is calculated based on the spatial frequency value, the edge energy value, the first low-frequency coefficient, and the second low-frequency coefficient. The first fusion coefficient is obtained by multiplying the ratio of the spatial frequency value to a preset spatial frequency value, the ratio of the edge energy value to a preset edge energy value, and the ratio of the first low-frequency coefficient to the second low-frequency coefficient.

[0153] The first fusion coefficient is the low-frequency fusion of the first feature image and the second feature image in combination with the first fusion algorithm.

[0154] In this embodiment, the first fusion algorithm is the algorithm used for low-frequency fusion.

[0155] In this embodiment, the second fusion algorithm is the algorithm used for high-frequency fusion.

[0156] In this embodiment, low-frequency contour wave fusion is the low-frequency fusion process of the contour wave transform algorithm.

[0157] In this embodiment, high-frequency contour wave fusion is the high-frequency fusion process of the contour wave transform algorithm.

[0158] In this embodiment, the first fused image is the image obtained after the first feature image and the corresponding second feature image are fused by low-frequency fusion and high-frequency fusion.

[0159] In this embodiment, the image to be identified is the image used to identify whether it is the target human body.

[0160] In this embodiment, the region of interest with the highest overlap is the region with the highest image feature similarity in the first fused image during the fusion process.

[0161] In this embodiment, the classifier is used to determine whether the image to be identified is the target human body.

[0162] The beneficial effects of the above technical solution are as follows: This invention is based on spatial domain processing and time-frequency analysis of infrared images, combined with multispectral band gradient feature fusion. Since common infrared images exhibit significant time-frequency features after wavelet transform, using support vector machines for judgment and screening offers advantages such as high detection rate, fast processing speed, low interference rate, and high reliability. Furthermore, the multispectral band infrared image feature-level fusion algorithm utilizes the spectral characteristics of the image to compensate for the shortcomings of using only time-frequency feature methods for detection, effectively improving the detection accuracy.

[0163] Example 6:

[0164] Based on the above embodiment 5, the sensor-type flushing device detection method, after obtaining the identification result, includes:

[0165] When a target human body appears in the monitoring area, the second infrared image corresponding to the image to be identified is acquired, and the compressed first infrared image is slid through a window according to a preset gradient window size and movement step size to obtain a preset number of first candidate windows.

[0166] Multiple local limb images are obtained from the standard pose image. All first candidate windows and local limb images are binary encoded to obtain the corresponding binary encoding matrix.

[0167] Based on the binary encoding matrix corresponding to the first candidate window and the local limb image, the first Hamming distance between each first candidate window and the local limb image is calculated, and the first candidate window with a first Hamming distance greater than a first preset threshold is determined as the second candidate window.

[0168] The second candidate window is mapped onto the first infrared image based on the mapping matrix to obtain the third candidate window. The third candidate window is binary encoded to determine a new binary encoding matrix. The second Hamming distance between the three candidate windows and the local limb image is calculated based on the new binary encoding matrix. The third candidate window with a second Hamming distance greater than a second preset threshold is determined as the fourth candidate window.

[0169] The region composed of all the fourth candidate windows is used to form the local limb image to be judged.

[0170] The algorithm detects the local limb image to be judged to obtain multiple corresponding first calibration points, and obtains the coordinates of the corresponding first calibration points based on the preset coordinate system;

[0171] Based on the image of the local limb to be judged, the corresponding calibration feature points in the preset calibration library are retrieved, and the coordinates of the first calibration point are corrected according to the offset distance between the coordinates of the first calibration point and the calibration feature points.

[0172] A first pose frame corresponding to the current pose of the target human body is constructed based on a preset construction algorithm and the coordinates of the first calibration point.

[0173] The first pose frame is extracted based on a preset feature data extraction neural network to obtain the current pose feature data of the target human body.

[0174] In this embodiment, the number of rows in the binary encoding matrix is ​​the sum of the number of the first candidate window and the number of the local limb images, and the number of columns in the data matrix is ​​the number of pixels in the first candidate window and the local limb images.

[0175] In this embodiment, the window size of the preset gradient is, for example, 1nm. 2 2nm 2 3nm2 ……wait;

[0176] The preset gradient movement step size can be, for example, 1nm, 2nm, 3nm, etc.

[0177] In this embodiment, the local limb image to be judged is the local limb image in the infrared image of the target human body that is inconsistent with multiple local limb images in the standard posture image.

[0178] In this embodiment, algorithm detection refers to the algorithm for detecting the calibration points of the target human body.

