Air outlet anti-leakage and anti-reverse-installation detection method based on image detection

Abstract

The invention discloses an air outlet anti-leakage and anti-reverse-installation detection method based on image detection, and belongs to the technical field of optical detection. The method comprises the following steps: determining a data acquisition range of image detection in an air outlet area, and delimiting a detection area according to structural parameters of the air outlet; by dividing conventional and error-prone areas, adopting a differential detection strategy and combining a three-dimensional point cloud comparison technology, the unqualified product rate and the detection time consumption of air outlet installation are effectively controlled, the labor cost and the repair rate are reduced, a threshold value is dynamically optimized based on historical data, the recognition precision of a complex area is improved, misjudgment of a difficult point area is solved, and the detection efficiency is improved. Double improvement of detection precision and efficiency is realized; meanwhile, through a three-level risk signal mechanism, conversion from passive detection to active early warning is realized, the risk response time is shortened, the product quality stability is guaranteed, the downtime is reduced, the overall efficiency of a production line is improved, and collaborative optimization of risk management and control and production continuity is realized.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of optical detection, in particular to an air outlet leakage and reverse installation detection method based on image detection. BACKGROUND

[0002] In the production and assembly of air conditioners, automobile air conditioners and the like, the air outlet is a core component, and the installation quality of the air outlet directly affects the air supply efficiency, sealing performance and service life of the equipment. The traditional detection method relies on manual visual inspection, and the installation state is judged by observing whether the buckle is tightly buckled and whether the direction mark is aligned or not. However, there are significant limitations: first, the efficiency is low, and it takes about 30 seconds to detect a single unit, which is difficult to meet the production line with a beat of 10 seconds per unit. Second, the precision is insufficient, and the human eye cannot identify the position deviation of 0.1 mm or the difference in curved surface fitting, resulting in about 5% of unqualified products flowing into the market.

[0003] With the application of image detection technology, some enterprises have introduced industrial cameras to realize automatic detection. However, the existing technology still has many problems: first, the universality is poor, the detection parameters are bound to specific models of air outlets, and the program needs to be re-adjusted when the model is changed, which takes up to 2 hours, and cannot adapt to the production mode of multiple varieties and small batches; second, the detection logic is fixed, and the identification precision of complex areas such as curved surface transitions is insufficient, and the misjudgment rate is high; third, the risk control is lagging behind, and only "qualified / unqualified" can be determined, and the risk level cannot be predicted based on historical fault data, resulting in high repair costs. Therefore, there is an urgent need for an air outlet detection method that can balance precision, efficiency and universality.

[0004] In view of the above technical defects, a solution is proposed. SUMMARY

[0005] The purpose of the present application is to provide an air outlet leakage and reverse installation detection method based on image detection to solve the problems.

[0006] To achieve the above purpose, the present application provides the following technical scheme: an air outlet leakage and reverse installation detection method based on image detection, comprising the following steps:

[0007] S1, determining the data acquisition range of image detection in the air outlet area, acquiring a plurality of image data of the installation state of the air outlet under the preset lighting condition through an industrial camera, and marking the image data as the original image data of the air outlet;

[0008] S2, jointly processing the historical record data and the original image data, dividing a conventional installation region set and an error-prone installation region set, and taking the historical data as a threshold reference source for subsequent detection;

[0009] S3, analyze the error-prone installation area, denoise, edge enhancement processing to the original image of the error-prone installation area, extract stable features and unstable features, obtain a feature recognition area set and a complex area set;

[0010] S4, obtain the scanning model data of the standard installation of the air outlet, compare the feature recognition area set, the complex area set and the model data one by one, generate an adaptive area signal and a non-adaptive area signal;

[0011] S5, retrieve the comparison data of the non-adaptive area, analyze the actual feature point cloud item, the model point cloud item and the deviation value item in combination with the historical fault cases, and generate a corresponding level risk signal according to the risk level.

