Truss laminated slab production quality inspection method

Abstract

The invention relates to the technical field of building material detection, and discloses a truss laminated slab production quality inspection method which comprises the following steps: firstly, translating a truss laminated slab into a shooting range of industrial cameras arranged in a linear array through a conveying platform; then, images of the sides, with galling, of the multiple truss laminated slabs are shot through an industrial camera; the truss laminated slab enters a sorting platform after being shot; the image of the truss laminated slab is transmitted to the upper computer to complete quality inspection, the upper computer sends a quality inspection result to the controller in the sorting platform, the controller controls the sorting platform to act, and the truss laminated slab with the galling area meeting the proportion requirement is sorted to a storage place of qualified products. And the truss laminated slabs with the galling areas not meeting the proportion requirement are sorted into the unqualified channels, and through the quality inspection method, the galling area occupation on the truss laminated slabs can be rapidly recognized under unmanned monitoring, and the quality inspection result is output.

Description

Technical Field

[0001] The invention relates to the technical field of building material detection, and in particular to a production quality inspection method for truss composite plates. Background Art

[0002] Truss laminate is a prefabricated component widely used in the construction industry. Its quality is directly related to the safety and durability of the building. Therefore, quality monitoring of the entire production process of truss laminate is very important.

[0003] According to the Technical Code for Prefabricated Concrete Structures JGJ1-2014, the rough surface area of ​​the truss laminate should not be less than 80% of the bonding surface. And in the Acceptance Code for Construction Quality of Concrete Structure Engineering GB50204-2015, the proportion of the rough surface of the truss laminate is listed as a quality inspection item for the production of truss laminates. The rough surface of the truss laminate is usually roughened by roughening when the truss laminate is prefabricated, leaving grooves on one side of the truss laminate to form a rough surface.

[0004] At present, the conventional method for detecting the rough surface of the truss composite plate is to use the sand-laying method, which is to spread sand of a certain particle size on the overlapping surface of the truss composite plate, and then measure the sand-laying area to estimate the roughening area. Although this method is simple, it is complicated to operate, slow, and not suitable for on-site measurement. Another method is to use the laser scanning method, which uses three-dimensional laser scanning technology to obtain detailed information on the surface of the truss composite plate, and calculate the depth and area of ​​the roughening through a formula. Although this method has high accuracy, the equipment cost is high and the operator's skills are required to be high. Therefore, how to achieve unmanned monitoring and quickly identify the proportion of the roughening area on the truss composite plate, no applicable method has been found so far. Summary of the invention

[0005] The present invention aims to provide a truss composite plate production quality inspection method to solve the problem of being able to quickly identify the roughening area on the truss composite plate and output the quality inspection results without human monitoring.

[0006] To achieve the above object, the present invention adopts a truss laminated plate production quality inspection method, comprising:

[0007] S1: The truss stacking plate is translated into the shooting range of the industrial camera arranged in a linear array through a conveying platform;

[0008] S2: Use an industrial camera to take multiple images of the roughened side of the truss composite plate;

[0009] S3: After the truss composite plate is photographed, it enters a sorting platform;

[0010] S4: The image of the truss composite plate is transmitted to the host computer for quality inspection, and the host computer performs the following steps:

[0011] S41: Preprocessing the image by graying, removing noise and enhancing contrast;

[0012] S42: The Canny edge detection algorithm is used to perform secondary processing on the image. First, the Gaussian filter is used to smooth the image again. Then, the gradient strength and direction of the image are calculated. The local maximum value is retained through the non-maximum suppression operation to form the edge. Then, the low threshold is 50 and the high threshold is 150 to determine the edge.