[0179] In this embodiment, the first calibration point is the calibration point determined by the algorithm detection of the local limb image to be judged, that is, the reference point of the human body part in the process of posture change.

[0180] In this embodiment, the calibration feature points are the preset calibration points in multiple local limb images of the standard pose image.

[0181] In this embodiment, the construction algorithm is a preset algorithm that constructs a limb posture framework based on the possible movement trajectory and range of human limbs.

[0182] In this embodiment, the first pose frame is the pose frame constructed based on the current pose of the target human body after determining that the suspected human body is the target human body.

[0183] In this embodiment, a preset feature data extraction neural network is used to extract the pose feature data corresponding to the pose frame.

[0184] The beneficial effects of the above technical solution are as follows: using Hamming distance for comparison, since the XOR operation of the computer's internal arithmetic unit can be used when calculating Hamming distance, the similarity calculation of multiple local limb images of the alignment pose image and the infrared image of the target human body can be completed within microseconds. By using Hamming distance to calculate the similarity between multiple local limb images of the standard pose image and the infrared image of the target human body, and implementing a two-stage screening method from coarse to fine, the calculation time is greatly reduced while ensuring the validity of the results.

[0185] Example 7:

[0186] Based on the above embodiment 6, the sensor-type flushing device detection method further includes determining a new binary encoding matrix by performing binary encoding on the third candidate window, including:

[0187] The binary encoding matrix is ​​reduced in dimensionality, and the mapping matrix is ​​initialized based on the number of bits of the reduced binary encoding matrix. The projection matrix is ​​obtained based on the reduced binary encoding matrix and the initialized mapping matrix.

[0188] Set the values ​​corresponding to the non-negative positions in the projection matrix to 1, and set the values ​​corresponding to the negative positions in the projection matrix to 0 to obtain the rotation matrix;

[0189] The transition matrix is ​​obtained based on the dimension-reduced binary encoding matrix, the initialized mapping matrix, and the transpose of the rotation matrix.

[0190] Calculate the first product of the transition matrix and the transpose of the transition matrix, the second product of the transpose of the transition matrix and the transition matrix, and set the third product of the singular value decomposition result of the first product and the singular value decomposition result of the second product as the new binary encoding matrix.

[0191] In this embodiment, the projection matrix is ​​obtained based on the dimension-reduced binary encoding matrix and the initialized mapping matrix by multiplying the dimension-reduced binary encoding matrix and the initialized mapping matrix.

[0192] In this embodiment, the transition matrix is ​​obtained based on the dimension-reduced binary encoding matrix, the initialized mapping matrix, and the transpose of the rotation matrix. The result of multiplying the dimension-reduced binary encoding matrix, the initialized mapping matrix, and the transpose of the rotation matrix in sequence is the transition matrix.

[0193] The beneficial effects of the above technical solution are as follows: the third candidate window is binary encoded, and a new binary encoding matrix is ​​determined based on the dimension-reduced binary encoding matrix and the initialized mapping matrix. While reducing the dimension, similar data remain similar after the dimension reduction, and dissimilar data remain dissimilar after the dimension reduction. In this way, the candidate window is mapped to the vertices of the binary hyperplane to obtain the binary encoding. The feature vectors of multiple local limb images of the pose image can be uniformly calculated in one operation. This realizes the similarity calculation of multiple local limb images of the standard pose image and the infrared image of the target human body, and realizes a more refined screening method. While ensuring the validity of the results, the computation time is greatly reduced.

[0194] Example 8:

[0195] Based on the above embodiment 3, the sensor-type flushing device detection method calculates the first posture change data of the target human body according to a preset period, including:

[0196] The infrared image of the target human body is detected by an algorithm to obtain multiple second calibration points. Based on the coordinates of the second calibration points in a preset coordinate system and a preset construction algorithm, a corresponding second pose frame is generated. Based on a preset point tracking algorithm, the second calibration points in the preset coordinate system are tracked.

[0197] Based on the tracking results, the movement trajectory of each second calibration point within a preset period is obtained. Based on the movement trajectory and the second posture frame, the first posture change data of the target human body in the preset coordinate system is obtained.

[0198] In this embodiment, the second calibration point is the calibration point determined by algorithm detection of the infrared image of the target human body, that is, the reference point of the human body part during the posture change process.

[0199] In this embodiment, the second posture frame is the posture frame constructed to further determine whether the target human body has made the first action after determining that the target human body's current posture is the first posture.