[0012] Further, the S1 step of determining the image detection range and the original data acquisition process is as follows:

[0013] Determine the detection boundary by obtaining the design drawing of the air outlet, the boundary range covers the installation reference surface, all buckle positions and direction marks of the air outlet, use a ring-shaped LED light source to provide uniform illumination, obtain 5 frames of image data of the air outlet installation state collected by the camera in succession, remove the blurred frames, and keep 3 clear images and mark them as original image data.

[0014] Further, the historical data processing and area division process in the S2 step is as follows:

[0015] Retrieve the installation history of the air outlet within the last 6 months and consistent with the current detection equipment model and mark it as a historical data set, filter and keep the effective records in the historical data set, divide the original image data and the historical data set into grids, calculate the qualified rate of the installation area in the corresponding grid of the original image data and the historical data set, compare the pre-stored standard qualified rate threshold with the qualified rate of each area; mark the area with a qualified rate ≥ the qualified rate threshold as a regular installation area, and form a regular installation area set; mark the area with a qualified rate < the qualified rate threshold as an error-prone installation area, and form an error-prone installation area set.

[0016] Further, the analysis process of the error-prone installation area set in the S3 step is as follows:

[0017] Preprocess the original image of the error-prone installation area, retrieve the pre-stored Gaussian filter model to remove noise, and histogram equalization to enhance the contrast of the image and increase the clarity of the feature edges, extract the image edge area, identify the repeatedly appearing stable features, mark the area with stable features as a feature recognition area set, and mark the area with unstable features as a complex area set.

[0018] Further, the model data acquisition and joint analysis process in the S4 step is as follows:

[0019] The scanning model data of the standard installation state of the air outlet is acquired by a scanner. The model data includes the three-dimensional coordinates and characteristic parameters of each feature point corresponding to the original image data. Each feature in the feature recognition region set is compared with the model data to obtain the difference value of the three-dimensional coordinates and mark it as the position deviation value, and the difference value of the characteristic parameters and mark it as the parameter deviation value. The pre-stored position deviation threshold and parameter deviation threshold are called and compared with the position deviation value and parameter deviation value. When the position deviation value is within the range of the position deviation threshold, and the parameter deviation value is within the range of the parameter deviation threshold, an adaptive region signal is generated. When the position deviation value is outside the range of the position deviation threshold, or the parameter deviation value is outside the range of the parameter deviation threshold, a recheck signal is generated.

[0020] Further, when the recheck signal is generated, the complex region set is acquired, the actual installation and representative key feature points are extracted and marked as actual feature point cloud, the model data for subsequent installation and representative key feature points are acquired and marked as model point cloud, the pre-stored fault tolerance difference threshold is called, and the actual feature point cloud and the model point cloud are compared by transverse splicing. When the actual feature point cloud and the model point cloud have a calibration difference value and the calibration difference value is greater than the fault tolerance difference threshold, a non-adaptive region signal is generated. When the actual feature point cloud and the model point cloud have a calibration difference value and the calibration difference value is less than the fault tolerance difference threshold, an error marking signal is generated.

[0021] Further, the analysis process of the non-adaptive signal in the S5 step is as follows:

[0022] The comparison data of the non-adaptive region and the historical fault cases existing in the historical data set are called, and the fault reason extraction of the historical fault cases is performed, including the actual feature point cloud item, the model point cloud item and the deviation value item. The actual feature point cloud item average threshold, the model point cloud item average threshold and the deviation value item threshold of the non-fault case in the historical data set are acquired. The average value of the difference value of the actual feature point cloud item, the model point cloud item and the deviation value item of the historical fault case and the actual feature point cloud item average threshold, the model point cloud item average threshold and the deviation value item threshold is taken as the risk judgment threshold. The comparison data and the related data of the historical fault cases are brought into comparison.

