CN112893186B - Rapid visual detection method and system for electrifying LED lamp filament - Google Patents

Abstract

The invention belongs to the technical field of LED detection, and specifically relates to a fast visual detection method and system for LED filament power-on, comprising the following steps: S1, collecting the power-on image of the LED filament; S2, cutting and performing the power-on image of the LED filament Value method threshold value processing; S3, divide the processed image into grid partitions according to the number of LED filaments and the number of luminous points of each LED filament; S4, set the qualified threshold, and calculate the luminescence of the luminous points in each grid Area, the luminous point whose luminous area is within the qualified threshold interval is judged as qualified, and the others are judged as unqualified. The image processing of the invention is simple and fast, greatly improves the power-on detection efficiency of the LED filament, and can be widely used in the field of LED detection.

Description

A fast visual inspection method and system for LED filament power-on

technical field

The invention belongs to the technical field of LED detection, and in particular relates to a fast visual detection method and system for LED filament power-on.

Background technique

In recent years, in the lighting industry, LED light-emitting products have occupied an increasing proportion in the market due to their advantages such as power saving and long life. After several years of development, LED lighting products have entered the Red Sea stage. How to increase production capacity and reduce costs in various links has become the key to the survival of various manufacturers. In some high-end applications, it is necessary to conduct power-on inspections on all products one by one, measure each luminous point, and eliminate those with abnormal, insufficient or non-luminous brightness. A typical example is the LED filament production industry. The ex-factory price of finished products after dispensing and drying is less than 10% of the original price. In order to improve efficiency, differentiated sales are often implemented, and the best quality is picked out at a higher price. Sell out to ensure the overall benefits, and the rest will be processed at a low price after secondary and tertiary screening. In order to achieve this goal, all products need to be powered on for testing. The photos after power-on are shown in Figure 1. Figure 1 is a part of 12 pieces in the field of view of a photo. In many cases, the overall bracket is 48 pieces in order to improve production efficiency. ;In the case of higher requirements, it is required that the brightness of all the luminous points on each filament is within the same narrow range, and those filaments whose brightness exceeds the standard, dim or not bright at individual points must be eliminated. This work has always been done manually, because human vision will quickly become fatigued, and it is easily affected by external conditions, such as light, fatigue and other factors, and the subjective standards of various quality inspectors are quite different, which leads to low production efficiency Moreover, the unqualified rate of the final product is high. Therefore, it is difficult to ensure the consistency of the power-on detection of the LED filament by manual inspection, and its efficiency is low.

There are also technical solutions for finding circles in the entire image through commercial image recognition software. But in actual use, the effect is very poor. On the one hand, it takes a long time. It often takes more than 100 seconds to process a picture. There is a problem that the recognition rate of the position of the luminous point is low and the misjudgment rate is high. Therefore, it is necessary to develop a fast visual detection method for LED filament power-on.

Contents of the invention

The invention overcomes the deficiencies in the prior art, and the technical problem to be solved is: to provide a fast visual detection method and system for LED filament power-on.

In order to solve the above-mentioned technical problems, the technical solution adopted in the present invention is: a fast visual detection method for LED filament power-on, comprising the following steps:

S1, collect the power-on image of the LED filament version;

S2. Cutting the power-on image of the LED filament version and performing binary threshold processing to remove areas not related to the LED filament and the halo around each luminous point in the image;

S3. Carry out grid partitioning of the processed image according to the number of filaments in the LED filament plate and the number of light-emitting points of each LED filament;

S4. Set a qualified threshold, calculate the luminous area of each luminous point in each grid, determine the luminous point whose luminous area is within the qualified threshold range as qualified, and judge the others as unqualified.

Before said step S1, the following steps are also included:

S0. Put the material box into the material rack, automatically push out a filament plate through the machine and power on;

After the step S4, the following steps are also included:

S5. Record the position of the unqualified filament, cut off the unqualified filament, and transfer the remaining filament plate to the blanking box.

In the step S3, the specific method of gridding and partitioning the image is as follows:

The image processed in step S2 is searched line by line from top to bottom to find the first bright pixel, and the line where the pixel is located is used as the horizontal baseline of the division.

In the step S3, before performing grid partitioning on the image, the steps of offsetting and rotating the image are also included.