[0013] S43: The edge detection result is combined with the calibration parameters of the camera to convert the pixel size into the actual size;

[0014] S44: Extracting the rough area boundary and rough texture features in the image through the trained ResNet-50 model;

[0015] S45: The extracted roughness feature value is compared with the corresponding value in the standard feature value library;

[0016] S46: After the feature comparison is completed, the number of pixels in the roughened area is calculated by using a pixel statistics method, and the number of pixels is converted into an actual area, and then the ratio of the roughened area to the total area is calculated;

[0017] S47: The system outputs the ratio of the roughened area to the dimension area, compares the ratio with the quality inspection qualified value, and outputs the quality inspection result of whether the inspected truss composite plate is qualified;

[0018] S5: The upper computer sends the quality inspection results to the controller in the sorting platform. The controller controls the action of the sorting platform to sort the truss composite panels whose roughened areas meet the requirements to the storage area for qualified products, and sort the truss composite panels whose roughened areas do not meet the requirements to the unqualified channel.

[0019] The principles and advantages of this solution are: in actual application, the truss composite plates that need to be inspected are gradually transmitted to the shooting range of the industrial camera through the transmission platform, and the image of the truss composite plate with the roughened side is captured by the industrial camera. After obtaining the image, the image is first pre-processed, and then the trained ResNet-50 model is used to identify the roughened features of the truss composite plate. After completing the feature comparison, the pixel statistics method is used to calculate the number of pixels in the roughened area and convert it into the actual area. The proportion of the roughened area is compared with the standard value (80%) to determine whether the roughened area is qualified. Finally, the system outputs the ratio of the roughened area to the size area. If it is ≥80%, it will output "qualified", and if it is <80%, it will output "unqualified".

[0020] Preferably, in step S1, the surface of the truss laminated plate is cleaned by a purge device, the conveying platform includes a base, a conveyor belt for conveying the truss laminated plate is installed on the base, and the purge device includes a door frame erected above the conveyor belt, and a purge fan with an adjustable purge angle is installed on the door frame. The residue on the surface of the truss laminated plate is cleared by the purge fan to avoid interference with taking photos.

[0021] Preferably, in order to obtain a high-definition image of the truss composite plate, in step S1, the industrial camera uses a camera mechanism to automatically adjust the shooting angle and fill light, and the camera mechanism includes a camera suspension beam and a light source suspension beam, and the light source suspension beam is an adjustable rotating rod, and an anti-shake gimbal is installed on the camera suspension beam, and the anti-shake gimbal is used to fix the industrial camera. The industrial camera cooperates with the fill light of the light source so that the camera shooting can clearly capture the surface details of the truss composite plate. In addition, an anti-shake gimbal is installed on the camera suspension beam to solve the problem of camera shaking caused by noise and mechanical vibration in the factory environment.

[0022] Preferably, a photoelectric sensor for detecting whether the truss stacking plate enters the shooting range is fixed on the conveying platform. Through the detection of the photoelectric sensor, the conditions for starting and stopping the camera shooting can be accurately controlled.

[0023] Preferably, in order to facilitate the distinction of the truss laminated plates after quality inspection, the sorting platform includes a supporting truss laminated plate placement platform, a hoisting gantry for transferring the truss laminated plates is fixedly installed above the placement platform, a mechanical claw that can simultaneously grab at least four grabbing points of the truss laminated plates is arranged on the gantry, and a controllable drip for marking is also arranged above the sorting platform, and the drip is filled with paint for marking the truss laminated plates. If the truss laminated plate is judged to be qualified, it is marked by controlling the drip to drip green, and if it is judged to be unqualified, it is marked by dripping red.

[0024] Preferably, in order to better complete the recognition of the image, in step S41, the noise removal method adopts Gaussian filtering, and the function of Gaussian filtering is Parameter σ = 1.0;

[0025] Preferably, in step S41, image contrast enhancement is achieved through histogram equalization, which includes calculating the histogram of the image, constructing a cumulative distribution function, and mapping the image pixel values ​​using the cumulative distribution function to make the grayscale values ​​of the image evenly distributed, thereby enhancing the rough texture in the image.

[0026] Preferably, a plurality of contrast-enhanced images of different sizes are synthesized using a multi-scale Retinex algorithm.