[0023] Further, when the actual feature point cloud, the model point cloud and the deviation value exceed the actual feature point cloud item mean value threshold, the model point cloud item mean value threshold and the deviation value item threshold, but do not exceed the risk judgment threshold, a first-level risk signal is generated; when the actual feature point cloud, the model point cloud and the deviation value exceed the actual feature point cloud item mean value threshold, the model point cloud item mean value threshold and the deviation value item threshold, and the risk judgment threshold, but do not exceed the actual feature point cloud item, the model point cloud item and the deviation value item, a second-level risk signal is generated; when the actual feature point cloud, the model point cloud and the deviation value exceed the actual feature point cloud item, the model point cloud item and the deviation value item, a third-level risk signal is generated.

[0024] The beneficial effects of the present application are:

[0025] 1. The present application effectively controls the unqualified rate of air outlet installation and detection time consumption by dividing the conventional / easy-to-mistake area and adopting a differentiated detection strategy combined with three-dimensional point cloud comparison technology, greatly reduces the labor cost and the repair rate, and dynamically optimizes the threshold based on historical data to improve the recognition accuracy of complex areas, solves the misjudgment problem of difficult areas such as curved surface transitions, and realizes the dual improvement of detection accuracy and efficiency.

[0026] 2. The present application realizes the transformation from passive detection to active early warning through a three-level risk signal mechanism, the first-level risk identifies slight deviation in advance, the second-level risk intercepts moderate problems in time, and the third-level risk handles serious faults in an emergency, shortens the risk response time, not only guarantees the stability of product quality, but also improves the overall efficiency of the production line by reducing downtime and reducing the repair rate, realizes the win-win of quality control and production benefit, and realizes the collaborative optimization of risk control and production continuity. BRIEF DESCRIPTION OF DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, a brief introduction will be given below to the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0028] Figure 1 The present application is a method flowchart. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0030] Embodiment one: please refer to Figure 1 As shown, the embodiment is an air outlet leakage and reverse installation detection method based on image detection, including the following steps:

[0031] S1, determine the data acquisition range of image detection in the air outlet area, determine the detection area according to the structure parameters of the air outlet, the structure parameters include size, installation reference line, acquire a plurality of image data of the air outlet installation state under the preset lighting condition through the industrial camera, and mark it as the original image data of the air outlet; The original image data of the air outlet contains RGB three color channels and depth information, the resolution is not less than 1920x1080, and the frame rate is 25fps; The determination of the image detection range and the original data acquisition process in the S1 step are as follows:

[0032] Determine the detection boundary by obtaining the design drawing of the air outlet, and the boundary range covers the installation reference surface, all buckle positions and direction marks of the air outlet, so as to ensure that there is no detection blind area; Adopt ring-shaped LED light source to provide uniform illumination, acquire 5 frames of image data of the air outlet installation state continuously collected by the camera, remove the blurred frames, adjust the blurred frame threshold according to the pre-stored standard frame threshold in the historical data set, remove all blurred frames less than the standard frame threshold, and keep 3 clear images and mark them as original image data; It should be noted that the industrial camera parameters are configured: the lens focal length is set to 16mm, the working distance is kept at 500±50mm; The light intensity is controlled at 800-1200lux; The original image data format is BMP, the storage path is associated with the equipment model and the time stamp of the detection collection, and the standard frame threshold: used to judge whether the image is blurred, which can be taken as 85 points, based on the definition score algorithm, the full score is 100 points, example: if the definition score of a certain frame image is 82 points, which is lower than 85 points, it is determined as a blurred frame and is removed; If the definition score of a certain frame image is 90 points, which is higher than 85 points, the original image data is kept.

