Automatic unstacking method for air conditioner condenser core

Abstract

The present application relates to the technical field of automation production, and especially relates to an automatic unstacking method for air conditioner condenser cores. The present application adopts the following technical scheme: the method comprises the following steps: a 3D camera is arranged above a condenser core incoming platform to identify the side of the heat dissipation fins of the condenser core; a robot drives a clamp to clamp and move the uppermost condenser core to a discharging conveying line according to the two fin sides of the condenser core identified by the 3D camera; a 2D camera is arranged on the discharging conveying line to identify and detect the end type of the condenser core; finally, the robot places the condenser core on the discharging conveying line with the U-shaped tube end downward according to the end type of the condenser core. The beneficial effect is that the identification efficiency can be improved under the premise of ensuring the identification accuracy; in addition, the 3D identification and 2D identification methods are optimized, the identification efficiency is further improved, and the automatic unstacking effect is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automation production, and particularly relates to an automatic unstacking method for air conditioner condenser cores. BACKGROUND

[0002] The air conditioner condenser is composed of multiple condenser cores arranged side by side, each condenser core includes multiple U-shaped copper pipes arranged side by side and heat dissipation fins inserted between the U-shaped copper pipes, and thus each condenser core includes two heat dissipation fin sides and one expanded pipe end and one U-shaped pipe end. During the processing of the air conditioner condenser, a single condenser core needs to be detached from the stacked condenser cores and placed on a conveying line for the next connecting and assembling process. At present, the unstacking operation of the air conditioner condenser core mainly includes manual unstacking and automatic unstacking. The manual unstacking has the problems of high labor intensity and low efficiency. The automatic unstacking uses a robot-driven clamp to clamp and unstack and transfer the condenser core. Since the types of the ends of the condenser core need to be recognized during the unstacking and transferring process, and the U-shaped pipe end is placed downward on the conveying line after the heat dissipation fin sides on both sides of the condenser core are clamped, and the sizes of the cores of different condenser models are different, the conventional automatic clamping equipment is difficult to adapt to the unstacking operation of the condenser cores of different models, and the areas for clamping on both sides of the condenser core are the heat dissipation fins, and if the clamping force is too large, the heat dissipation fins are easily damaged. The conventional visual recognition system is used to recognize the condenser core to improve the accuracy of clamping during the unstacking process. Since the condenser core needs to be recognized twice (once when the condenser core is detached from the stacked condenser cores to detect the heat dissipation fin sides for clamping, at this time, the expanded pipe end and the U-shaped pipe end are shielded and cannot be accurately recognized, and the other time when the condenser core is placed on the unloading line to recognize whether the upper end is the expanded pipe end or the U-shaped pipe end), if the conventional visual recognition method is used, the recognition time is too long, the unstacking cycle is prolonged, and the unstacking and transferring efficiency is reduced. SUMMARY

[0003] The present application aims to provide an automatic unstacking method for air conditioner condenser cores, and particularly to provide an automatic unstacking method for air conditioner condenser cores which can quickly recognize the ends of the condenser core and automatically unstack.

[0004] To achieve the above-mentioned purpose, the present application adopts the following technical scheme: an automatic unstacking method for air conditioner condenser cores, comprising the following steps:

[0005] S1, a 3D camera is arranged above a condenser core feeding platform, the 3D camera is used to take a photo of the uppermost condenser core on the condenser core feeding platform and recognize the two fin sides of the condenser core.

[0006] S2, using the robot to drive the clamp to clamp and move the uppermost condenser core to the unloading conveying line according to the position of the condenser core and the positions of the two fin sides identified by the 3D camera.

[0007] S3, a 2D camera is arranged on the unloading conveying line to take a picture of the condenser core clamped by the robot and identify the type of the end of the condenser core as a U-shaped tube end or an expanded tube end.

[0008] S4, the robot places the condenser core on the unloading conveying line with the U-shaped tube end downward according to the type of the end of the condenser core identified in step S3.

[0009] Specifically, step S1 specifically includes the following steps:

[0010] S11, the 3D camera acquires the original point cloud data of the condenser core by using structured light, and filters and denoises the original point cloud data.

[0011] S12, the PCA principal component analysis algorithm is used to calculate the filtered and denoised point cloud data, and the global principal direction calculated by PCA is used for contour feature extraction to determine the position of the fin side.