[0027] Preferably, in step S44, the training process of the ResNet-50 model includes: collecting surface images of truss composite plates under different lighting and shooting angles, and transforming the images by angle and size through a distributed algorithm to form no less than 1,000 training samples, using the LabelImg tool to mark the roughness in the truss composite plate images, and normalizing the images to meet the input requirements of the ResNet-50 model; after iterative training, randomly collecting no less than 500 uninspected truss composite plates, and continuously adjusting the hyperparameters of the initial ResNet-50 model, and completing the model training when the recognition accuracy reaches more than 95%.

[0028] Preferably, after step S47, when the system completes the identification of the roughening ratio of the truss composite plate, the system calibrates the roughening features of the truss composite plate, stores the roughening features in the truss composite plate with a qualified roughening area ratio into a qualified roughening feature library, trains the truss composite plate with an unqualified roughening area ratio through the defect detection model of Faster R-CNN, classifies the roughening defects of the unqualified truss composite plate, and outputs the unqualified roughening type in the host computer. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a schematic diagram of the hardware structure used for the production quality inspection of truss composite plates in an embodiment of the present invention.

[0030] Figure 2 Flow chart of the quality inspection process steps in an embodiment of the present invention.

[0031] Figure 3 This is a flowchart of determining the roughness ratio performed by the host computer in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following is further described in detail through specific implementation methods:

[0033] The figure marks in the drawings of the specification include: conveying platform 1, truss composite plate 2, purge fan 3, camera mechanism 4, camera suspension beam 5, light source suspension beam 6, photoelectric sensor 7, drip 8, mechanical claw 9, sorting platform 10.

[0034] Example 1

[0035] As attached Figure 1As shown: the truss laminated plate 2 to be inspected is sent to the conveying platform 1, and the conveying platform 1 drives the truss laminated plate 2 to move horizontally into the shooting range of the industrial camera arranged in a linear array; the conveying platform includes a base, on which a conveyor belt for transmitting the truss laminated plate is installed, and the purge device includes a door frame erected above the conveyor belt, on which a purge fan 3 with an adjustable purge angle is installed. Before entering the shooting range, the surface of the truss laminated plate is cleaned by the purge fan 3, and the residue on the surface of the truss laminated plate 2 is removed by the purge fan 3 to avoid interference with the shooting of photos.

[0036] The conveying platform 2 is provided with a camera mechanism 4, which includes a camera suspension beam 5 and a light source suspension beam 6. The light source suspension beam 6 is an adjustable rotating rod. An anti-shake gimbal is installed on the camera suspension beam 5. The anti-shake gimbal is used to fix the industrial camera. The industrial camera cooperates with the fill light of the light source so that the camera can clearly capture the surface details of the truss composite plate. In addition, an anti-shake gimbal is installed on the camera suspension beam to solve the problem of camera shaking caused by noise and mechanical vibration in the factory environment.

[0037] A photoelectric sensor 7 is fixed on the conveying platform 1 for detecting whether the truss composite plate 2 enters the shooting range. Through the detection of the photoelectric sensor 7, the conditions for starting and stopping the camera shooting can be accurately controlled.

[0038] A sorting platform 10 is arranged behind the conveying platform 1. The sorting platform 10 includes a supporting truss laminate placement platform. A hoisting gantry for transferring truss laminates is fixedly installed above the placement platform. A mechanical claw 9 that can simultaneously grab at least four grabbing points of the truss laminate is arranged on the gantry. A controllable drip 8 for marking is also arranged above the sorting platform 10. The drip 8 is filled with paint for marking truss laminates. If the truss laminate is judged to be qualified, it is marked by controlling the drip to drip green. If it is judged to be unqualified, it is marked by dripping red.