In the step S3, the calculation method of the image offset and rotation is:

After adjusting the position and angle of the standard product in place in advance, record the position of the four mark points forming a rectangle on the product in the image, as well as the pixel equivalent of the image, and calculate the center position and angle of the rectangle;

After collecting the power-on image of the LED filament to be detected, calculate the center position and angle of the rectangle according to the positions of the four mark points in the image, compare it with the data of the standard product, and calculate the offset and rotation .

In addition, the present invention also provides a fast visual inspection system for LED filament power-on, including:

Feeding unit: used for automatic feeding;

Image acquisition unit: used to acquire the image of the LED filament plate after power-on;

Calculation unit: used to calculate and process the image of the LED filament plate, and judge the qualification;

PLC control unit: used to control the servo mechanism to cut off the unqualified filament according to the judgment result of the calculation unit;

Unloading unit: used to transfer the processed filament plate to the unloading box for unloading.

The calculation unit calculates and processes the image of the LED filament plate, and the specific method for judging the qualification is as follows:

The power-on image of the LED filament is cropped and thresholded by the binary method, and the area not related to the LED filament in the image and the halo around each luminous point are removed;

The processed image is gridded and partitioned according to the number of LED filaments and the number of luminous points of each LED filament;

Set the qualified threshold, calculate the area of the calculated luminous point in each grid, judge the luminous point whose area is within the qualified threshold interval as qualified, and judge the others as unqualified.

The specific method for the computing unit to perform grid partitioning on the image is as follows:

The processed image is searched row by row from top to bottom to find the first bright pixel, and the row where the pixel is located is used as the horizontal baseline of the division.

Compared with the prior art, the present invention has the following beneficial effects: the present invention provides a fast visual detection method and system for powering on the LED filament, and can greatly improve the efficiency of LED detection by powering on the LED filament and taking pictures and automatically identifying it , and, its accuracy is high, the system is stable and reliable, and can be widely used in the field of LED filament power-on detection.

Description of drawings

Fig. 1 is a control flow chart of a fast visual inspection method for powering on an LED filament provided by an embodiment of the present invention;

FIG. 2 is a flow chart of image processing in Embodiment 1 of the present invention;

FIG. 3 is a schematic diagram of the result of gridding and partitioning the image in Embodiment 1 of the present invention;

Fig. 4 is the result schematic diagram that obtains in the embodiment of the present invention;

Fig. 5 is another result schematic diagram obtained in the embodiment of the present invention;

FIG. 6 is a structural block diagram of a fast visual inspection system for LED filament power-on provided by Embodiment 2 of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, rather than All the embodiments; based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts all belong to the protection scope of the present invention.

Embodiment one

As shown in Figure 1, Embodiment 1 of the present invention provides a fast visual inspection method for LED filament power-on, including the following steps:

S0. Put the material box into the material rack, automatically push out a filament plate through the machine and power on.

S1. Collect the power-on image of the LED filament.

S2. Perform cropping and binary threshold processing on the power-on image of the LED filament, and remove areas not related to the LED filament in the image and halos around each luminous point.

S3. Perform grid partitioning on the processed image according to the number of LED filaments and the number of light-emitting points of each LED filament.

Among them, the specific method of grid partitioning the image is as follows:

The image processed in step S2 is searched line by line from top to bottom to find the first bright pixel, and the line where the pixel is located is used as the horizontal baseline of the division.

When taking pictures, after letting the product pass through the flow channel smoothly and directly below the camera, there will be a certain gap between the upper and lower sides, and the resulting image will have a certain drift in the up and down direction. In order to improve the accuracy of recognition, it is necessary to find the place where the luminous point starts as the horizontal baseline. In the production of product brackets, the consistency in the vertical direction can meet the requirements. After the switch is detected, the position of the LED filament plate is relatively fixed. After the image is divided evenly in this direction each time, each filament can be distributed in the grid without crossing the grid, so there is no need to consider its baseline. question. In this implementation of the LED filament version, it is searched line by line from top to bottom. All filaments in the image, no matter which one, as long as there is a point with a brightness that meets the standard and is at the highest position, this position is used as the baseline, not just Find it on a certain filament. In extreme cases, when the first luminous point of all filaments is not lit, the baseline will move down by one luminous point. At this time, the calculation result will consider that all points in the bottom row are not luminous, so that all filaments They are all waste products and must be completely removed. In this case, the machine will stop and notify manual inspection.