[0012] S13, the centroid coordinates of the condenser core are calculated according to the filtered and denoised point cloud data to determine the position of the condenser core.

[0013] Specifically, in step S11, when the 3D camera acquires the original point cloud data of the condenser core by using structured light, HDR high dynamic range imaging is added, and three exposures are performed with exposure times of 0.5ms, 2ms and 8ms respectively, and then integrated to obtain; at the same time, when the 3D camera samples the original point cloud data of the condenser core, the point cloud density sampling is 10000 points / frame at the edge part of the condenser core, and the point cloud density sampling is 2000 points / frame at the middle plane area of the condenser core.

[0014] Specifically, in step S12, when the PCA principal component analysis algorithm is used to calculate the filtered and denoised point cloud data, the point cloud data is divided into blocks, and PCA calculation is performed on each point cloud data block in a multi-thread parallel manner, and then the global principal direction is obtained after all point cloud data blocks are calculated and combined.

[0015] Specifically, in step S11, when filtering and denoising the original point cloud data, a Gaussian filtering algorithm is used.

[0016] Specifically, in step S3, when using a 2D camera to photograph the condenser core picked up by the robot and identify the type of the end of the condenser core, Hough transform is used to detect the circular hole feature to identify the expansion tube end, and Canny edge detection is used to extract the U-shaped contour to identify the U-shaped tube end.

[0017] Specifically, in step S3, when using a 2D camera to take pictures of the condenser core picked up by the robot and identify the type of the end of the condenser core, the image of the end area of ​​the condenser core in the condenser core photo taken by the 2D camera is cropped out and the end type of the condenser core is identified.

[0018] Specifically, in step S1, a 3D camera is used to take a picture of the condenser core at the top of the condenser core receiving platform, and in step S3, a 2D camera is used to take a picture of the condenser core. Deformation detection is performed on the condenser core. If the condenser core shows obvious deformation, it is determined to be a defective product. In step S4, the robot places it on the unloading conveyor line, and the unloading conveyor line rejects it.

[0019] The beneficial effects of this invention are as follows: by setting a 3D camera above the condenser core material receiving platform and a 2D camera on the unloading conveyor line, the side and end types of the condenser core heat dissipation fins can be identified and detected respectively, which can improve the identification efficiency while ensuring the identification accuracy. In addition, the 3D identification and 2D identification methods are optimized to further improve the identification efficiency and improve the automatic destacking effect. Attached Figure Description

[0020] Appendix Figure 1 This is a schematic diagram showing the location distribution of various equipment and products during the automatic unstacking process of the air conditioner condenser core in the embodiment.

[0021] Appendix Figure 2 This is a schematic diagram of the specific structure of the robot gripper holding the condenser core in the embodiment. Detailed Implementation

[0022] Example 1, referring to Figures 1-2 An automatic unpacking method for air conditioner condenser cores includes the following steps:

[0023] S1. Set up a 3D camera 2 above the condenser core material receiving platform 1, and use the 3D camera 2 to take pictures of the condenser core 10 at the top of the condenser core material receiving platform 1 and identify the two fin side surfaces 11 of the condenser core 10.

[0024] S2. Using the position of the condenser core 10 identified by the 3D camera 2 and the position of the two fins, the robot 3 drives the clamp 31 to clamp the uppermost condenser core 10 and move it to the unloading conveyor line 4.

[0025] S3, a 2D camera 5 is arranged on the blanking conveying line to take a picture of the condenser core 10 gripped by the robot 3 to identify the type of the end of the condenser core 10, which belongs to the U-shaped tube end 12 or the expanded tube end 13.

[0026] S4, the robot places the condenser core 10 on the blanking conveying line 4 in a direction with the U-shaped tube end downward according to the type of the end of the condenser core 10 identified in step S3; wherein, the blanking conveying line shown in Figure 1 The transition conveying roller in the middle of the two conveying lines is used as a transition between the two conveying lines. Since the U-shaped tube end is composed of multiple U-shaped copper tubes, and the diameter of the U-shaped copper tube is small, the transition conveying roller can stably convey the U-shaped tube end. The main body of the conveying line is a plate chain structure, that is, a chain is used to drive a plate-shaped chain plate, and a rubber plate for preventing slipping and collision is arranged on the surface of the chain plate, which can effectively improve the conveying stability of the condenser core. Figure 1 As shown in