[0039] Example 2

[0040] As attached Figure 2 and Figure 3 A truss composite plate production quality inspection method is shown, comprising:

[0041] S1: The truss stacking plate is translated into the shooting range of the industrial camera arranged in a linear array through a conveying platform;

[0042] S2: Use an industrial camera to take multiple images of the roughened side of the truss composite plate;

[0043] S3: After the truss composite plate is photographed, it enters a sorting platform;

[0044] S4: The image of the truss composite plate is transmitted to the host computer for quality inspection. The host computer performs the following steps:

[0045] S41: Preprocess the image by graying, removing noise and enhancing contrast; Gaussian filtering is used for noise removal, and the function of Gaussian filtering is: The parameter σ = 1.0. The image contrast is enhanced by histogram equalization, which includes calculating the image histogram, constructing the cumulative distribution function, and mapping the image pixel values ​​using the cumulative distribution function to make the grayscale value of the image evenly distributed, enhance the rough texture in the image, and synthesize multiple images of different sizes with contrast enhancement using the multi-scale Retinex algorithm.

[0046] S42: The Canny edge detection algorithm is used to perform secondary processing on the image. First, the Gaussian filter is used to smooth the image again. Then, the gradient strength and direction of the image are calculated. The local maximum value is retained through the non-maximum suppression operation to form the edge. Then, the low threshold is 50 and the high threshold is 150 to determine the edge.

[0047] S43: The edge detection result is combined with the calibration parameters of the camera to convert the pixel size into the actual size;

[0048] S44: Extract the rough area boundary and rough texture features in the image through the trained ResNet-50 model.

[0049] The training process of the ResNet-50 model includes: collecting surface images of truss composite plates under different lighting and shooting angles, and transforming the images by angle and size through a distributed algorithm to form no less than 1,000 training samples, using the LabelImg tool to mark the roughness in the truss composite plate images, and normalizing the images to meet the input requirements of the ResNet-50 model; after iterative training, randomly collecting no less than 500 uninspected truss composite plates, and continuously adjusting the hyperparameters of the initial ResNet-50 model. When the recognition accuracy reaches more than 95%, the model training is completed.

[0050] After the system completes the identification of the roughness ratio of the truss composite plate, the system will calibrate the roughness features of the truss composite plate, store the roughness features of the truss composite plate with a qualified roughness area ratio into the qualified roughness feature library, train the truss composite plate with an unqualified roughness area ratio through the Faster R-CNN defect detection model, classify the roughness defects of the unqualified truss composite plate, and output the unqualified roughness type in the host computer.

[0051] S45: The extracted roughness feature value is compared with the corresponding value in the standard feature value library;

[0052] S46: After the feature comparison is completed, the number of pixels in the roughened area is calculated by using a pixel statistics method, and the number of pixels is converted into an actual area, and then the ratio of the roughened area to the total area is calculated;

[0053] S47: The system outputs the ratio of the roughened area to the dimension area, compares the ratio with the quality inspection qualified value, and outputs the quality inspection result of whether the inspected truss composite plate is qualified;

[0054] S5: The upper computer sends the quality inspection results to the controller in the sorting platform. The controller controls the action of the sorting platform to sort the truss composite panels whose roughened areas meet the requirements to the storage area for qualified products, and sort the truss composite panels whose roughened areas do not meet the requirements to the unqualified channel.

[0055] The above is only an embodiment of the present invention, and the common knowledge such as the known specific technical solutions and / or characteristics in the solution is not described in detail here. It should be pointed out that for those skilled in the art, without departing from the technical solution of the present invention, several modifications and improvements can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A truss composite plate production quality inspection method, characterized in that: The following steps are included: S1: The truss stacking plate is translated into the shooting range of the industrial camera arranged in a linear array through a conveying platform; S2: Use an industrial camera to take multiple images of the roughened side of the truss composite plate; S3: After the truss composite plate is photographed, it enters a sorting platform; S4: The image of the truss composite plate is transmitted to the host computer for quality inspection, and the host computer performs the following steps: S41: Preprocessing the image by graying, removing noise and enhancing contrast; S42: The Canny edge detection algorithm is used to process the image twice. First, the Gaussian filter is used to smooth the image again, and then the gradient strength and direction of the image are calculated. The local maximum value is retained through the non-maximum suppression operation to form the edge; then the low threshold is 50 and the high threshold is 150 to determine the edge. S43: The edge detection result is combined with the calibration parameters of the camera to convert the pixel size into the actual size; S44: Extracting the rough area boundary and rough texture features in the image through the trained ResNet-50 model; S45: The extracted roughness feature value is compared with the corresponding value in the standard feature value library; S46: After the feature comparison is completed, the number of pixels in the roughened area is calculated by using a pixel statistics method, and the number of pixels is converted into an actual area, and then the ratio of the roughened area to the total area is calculated; S47: The system outputs the ratio of the roughened area to the dimension area, compares the ratio with the quality inspection qualified value, and outputs the quality inspection result of whether the inspected truss composite plate is qualified; S5, the upper computer sends the quality inspection results to the controller in the sorting platform, and the controller controls the action of the sorting platform to sort the truss composite panels whose roughening area meets the ratio requirements to the storage area of ​​qualified products, and sort the truss composite panels whose roughening area does not meet the ratio requirements to the unqualified channel.