As shown in FIG. 2 , it is a schematic diagram of the grid partitioning of the image in the embodiment of the present invention.

In addition, in this embodiment, before performing grid partitioning on the image, steps of offsetting and rotating the image are also included.

The calculation method of image offset and rotation is:

After adjusting the position and angle of the standard product in place in advance, record the position of the four mark points forming a rectangle on the product in the image, as well as the pixel equivalent of the image, and calculate the center position and angle of the rectangle;

After collecting the power-on image of the LED filament to be detected, calculate the center position and angle of the rectangle according to the positions of the four mark points in the image, compare it with the data of the standard product, and calculate the offset and rotation .

In specific implementation, a state matrix (such as 100×50) can be established for all luminous points on the entire bracket, and each filament corresponds to a row. Before replacing each bracket, use the memory block operation command to initialize all points to zero, that is, all The luminous points are qualified, which is feasible, because the actual situation is that most of them are qualified. Before testing, use a fully qualified bracket as a template to evenly divide the matrix area on the screen, with exactly one luminescent point, and make a record for future use. In order to improve the recognition accuracy, after each photo is taken and binarized, the first bright pixel is found line by line from top to bottom, and it is used as the baseline of the division. Then count the area of the luminous pixels in each area and mark it on the screen at the same time.

In the case of poor product quality or a large gap in the conveying channel, there will be a situation where the luminous point is relatively large after taking a photo, and the collected photos need to be offset and rotated.

Specifically, there are four circular MARK points forming a rectangular vertex on the border of each bracket. The positions of the MARK points on different brackets are very consistent. You can find the positions of these points before detecting each bracket. Use The difference between the coordinates of the MARK points in the image collected during the measurement and the coordinates of the MARK points in the image collected when the bracket is in the standard position is calculated to calculate the offset and rotation angle of the bracket, and then the image is processed according to these two parameters. After rotation and offset, the luminescent point area is divided.

In actual operation, there is a simple algorithm. It is enough to only take pictures of three mark points. Use two points on the long side of the rectangle to calculate the rotation angle, and use the midpoint of the two points on the diagonal as the center of the rectangle. , which is the location. Practice has proved that this precision can meet the production requirements.

S4. Set a qualified threshold, calculate the luminous area of each luminous point in each grid, determine the luminous point whose luminous area is within the qualified threshold range as qualified, and judge the others as unqualified. Specifically, it is assumed that the threshold interval is [A, B], where A<B. When judging, if the luminous area is larger than A and smaller than B, it is judged as qualified, and if the luminous area is smaller than A or larger than B, it is judged as unqualified.

Each luminous point in the LED filament is the light emitted by each luminous LED particle after it is powered on and passes through the colored glue. It is inevitable that there will be divergence, with different shapes and blurred edges, rather than the first impression. round. This also leads to time-consuming and labor-intensive recognition with a high error rate when the circle is used as a template for recognition. In this embodiment, after discretization and binarization of the collected images, all points of the entire image are divided into two states: bright and not bright, and the threshold can be adjusted according to requirements to control the standard of qualified products. The number of pixels in the bright area accurately reflects the area of the luminous point, so it is a good and efficient way to count the number of pixels whose brightness reaches the standard in each area as a measure of the area of the luminous point. The fundamental purpose of the area is to isolate each luminous point.

As shown in Figure 4 and Figure 5, it is a schematic diagram of displaying the inverted image and the final judgment result on the display screen in the embodiment of the present invention. The luminous area is displayed in the corresponding area. The unqualified area is enlarged in the lower part of Figure 4, and the numbers in it show the luminous area of the luminous point whose brightness does not meet the standard, while the unqualified area is enlarged in the right part of Figure 5 , the numbers in it show the luminous area of the luminous points whose brightness exceeds the standard, and the areas without numbers are qualified products. Wherein, the light-emitting area of the light-emitting point is related to the number of pixels whose brightness reaches the standard in the area where it is located.

S5. Record the position of the unqualified filament, cut off the unqualified filament, and transfer the remaining filament plate to the blanking box.