[0027] Specifically, in step S1, the following steps are specifically included:

[0028] S11, the 3D camera acquires the original point cloud data of the condenser core by using structured light, and filters and denoises the original point cloud data. When the 3D camera acquires the original point cloud data of the condenser core by using structured light, HDR high dynamic range imaging is added, and three exposures are performed with exposure times of 0.5 ms, 2 ms and 8 ms, respectively, and then integrated to obtain; at the same time, when the 3D camera samples the original point cloud data of the condenser core, a point cloud density of 10000 points / frame is used to sample the edge part of the condenser core, and a point cloud density of 2000 points / frame is used to sample the middle plane area of the condenser core, so that the time of the 3D camera for collecting and processing the point cloud data can be reduced. Through actual test, the time of the 3D camera for collecting and processing the point cloud data can be reduced from 30 ms to 5 ms. In addition, when filtering and denoising the original point cloud data, a Gaussian filtering algorithm is used, the kernel size of the Gaussian filtering is generally 5x5 pixels, which can be adjusted according to the actual point cloud density, and the standard deviation is 1.5, which is an empirical value and is set according to the identification requirements of the condenser core in this embodiment to balance the smoothness and edge definition of the point cloud data.

[0029] S12, using PCA principal component analysis algorithm to calculate the point cloud data after filtering and denoising, and according to the global principal direction calculated by PCA, the profile feature is extracted, and the position of the fin side is judged. When using PCA principal component analysis algorithm to calculate the point cloud data after filtering and denoising, the point cloud data is divided into blocks, and then the PCA calculation of each point cloud data block is carried out in multi-thread parallel mode. After all the point cloud data blocks are calculated, the global principal direction is obtained by merging. Through the way of block and multi-thread parallel calculation of local PCA, the time-consuming of calculation can be effectively reduced. According to the actual test, the PCA calculation time can be reduced from 20ms to 6ms, thereby effectively improving the recognition efficiency.

[0030] S13, the centroid coordinates of the condenser core are calculated according to the point cloud data after filtering and denoising, so as to determine the position of the condenser core; it is convenient for the robot to control the clamp to reach the position of the condenser core for clamping. Wherein, the centroid coordinates of the condenser core are obtained by calculating the average value of all point cloud data coordinates, and the calculation formula is as follows:

[0031]

[0032] Wherein, x i , y i , z i is the coordinate value of the i-th point in the point cloud data, i=1, 2, 3……N, and N is the total number of point cloud data.

[0033] Specifically, in step S3, when the 2D camera is used to take a picture of the condenser core picked up by the robot and identify the type of the end of the condenser core, the Hough transform is used to detect the circular hole feature to identify the expanded tube end, and the U-shaped profile is extracted based on the Canny edge detection to identify the U-shaped tube end. In addition, when identifying, the image of the end region of the condenser core in the picture taken by the 2D camera is cropped to identify the type of the end of the condenser core. When the Hough transform is used to detect the circular hole feature to identify the expanded tube end, the image is first preprocessed by using Gaussian filtering and adaptive binarization in sequence, and then the Hough transform is used to detect the circular hole to identify the expanded tube end. When the Hough transform is detected, the parameters are set as follows: dp=1.8, minDist=30, param1=200, param2=0.92, minRadius=15, maxRadius=25. The value of dp is the inverse ratio of the accumulator resolution and the image resolution, which is generally in the range of 1.0-2.0. The larger the value, the faster the calculation speed and the lower the accuracy. In this embodiment, 1.8 is taken according to experience and identification requirements. minDist is the minimum distance between the detected circle centers (in pixels), which is used to prevent repeated detection of circles. For densely arranged expanded tube holes, it should be set to be slightly larger than 2 times the maximum radius. param1 is the high threshold value of edge detection, which is in the range of 0-300, and is generally set to 150. The larger the value, the more strict the edge detection. In this application, it is set to 200. param2 is the accumulator threshold value, which is in the range of 0-1.0. The closer the value to 1.0, the more strict the detection standard. It is usually set to 0.8-0.95 to obtain the best effect. In this embodiment, 0.92 is taken. When the U-shaped profile is extracted based on the Canny edge detection to identify the U-shaped tube end, considering the low-contrast region of the U-shaped bottom and the convenience of threshold adjustment, the U-shaped template matching is added, that is, the standard U-shaped profile template (the template uses SVG vector graphics) is pre-stored, and the pre-stored template is used for matching identification. After identifying whether the top end of the condenser core is an expanded tube end or a U-shaped tube end, the robot can control the clamp to place the condenser core on the unloading conveying line with the U-shaped tube end facing downward.