[0033] S2, joint processing of historical record data and original image data, divide the conventional installation area set and the error-prone installation area set, the historical data is used as the threshold reference source for subsequent detection, and the influence of reflection on image quality is avoided; The historical data processing and area division process in the S2 step are as follows:

[0034] Call the air outlet installation history record consistent with the current detection equipment model within nearly 6 months and mark it as a historical data set, the historical data set includes detection time, installation area coordinates, qualified / unqualified judgment results and fault types, filter and retain valid records in the historical data set, sample size ≥ 500, grid division is performed on the original image data and the historical data set, and the qualified rate of the installation area in the corresponding grid of the original image data and the historical data set is calculated: qualified rate = (qualified record number / total record number) × 100%, compare the pre-stored standard qualified rate threshold with the qualified rate of each area, it should be noted that the standard qualified rate threshold: the value can be 95%, used to divide the normal and error-prone areas, for example: in the historical detection of a certain buckle position, there are 480 qualified records, a total of 500 records, and the qualified rate is 96%, which is ≥ 95%, and is divided into a normal installation area; the qualified record of a certain curved transition area is 420, the total record is 500, and the qualified rate is 84%, which is < 95%), and is divided into an error-prone installation area;

[0035] Regions with a qualified rate ≥ the qualified rate threshold are marked as normal installation areas, and a normal installation area set is formed, and the subsequent detection frequency of the region can be reduced to once every 3 frames of images;

[0036] Regions with a qualified rate < the qualified rate threshold are marked as error-prone installation areas, and an error-prone installation area set is formed, and subsequent detection is required for each frame of image, and the detection qualified rate threshold is increased by 20%;

[0037] The historical data set is automatically updated every 24 hours, and the new detection records are included in the database after being manually reviewed and confirmed; when the qualified rate of a certain installation area is ≥ the standard qualified rate threshold for 30 consecutive days, it is adjusted from the error-prone installation area set to the normal installation area set; if the qualified rate is < the standard qualified rate threshold for 5 consecutive days, it is adjusted from the normal installation area set to the error-prone installation area set, to ensure the dynamic adaptability of the region division.

[0038] S3, analyze the error-prone installation area, perform denoising and edge enhancement processing on the original image of the error-prone installation area, extract stable features and unstable features, and obtain a feature recognition area set and a complex area set; the analysis process of the S3 step on the error-prone installation area set is as follows:

[0039] The original image of the error-prone installation area is acquired for preprocessing, a pre-stored Gaussian filter model is called to remove noise, the filter kernel size is 3*3, that is, the neighborhood pixels within the range of 3*3 around each pixel point participate in calculation, a corresponding weight matrix is generated according to the Gaussian function, the pixel weight closer to the center of the matrix is greater, which conforms to the Gaussian distribution law; then the gray value of each pixel point in the original image is convolved with the weight matrix, that is, the new gray value of each pixel is equal to the product of its own gray value and the corresponding weight plus the sum of the product of the gray value of the neighborhood pixels and the corresponding weight, the image is smoothed by this weighted average method, and the interference of random noise on subsequent feature extraction is reduced;

[0040] And the histogram equalization enhances the contrast of the image and increases the edge definition of the features, and the specific process is as follows: the gray histogram of the original image data is counted, that is, the number of pixels corresponding to each gray level (0-255) is calculated; then the cumulative distribution function (CDF) is calculated, which reflects the proportion of pixels less than or equal to a certain gray level in the total pixels; finally, the gray value is mapped and converted according to the cumulative distribution function, the gray interval with dense distribution in the original image is stretched to a wider range, the dark part is darker and the bright part is brighter, so as to enhance the contrast between features and background in the image and highlight the edge details;

[0041] The edge region of the image is extracted, and the repeatedly appearing judgment is identified as a stable feature, such as a circular buckle hole with a diameter of 3±0.5mm and a directional scale line with a length of 10±1mm, the region where the stable feature is located is marked as a feature recognition region set, the edge blur region with surface transition and multi-feature overlapping region in the extracted edge region is extracted, which is easy to be affected by light, and is determined as an unstable feature, and the region of the unstable feature is marked as a complex region set.