[0200] In this embodiment, the point tracking algorithm is used to track the calibration points of the target human body and record the movement trajectory and movement distance of the calibration points.

[0201] The beneficial effects of the above technical solution are: by constructing a posture framework and tracking calibration points for the target human body, the limb changes of the target human body can be effectively and accurately recorded, and the limb changes can be recorded as specific data to obtain accurate first posture change data of the target human body, providing a data foundation for subsequent motion analysis of the target human body.

[0202] Example 9:

[0203] Based on the above embodiment 4, the sensor-activated flushing device detection method, S4: acquiring the sensing result of the capacitive sensing device and, in conjunction with the judgment result, controlling the sensor-activated flushing device to perform corresponding operations, further includes:

[0204] When a target is detected within the preset sensing area, if the determination result is that the start signal generated when the first action was performed was not received, and the real-time monitoring did not detect the target human body, then the first alarm message is issued.

[0205] When a target is detected within a preset sensing area, if no activation signal generated when the first action occurs is received and the target human body is detected in real time, the second posture change data corresponding to the target human body within the period of the first moment when the target is detected within the preset sensing area is obtained, and the preset action analysis neural network is corrected based on the second posture change data.

[0206] In this embodiment, real-time monitoring is achieved through an infrared detection device, which is the aforementioned device for acquiring and detecting infrared images. It is installed together with a capacitive sensing device on the wall behind the toilet.

[0207] In this embodiment, the second posture change data is the posture change data of the target human body within the period of the first moment when the target is sensed in the preset sensing area.

[0208] The beneficial effects of the above technical solution are: when the infrared device does not detect a target human body or the first action that can trigger the flushing device, but the capacitive sensing device senses the target that triggers the flushing device, it can be determined whether the capacitive sensing device is malfunctioning or the infrared sensing device is insensitive based on whether the infrared detection device detects the target human body. Based on the judgment result, the device and detection process can be further corrected, which is conducive to maintaining the accuracy of the detection process.

[0209] Example 10:

[0210] This invention provides an induction-type flusher detection system, comprising:

[0211] Monitoring module: Real-time monitoring of preset monitoring areas;

[0212] First judgment module: When a target human body appears in the monitoring area, it determines whether the current posture of the target human body is the first posture. If so, it performs motion analysis on the target human body.

[0213] The second judgment module: Based on the analysis results, determines whether the target human body has performed the first action;

[0214] Control module: Acquires the sensing results from the capacitive sensing device and, based on the judgment results, controls the sensor-operated flushing device to perform corresponding operations.

[0215] The beneficial effects of the above technical solution are as follows: by setting up a monitoring module, a first judgment module, a second judgment module, and a control module, and adopting a sensing method that combines capacitive sensing and infrared sensing, it not only retains the advantages of fast response and strong anti-interference performance of capacitive sensing technology, but also solves the problems of existing inductive flushing devices, such as insensitive detection leading to failure to flush in time or detection errors leading to accidental flushing when flushing is not needed, thus wasting water resources.

[0216] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for detecting an induction-type flushing device, characterized in that, include: S1: Real-time monitoring of the preset monitoring area; when the sensor flushing device detection system is connected to power, the first infrared image within the monitoring area is acquired in real time. Determine whether there is a suspected human body in the first infrared image. If so, remove the background from the first infrared image and determine whether the remaining image to be identified after removal is the target human body. The determination of whether the remaining image to be identified after removal is the target human body includes: The difference region between the first infrared image and the preset image is determined, and multiple first images corresponding to the first infrared image are obtained based on the preset spectral band gradient; The first image is denoised, and the contrast of the denoised first image is set to a preset contrast. Preset morphological operations are performed on the difference regions in the processed first image, and the corresponding spatial domain morphological features are extracted to generate a first feature image. At the same time, time-frequency analysis is performed on the difference regions in the processed first image, and the corresponding time-frequency features are extracted to generate a second feature image. The first feature image and the second feature image are subjected to contour wave transform to obtain the first high-frequency coefficient and the first low-frequency coefficient corresponding to the first feature image and the second high-frequency coefficient and the second low-frequency coefficient corresponding to the second feature image; Based on the row and column frequencies of the first feature image and the gray value of each pixel, the corresponding spatial frequency value is calculated; based on the preset window weighting coefficient, directional filtering operator and the gray value of each pixel in the second feature image, the corresponding edge energy value is calculated. Based on the spatial frequency value, the edge energy value, the first low-frequency coefficient, and the second low-frequency coefficient, a first fusion coefficient is calculated; based on the first fusion coefficient and a first preset fusion algorithm, the first feature image and the second feature image corresponding to each first image are fused using low-frequency contour waves, and based on a preset convolutional neural network model and the first high-frequency coefficient and the second high-frequency coefficient, the first feature image and the second feature image corresponding to each first image are fused using high-frequency contour waves to obtain a first fused image corresponding to each first image; Based on the second preset fusion algorithm and the preset spectral band gradient, all first fusion images are fused. The region of interest with the highest overlap in all first fusion images during the image fusion process is marked as the background region. The background regions in the first feature image and the second feature image are removed, and the remaining region is used as the image to be identified. First feature data and second feature data corresponding to the image to be identified in the first feature image and the second feature image are extracted and input into a classifier trained on samples to obtain a recognition result; wherein, the recognition result includes: the image to be identified is a target human body or the image to be identified is not a target human body; S2: When a target human body appears in the monitoring area, it is determined whether the current posture of the target human body is the first posture. If so, motion analysis is performed on the target human body. S3: Based on the analysis results, determine whether the target human body has performed a first action; S4: Obtain the sensing result from the capacitive sensing device, and combine it with the judgment result to control the sensor-type flusher to perform the corresponding operation.