[0034] In addition, when the 3D camera is used to take a picture of the uppermost condenser core on the condenser core incoming platform in step S1, and the 2D camera is used to take a picture of the condenser core in step S3, the condenser core is subjected to deformation detection. If the condenser core is obviously deformed, it is judged that the condenser core is a defective product. The robot can place it on the unloading conveying line in step S4, and the unloading conveying line can remove it.

[0035] Of course, the above is only a preferred embodiment of the present application, and is not intended to limit the scope of use of the present application. Therefore, any equivalent changes made on the principle of the present application should be included in the protection scope of the present application.

Claims

1. An automatic unstacking method of an air conditioner condenser core, characterized by, Comprise the following steps: S1, 3D camera is arranged above the condenser core incoming platform, the uppermost condenser core on the condenser core incoming platform is photographed using the 3D camera, and the two fin sides of the condenser core are identified, specifically comprising the following steps: S11, the 3D camera obtains the original point cloud data of the condenser core using structured light, and filters and denoises the original point cloud data;When the 3D camera obtains the original point cloud data of the condenser core using structured light, HDR high dynamic range imaging is added, and three exposures are performed with exposure times of 0.5ms, 2ms and 8ms respectively, and then integrated;At the same time, when the 3D camera samples the original point cloud data of the condenser core, the point cloud density sampling is 10000 points / frame at the edge of the condenser core, and the point cloud density sampling is 2000 points / frame in the middle plane area of the condenser core; S12, the PCA principal component analysis algorithm is used to calculate the filtered and denoised point cloud data, and the global principal direction calculated by PCA is used to extract the contour feature, and the position of the fin side is judged; S13, the centroid coordinates of the condenser core are calculated according to the filtered and denoised point cloud data, so as to determine the position of the condenser core; S2, the robot drives the clamp to clamp and move the uppermost condenser core to the unloading conveying line according to the position of the condenser core and the position of the two fin sides identified by the 3D camera; S3, a 2D camera is arranged on the unloading conveying line to photograph and identify the type of the end of the condenser core clamped by the robot, which belongs to U-shaped tube end or expanded tube end; S4, the robot places the condenser core on the unloading conveying line with the U-shaped tube end downward according to the type of the end of the condenser core identified in step S3; In step S1, the uppermost condenser core on the condenser core incoming platform is photographed using the 3D camera, and in step S3, the condenser core is photographed using the 2D camera, and deformation detection is performed on the condenser core, if the condenser core is deformed obviously, it is judged that the condenser core is a defective product, and the robot places it on the unloading conveying line in step S4, and the unloading conveying line removes it.

2. An automatic unstacking method of an air conditioner condenser core according to claim 1, characterized in that: In step S12, when the PCA principal component analysis algorithm is used to calculate the filtered and denoised point cloud data, the point cloud data is divided into blocks, and the PCA calculation is performed on each point cloud data block in a multi-thread parallel manner, and the global principal direction is obtained after all point cloud data blocks are calculated.

3. An automatic unstacking method of an air conditioner condenser core according to claim 1, characterized in that: In step S11, the original point cloud data is filtered and denoised using Gaussian filtering algorithm.

4. An automatic unstacking method of an air conditioner condenser core according to claim 1, characterized in that: In step S3, the 2D camera is used to photograph the condenser core clamped by the robot and identify the type of the end of the condenser core, the Hough transform is used to detect the circular hole feature to identify the expanded tube end, and the U-shaped contour is extracted based on Canny edge detection to identify the U-shaped tube end.

5. An automatic unstacking method of an air conditioner condenser core according to claim 4, characterized in that: In the step S3, when the 2D camera is used to take a photo of the condenser core picked up by the robot and identify the type of the end portion of the condenser core, the image of the end portion region of the condenser core in the photo of the condenser core taken by the 2D camera is cropped out before the identification of the type of the end portion of the condenser core is performed.

Images