[0042] S4, the scanning model data of the standard installation of the air outlet is acquired, including three-dimensional coordinates and feature parameters, the feature recognition region set, the complex region set and the model data are compared one by one to generate an adaptive region signal and a non-adaptive region signal; the acquisition and joint analysis process of the model data in the S4 step is as follows:

[0043] The scanning model data of the standard installation state of the air outlet is obtained by a 3D scanner. The model data includes three-dimensional coordinates and characteristic parameters of each feature point corresponding to the original image data. The three-dimensional coordinates include X, Y, and Z axis accuracy of ±0.01 mm. The characteristic parameters include the diameter of the buckle hole and the angle of the scale line. Each feature in the feature recognition area set is compared with the model data to obtain the difference value of the three-dimensional coordinates and mark it as the position deviation value, such as ΔX, ΔY, and ΔZ. The difference value of the characteristic parameters is marked as the parameter deviation value, such as the diameter deviation ΔD. The pre-stored position deviation threshold and parameter deviation threshold are retrieved and compared with the position deviation value and parameter deviation value. It should be noted that the position deviation threshold can be ±0.1 mm, based on the three-dimensional coordinate accuracy of ±0.01 mm. For example, the standard X coordinate of a certain buckle hole is 100.00 mm, and the actual detection is 100.08 mm, with a deviation of 0.08 mm≤0.1 mm, which meets the threshold requirement. If the actual value is 100.12 mm, the deviation is 0.12 mm>0.1 mm, which exceeds the threshold. The parameter deviation threshold can be 5%, based on the characteristic parameter. For example, the standard length of a certain direction scale line is 10 mm, and the actual detection is 10.4 mm, with a deviation of 4%≤5%, which meets the requirement. If the actual value is 10.6 mm, the deviation is 6%>5%, which exceeds the threshold.

[0044] When the position deviation value is within the range of the position deviation threshold, and the parameter deviation value is within the range of the parameter deviation threshold, an adaptive region signal is generated, indicating that the installation of the region meets the standard. The region qualified information is automatically recorded, and the conveyor belt is triggered to send the air outlet to the next process. At the same time, the historical data set is updated, and the qualified record is accumulated to optimize the region division threshold. Emergency measures: if the adaptive signal is continuous for 3 times, but the downstream process feedback is abnormal, the re-inspection program is automatically started.

[0045] When the position deviation value is outside the range of the position deviation threshold, or the parameter deviation value is outside the range of the parameter deviation threshold, a re-inspection signal is generated, indicating that the feature recognition area has deviation and needs to be confirmed again. The industrial camera is controlled to re-acquire 3 frames of images of the region. Higher precision algorithms such as sub-pixel level edge detection are used to re-calculate the position deviation value and the parameter deviation value. If the threshold is still exceeded, the system automatically switches to the complex region point cloud comparison mode. Emergency measures: if the re-inspection times are ≥3 times and still abnormal, the audible and light alarm is triggered to prompt the operator to check the camera focal length or light source stability.

[0046] In generating the re-inspection signal, the complex region set is obtained, the actual installation and representative key feature points are extracted and marked as actual feature point cloud, the key feature points in the model data for subsequent installation and representative are obtained and marked as model point cloud, the pre-stored tolerance difference threshold is called, and the actual feature point cloud and the model point cloud are transversely spliced and compared. It should be noted that the tolerance difference threshold: the value can be 0.08mm, which is used for complex region point cloud comparison. For example, the calibration difference between the actual feature point cloud and the model point cloud is 0.07mm, 0.07mm<0.08mm, and it is determined as acceptable error; if the difference is 0.09mm, 0.09mm>0.08mm, it is determined as not suitable:

[0047] When the actual feature point cloud and the model point cloud have a calibration difference and the calibration difference is greater than the tolerance difference threshold, a misfit region signal is generated, indicating that the installation deviation of the complex region is significant, the production line is immediately suspended, the three-dimensional coordinates and the deviation value of the misfit region are marked on the display screen, the control execution mechanism such as the mechanical arm is pushed to the return repair station, and the return repair guide is generated, such as "the buckle hole position deviation is 0.15mm, it is suggested to adjust leftward", which is displayed in the form of text on the display of the supervisor; if the same type of outlet appears 5 times in succession within 1 hour, the equipment is automatically locked, and it is prompted to check whether the scanning model data matches the current product type;