2. The method for detecting an induction-type flushing device according to claim 1, characterized in that, S2: When a target human body appears within the monitoring area, it is determined whether the current posture of the target human body is the first posture. If so, motion analysis is performed on the target human body, including: When a target human body appears within the monitoring area, the current posture feature data of the target human body is acquired; The current posture feature data is input into a preset posture judgment neural network to determine whether the current posture of the target human body is the first posture; If so, the first posture change data of the target human body is calculated according to a preset period, and the first posture change data is input into a preset motion analysis neural network to determine whether the target human body has performed a first action.

3. The method for detecting an induction flushing device according to claim 1, characterized in that, S4: Obtain the sensing result from the capacitive sensing device, and based on the judgment result, control the sensor-operated flushing device to perform corresponding operations, including: When the sensor flushing detection system is connected to power, it begins to sense whether a target is detected within the preset sensing area. If present, the sensor-operated flushing device will be controlled to start flushing, based on the activation signal generated when the first action occurs. Otherwise, output the first alarm message.

4. The method for detecting an induction flushing device according to claim 1, characterized in that, After obtaining the recognition results, the following are included: When a target human body appears in the monitoring area, the second infrared image corresponding to the image to be identified is acquired, and the compressed first infrared image is slid through a window according to a preset gradient window size and movement step size to obtain a preset number of first candidate windows. Multiple local limb images are obtained from the standard pose image. All first candidate windows and local limb images are binary encoded to obtain the corresponding binary encoding matrix. Based on the binary encoding matrix corresponding to the first candidate window and the local limb image, the first Hamming distance between each first candidate window and the local limb image is calculated, and the first candidate window with a first Hamming distance greater than a first preset threshold is determined as the second candidate window. The second candidate window is mapped onto the first infrared image based on the mapping matrix to obtain the third candidate window. The third candidate window is binary encoded to determine a new binary encoding matrix. The second Hamming distance between the third candidate window and the local limb image is calculated based on the new binary encoding matrix. The third candidate window with a second Hamming distance greater than a second preset threshold is determined as the fourth candidate window. The region composed of all the fourth candidate windows is used to form the local limb image to be judged. The algorithm detects the local limb image to be judged and obtains multiple corresponding first calibration points. Based on a preset coordinate system, the coordinates of the corresponding first calibration points are obtained. Based on the image of the local limb to be judged, the corresponding calibration feature points in the preset calibration library are retrieved, and the coordinates of the first calibration point are corrected according to the offset distance between the coordinates of the first calibration point and the calibration feature points. A first pose frame corresponding to the current pose of the target human body is constructed based on a preset construction algorithm and the coordinates of the first calibration point. The first pose frame is extracted based on a preset feature data extraction neural network to obtain the current pose feature data of the target human body.

5. The method for detecting an induction flushing device according to claim 4, characterized in that, The third candidate window is binary encoded to determine a new binary encoding matrix, including: The binary encoding matrix is ​​reduced in dimensionality, and the mapping matrix is ​​initialized based on the number of bits of the reduced binary encoding matrix. The projection matrix is ​​obtained based on the reduced binary encoding matrix and the initialized mapping matrix. Set the values ​​corresponding to the non-negative positions in the projection matrix to 1, and set the values ​​corresponding to the negative positions in the projection matrix to 0 to obtain the rotation matrix; The transition matrix is ​​obtained based on the dimension-reduced binary encoding matrix, the initialized mapping matrix, and the transpose of the rotation matrix. Calculate the first product of the transition matrix and the transpose of the transition matrix, the second product of the transpose of the transition matrix and the transition matrix, and set the third product of the singular value decomposition result of the first product and the singular value decomposition result of the second product as the new binary encoding matrix.