2. A truss composite plate production quality inspection method according to claim 1, characterized in that: In step S1, it also includes cleaning the surface of the truss composite plate by a blowing device, the conveying platform includes a base, a conveyor belt for transmitting the truss composite plate is installed on the base, the blowing device includes a portal frame erected above the conveyor belt, and a blowing fan with an adjustable blowing angle is installed on the portal frame.

3. A truss composite plate production quality inspection method according to claim 2, characterized in that: In step S1, the industrial camera uses a camera mechanism to automatically adjust the shooting angle and fill light. The camera mechanism includes a camera suspension beam and a light source suspension beam. The light source suspension beam is an adjustable rotating rod. An anti-shake gimbal is installed on the camera suspension beam. The anti-shake gimbal is used to fix the industrial camera.

4. A truss composite plate production quality inspection method according to claim 3, characterized in that: In step S1, a photoelectric sensor for detecting whether the truss stacking plate enters the shooting range is fixed on the conveying platform.

5. A truss composite plate production quality inspection method according to claim 4, characterized in that: The sorting platform includes a supporting truss composite plate placement platform, a lifting gantry for transferring the truss composite plates is fixedly installed above the placement platform, a mechanical claw that can simultaneously grasp at least four grasping points of the truss composite plates is arranged on the gantry, and a controllable drip for marking is also arranged above the sorting platform, and the drip is filled with paint for marking the truss composite plates.

6. A truss composite plate production quality inspection method according to claim 5, characterized in that: In step S41, the noise removal method adopts Gaussian filtering, and the function of Gaussian filtering is Parameter σ=1.

0.

7. A truss composite plate production quality inspection method according to claim 5, characterized in that: In step S41, the contrast of the image is enhanced by histogram equalization. The histogram equalization includes calculating the histogram of the image, constructing a cumulative distribution function, and mapping the image pixel values ​​using the cumulative distribution function to make the grayscale values ​​of the image evenly distributed, thereby enhancing the rough texture in the image.

8. A truss composite plate production quality inspection method according to claim 7, characterized in that: Multiple contrast-enhanced images of different sizes are synthesized using the multi-scale Retinex algorithm.

9. A truss composite plate production quality inspection method according to claim 8, characterized in that: In step S44, the training process of the ResNet-50 model includes: collecting qualified truss composite plate surface images, deforming the images in angle and size through a distributed algorithm to form no less than 1,000 training samples, and after iterative training, randomly collecting no less than 500 uninspected truss composite plates, and continuously adjusting the hyperparameters of the initial ResNet-50 model. When the recognition accuracy reaches more than 95%, the model training is completed.

10. A truss composite plate production quality inspection method according to claim 9, characterized in that: After step S47, when the system completes the identification of the roughness ratio of the truss composite plate, the system calibrates the roughness features of the truss composite plate, stores the roughness features of the truss composite plate with a qualified roughness area ratio into a qualified roughness feature library, trains the truss composite plate with an unqualified roughness area ratio through the Faster R-CNN defect detection model, classifies the roughness defects of the unqualified truss composite plate, and outputs the unqualified roughness type in the host computer.

Images