[0048] When the actual feature point cloud and the model point cloud have a calibration difference and the calibration difference is less than the tolerance difference threshold, an error marking signal is generated, indicating that there is a slight deviation in the complex region but it is acceptable, the region deviation value is marked in the detection report, such as "the curved surface fitting deviation is 0.06mm" in the form of text, and the data is synchronized to the production management system for statistical analysis of the common error distribution of the same type of outlet. In subsequent detection, the attention to this region is increased, such as increasing the number of feature points extracted by 10%, and if the error marking signal accounts for ≥30% of the same batch of products, the operator is prompted to fine-tune the positioning reference of the installation tooling.

[0049] Embodiment two

[0050] S5, call the comparison data of the misfit region, analyze the actual feature point cloud item, the model point cloud item and the deviation value item in combination with historical fault cases, and generate corresponding level risk signals according to the risk level. The analysis process of the misfit signal in the S5 step is as follows:

[0051] The comparison data of the mismatched area and the historical fault cases existing in the historical data set are called, and the fault cause extraction is performed on the historical fault cases, including the actual feature point cloud item, the model point cloud item and the deviation value item, the mean value threshold of the actual feature point cloud item, the model point cloud item and the deviation value item in the historical data set is obtained, and the difference average value of the actual feature point cloud item, the model point cloud item and the deviation value item of the historical fault case and the mean value threshold of the actual feature point cloud item, the model point cloud item and the deviation value item is taken as the risk judgment threshold. The related data of the comparison data and the historical fault cases are substituted into the comparison. It should be noted that the mean value threshold of the actual feature point cloud item: taking the X coordinate of a certain curved surface feature point as an example, the value can be 50.00±0.05mm, based on the historical non-fault case statistics, example: the actual detection X coordinate is 50.06mm, which exceeds the mean value threshold, and needs to be further compared with the risk judgment threshold;

[0052] When the actual feature point cloud, the model point cloud and the deviation value exceed the mean value threshold of the actual feature point cloud item, the mean value threshold of the model point cloud item and the threshold of the deviation value item, but do not exceed the risk judgment threshold, a first-level risk signal is generated, indicating that there is a slight risk, a yellow warning icon is displayed on the terminal, and the risk area and the deviation value are marked, such as the "direction marking line angle deviation 3°" style text. The production is not suspended but the risk information is recorded; when the first-level risk signal of the same air outlet is ≥2, it is automatically upgraded to the second-level risk; the operator summarizes the first-level risk signal every hour to analyze whether there is a common deviation, such as size fluctuation of a batch of parts;

[0053] When the actual feature point cloud, the model point cloud and the deviation value exceed the mean value threshold of the actual feature point cloud item, the mean value threshold of the model point cloud item and the threshold of the deviation value item, and the risk judgment threshold, but do not exceed the actual feature point cloud item, the model point cloud item and the deviation value item, a second-level risk signal is generated, indicating that there is a moderate risk, triggering an orange alarm, suspending the circulation of the air outlet, and prompting the operator to manually review; after the review is qualified, "risk release" needs to be manually confirmed to continue production; if the review is unqualified, it will be transferred to the repair process; if three second-level risk signals appear continuously, the historical fault cases are automatically called, and the solutions of similar faults are pushed, such as the "reference 20240512 case, adjust the buckle installation force" style text;

[0054] When the actual feature point cloud, the model point cloud and the deviation value exceed the actual feature point cloud item, the model point cloud item and the deviation value item, a third-level risk signal is generated, indicating that there is a serious risk, such as missing installation or installation in reverse, starting a red emergency alarm, cutting off the power supply of the production line, and sending warning information to the manager's mobile phone at the same time, including fault area pictures and deviation data; after manual troubleshooting and replacement of qualified air outlets, the administrator password needs to be input to restart the equipment. If two third-level risk signals appear within 24 hours, the quality traceability program is started, and the material batch and installation process parameters are checked.