6. The method for detecting an induction flushing device according to claim 2, characterized in that, The first posture change data of the target human body is calculated according to a preset period, including: The infrared image of the target human body is detected by an algorithm to obtain multiple second calibration points. Based on the coordinates of the second calibration points in a preset coordinate system and a preset construction algorithm, a corresponding second pose frame is generated. Based on a preset point tracking algorithm, the second calibration points in the preset coordinate system are tracked. Based on the tracking results, the movement trajectory of each second calibration point within a preset period is obtained. Based on the movement trajectory and the second posture frame, the first posture change data of the target human body in the preset coordinate system is obtained.

7. The method for detecting an induction flushing device according to claim 3, characterized in that, S4: Obtain the sensing result from the capacitive sensing device, and based on the judgment result, control the sensor-operated flushing device to perform corresponding operations, including: When a target is detected within the preset sensing area, if the determination result is that the start signal generated when the first action was performed was not received, and the real-time monitoring did not detect the target human body, then the first alarm message is issued. When a target is detected within a preset sensing area, if no activation signal generated when the first action occurs is received and the target human body is detected in real time, the second posture change data corresponding to the target human body within the period of the first moment when the target is detected within the preset sensing area is obtained, and the preset action analysis neural network is corrected based on the second posture change data.

8. A sensor-based flushing device detection system, characterized in that, The method for detecting an induction flusher according to any one of claims 1 to 7 includes: Monitoring module: Real-time monitoring of the preset monitoring area; when the sensor flushing detection system is connected to power, it acquires the first infrared image of the monitoring area in real time. Determine whether there is a suspected human body in the first infrared image. If so, remove the background from the first infrared image and determine whether the remaining image to be identified after removal is the target human body. The determination of whether the remaining image to be identified after removal is the target human body includes: The difference region between the first infrared image and the preset image is determined, and multiple first images corresponding to the first infrared image are obtained based on the preset spectral band gradient; The first image is denoised, and the contrast of the denoised first image is set to a preset contrast. Preset morphological operations are performed on the difference regions in the processed first image, and the corresponding spatial domain morphological features are extracted to generate a first feature image. At the same time, time-frequency analysis is performed on the difference regions in the processed first image, and the corresponding time-frequency features are extracted to generate a second feature image. The first feature image and the second feature image are subjected to contour wave transform to obtain the first high-frequency coefficient and the first low-frequency coefficient corresponding to the first feature image and the second high-frequency coefficient and the second low-frequency coefficient corresponding to the second feature image; Based on the row and column frequencies of the first feature image and the gray value of each pixel, the corresponding spatial frequency value is calculated; based on the preset window weighting coefficient, directional filtering operator and the gray value of each pixel in the second feature image, the corresponding edge energy value is calculated. Based on the spatial frequency value, the edge energy value, the first low-frequency coefficient, and the second low-frequency coefficient, a first fusion coefficient is calculated; based on the first fusion coefficient and a first preset fusion algorithm, the first feature image and the second feature image corresponding to each first image are fused using low-frequency contour waves, and based on a preset convolutional neural network model and the first high-frequency coefficient and the second high-frequency coefficient, the first feature image and the second feature image corresponding to each first image are fused using high-frequency contour waves to obtain a first fused image corresponding to each first image; Based on the second preset fusion algorithm and the preset spectral band gradient, all first fusion images are fused. The region of interest with the highest overlap in all first fusion images during the image fusion process is marked as the background region. The background regions in the first feature image and the second feature image are removed, and the remaining region is used as the image to be identified. First feature data and second feature data corresponding to the image to be identified in the first feature image and the second feature image are extracted and input into a classifier trained on samples to obtain a recognition result; wherein, the recognition result includes: the image to be identified is a target human body or the image to be identified is not a target human body; First judgment module: When a target human body appears in the monitoring area, it determines whether the current posture of the target human body is the first posture. If so, it performs motion analysis on the target human body. The second judgment module: Based on the analysis results, determines whether the target human body has performed the first action; Control module: Acquires the sensing results from the capacitive sensing device and, based on the judgment results, controls the sensor-operated flushing device to perform corresponding operations.

Images