[0055] In combination with Embodiment One and Embodiment Two, the present application effectively controls the unqualified product rate and detection time of the air outlet installation by dividing the conventional and error-prone areas, adopting a differentiated detection strategy and combining three-dimensional point cloud comparison technology, reduces the labor cost and rework rate, dynamically optimizes the threshold based on historical data, improves the identification accuracy of complex areas, solves the misjudgment of difficult areas, realizes the dual improvement of detection accuracy and efficiency; at the same time, through the three-level risk signal mechanism, the transformation from passive detection to active early warning is realized, the risk response time is shortened, the product quality stability is guaranteed, the downtime is reduced, the overall efficiency of the production line is improved, and the risk control and production continuity are optimized.

[0056] The above is only an example and description of the structure of the present application, and those skilled in the art can make various modifications or supplements to the described specific examples or replace them with similar ways, as long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, which shall be within the protection scope of the present application.

[0057] In the description of the present application, the description of the terms "one embodiment", "example", "specific example" and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner, and the relevant fittings include couplings, lead screws, gears, gaskets and other commonly used mechanical connecting parts in the field, and are not limited thereto. The connecting mode is replaced and used according to the actual use.

[0058] The preferred embodiments of the present application disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details and limit the present application to the specific embodiments. Obviously, many modifications and changes can be made according to the content of the present application. The present application selects and describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can well understand and utilize the present application. The present application is limited by the claims and their entire scope and equivalents.

Claims

1. An air outlet leakage and reverse installation detection method based on image detection, characterized in that, Comprise the following steps; S1, determine the image detection in the data acquisition range of the air outlet area, acquire several image data of the air outlet installation state under the preset lighting condition through the industrial camera, and mark it as the original image data of the air outlet; S2, acquire historical record data and original image data joint processing, divide the conventional installation area set and the error prone installation area set, and the historical data is used as the threshold reference source for subsequent detection; S3, analyze the error prone installation area, denoise and edge enhancement processing are carried out on the original image of the error prone installation area, stable features and unstable features are extracted, and the feature recognition area set and the complex area set are obtained; S4, acquire the scanning model data of the standard installation of the air outlet (including three-dimensional coordinates and feature parameters), compare the feature recognition area set, the complex area set and the model data one by one, and generate the adaptive area signal and the non adaptive area signal; S5, call the comparison data of the non adaptive area, analyze the actual feature point cloud item, the model point cloud item and the deviation value item combined with the historical fault case, and generate the corresponding level risk signal according to the risk level.

2. The air outlet leakage and reverse installation detection method based on image detection according to claim 1, characterized in that, The determination of the image detection range and the acquisition of the original data in the S1 step are as follows: The design drawing of the air outlet is acquired to determine the detection boundary, the boundary range covers the installation reference surface, all buckle positions and direction marks of the air outlet, uniform illumination is provided by adopting ring LED light source, 5 frames of image data of the air outlet installation state are continuously acquired by the camera, the blurred frames are removed, 3 clear images are reserved and marked as original image data.

3. The air outlet leakage and reverse installation detection method based on image detection according to claim 1, characterized in that, The historical data processing and area division process in the S2 step are as follows: The installation history records of the air outlet consistent with the current detection equipment model in the past 6 months are called and marked as historical data set, the effective records in the historical data set are screened and reserved, the original image data and the historical data set are subjected to grid division, the qualified rate of the installation area in the corresponding grid of the original image data and the historical data set is counted, the pre stored standard qualified rate threshold is compared with the qualified rate of each area; the area with qualified rate greater than or equal to the threshold is marked as the conventional installation area, and the conventional installation area set is formed, the subsequent detection frequency of this area can be reduced to once every 3 frames of image detection; the area with qualified rate less than the threshold is marked as the error prone installation area, and the error prone installation area set is formed, subsequent detection is required for every frame of image, and the detection accuracy threshold is increased by 20%.

4. The air outlet leakage and reverse installation detection method based on image detection according to claim 1, characterized in that, The analysis process of the error prone installation area set in the S3 step is as follows: The original image of the error prone installation area is preprocessed, the pre stored Gaussian filter model is called to remove noise, and the histogram equalization is enhanced to increase the contrast of the image and the clarity of the feature edge, the edge region of the image is extracted, the repeatedly appearing judgment is recognized as stable features, such as circular buckle hole with diameter of 3±0.5mm and direction scale line with length of 10±1mm, the area where the stable features are located is marked as the feature recognition area set, the edge blur area where there are curved transition and multi feature overlapping area in the extracted edge region is judged as unstable features, and the area of the unstable features is marked as the complex area set.

5. The air outlet leakage and reverse installation detection method based on image detection according to claim 1, characterized in that, The acquisition and joint analysis process of the model data in the S4 step is as follows: The scanning model data of the standard installation state of the air outlet is acquired by a scanner, the model data includes the three-dimensional coordinates and characteristic parameters of each feature point corresponding to the original image data, each feature in the feature recognition area set is compared with the model data, the difference value of the three-dimensional coordinates is obtained and marked as the position deviation value, the difference value of the characteristic parameters is obtained and marked as the parameter deviation value, and the pre-stored position deviation threshold and parameter deviation threshold are called and compared with the position deviation value and the parameter deviation value; when the position deviation value is within the range of the position deviation threshold, and the parameter deviation value is within the range of the parameter deviation threshold, an adaptive region signal is generated; when the position deviation value is outside the range of the position deviation threshold, or the parameter deviation value is outside the range of the parameter deviation threshold, a recheck signal is generated.

6. The air outlet leakage and reverse installation detection method based on image detection according to claim 5, characterized in that, When the recheck signal is generated, the complex region set is acquired, the actual installation and representative key feature points are extracted and marked as actual feature point cloud, the key feature points in the model data for subsequent installation and representative are acquired and marked as model point cloud, the pre-stored fault tolerance difference threshold is called, and the actual feature point cloud and the model point cloud are transversely spliced and compared: when the actual feature point cloud and the model point cloud exist calibration difference value and are greater than the fault tolerance difference threshold, a non-adaptive region signal is generated; when the actual feature point cloud and the model point cloud exist calibration difference value and are less than the fault tolerance difference threshold, an error marking signal is generated.

7. The air outlet leakage and reverse installation detection method based on image detection according to claim 1, characterized in that, The analysis process of the non-adaptive signal in the S5 step is as follows: The comparison data of the non-adaptive region and the historical fault cases existing in the historical data set are called, and the fault causes of the historical fault cases are extracted, including the actual feature point cloud item, the model point cloud item and the deviation value item, the actual feature point cloud item average threshold, the model point cloud item average threshold and the deviation value item threshold in the historical data set are acquired, the average value of the difference between the actual feature point cloud item, the model point cloud item and the deviation value item of the historical fault cases and the actual feature point cloud item average threshold, the model point cloud item average threshold and the deviation value item threshold is taken as the risk judgment threshold, and the related data of the comparison data and the historical fault cases are brought into comparison.

8. The air outlet leakage and reverse installation detection method based on image detection according to claim 7, characterized in that, When the actual feature point cloud, the model point cloud and the deviation value exceed the actual feature point cloud item average threshold, the model point cloud item average threshold and the deviation value item threshold, but do not exceed the risk judgment threshold, a first-level risk signal is generated; when the actual feature point cloud, the model point cloud and the deviation value exceed the actual feature point cloud item average threshold, the model point cloud item average threshold and the deviation value item threshold, and the risk judgment threshold, but do not exceed the actual feature point cloud item, the model point cloud item and the deviation value item, a second-level risk signal is generated; When the actual feature point cloud, the model point cloud and the deviation value exceed the actual feature point cloud item, the model point cloud item and the deviation value item, a third-level risk signal is generated.

Images