In this embodiment, by detecting and analyzing the power-on of the LED filament, it can also reduce the defect rate and play a guiding role in the production process. Analyzing Figure 4, it can be seen that there are a large proportion of filaments with unqualified brightness, and the points with insufficient brightness are concentrated at the bottom of the image. Basically, the problem of the filament hardware itself can be ruled out. The possibility of such a thing happening is too small. After inspection and inquiry on the production site, I learned that in the production process, in order to ensure the beautiful appearance of the product, in the final stage of glue collection, instead of directly lifting the glue needle, there is a back-wiping action, which leads to the last The amount of glue on the light-emitting point is too much, blocking the light from coming through. Based on this point of view, it is recommended to run the dispensing action 2mm longer during production, and then perform the wipe-back action. After this improvement, the yield rate is greatly improved. Analyzing Figure 5, it is found that the brightness of many light-emitting points of the far right filament exceeds the standard. There are basically two reasons for this, one is that the point is off, and light leakage occurs, which was ruled out in this case; another reason is that the amount of glue is insufficient, and it is suspected that the glue is not flowing smoothly. After replacing the glue needle, The yield rate of this filament at this position is immediately improved a lot.

Embodiment two

As shown in Figure 6, Embodiment 2 of the present invention provides a fast visual inspection system for LED filament power-on, including:

Image acquisition unit: used to acquire the image of the LED filament plate after power-on;

Calculation unit: used to calculate and process the image of the LED filament plate, and judge the qualification;

PLC control unit: used to control the servo mechanism to cut off the unqualified filament according to the judgment result of the calculation unit;

Specifically, in this embodiment, the calculation unit performs calculation processing on the image of the LED filament plate, and the specific method for determining the qualified situation is as follows:

1. Cut the power-on image of the LED filament and perform binary threshold processing to remove areas not related to the LED filament in the image and the halo around each luminous point;

2. Grid partition the processed image according to the number of LED filaments and the number of luminous points of each LED filament;

3. Set the qualified threshold, calculate the luminous area of the luminous points in each grid, judge the luminous points whose luminous area is within the qualified threshold range as qualified, and judge the others as unqualified. Specifically, assuming that the threshold interval is [A, B], when judging, if the luminous area is larger than A and smaller than B, it is judged as qualified, and if the luminous area is smaller than A or larger than B, it is judged as unqualified.

Further, a fast visual inspection system for LED filament power-on in this embodiment also includes:

Feeding unit: used for automatic feeding;

Unloading unit: used to transfer the processed filament plate to the unloading box for unloading.

Further, a fast visual inspection system for LED filament power-on in this embodiment also includes an input unit, an output unit and a motion control unit connected to the PLC control unit, wherein the input unit includes various pushbutton switches, limit switches and The origin switch is used for the PLC control unit to monitor the various states of the system. The motion control unit includes the feeding axis, the flow channel axis, the cutter advance and retreat, the cutter lift and the discharge axis. The PLC control unit controls the system according to the status of each sensor. Loading, feeding, cutting, unloading and other operations. In addition, the output unit includes relays, air valves, etc. The PLC control unit can control other operating steps of the system by controlling the status of the relays and air valves.

In addition, in this embodiment, the image acquisition unit may specifically be a commercial camera. The system can also include a touch screen for displaying collected images and for man-machine interaction.

Specifically, in this embodiment, the specific method for the computing unit to perform grid partitioning on the image is as follows:

The processed image is searched row by row from top to bottom to find the first bright pixel, and the row where the pixel is located is used as the horizontal baseline of the division.

The LED filament power-on fast visual inspection system provided by the embodiment of the present invention has relatively low requirements on hardware such as cameras and operating systems. The industrial camera has a USB3.0 interface, 5 million pixels, and an image resolution of 2448X2048. Specifically, the calculation unit can be written in C# language using Open CV image processing technology in Visual Studio 2019 environment under 64-bit Windows 10.

To sum up, the present invention provides a fast visual detection method and system for LED filament power-up, through computer vision to detect and research the light-emitting point of LED filament power-on, and identify the location of an entire bracket LED filament under the 6th generation i3 processor The measured time is within 0.2 seconds. When using this detection system, 32K filaments are continuously selected on one machine during equipment acceptance, and all are qualified through manual re-inspection one by one. Therefore, the present invention can improve the production efficiency and quality of the detection system, and the detection consistency is good. According to statistics, one machine and one employee perform the auxiliary work of loading and unloading boxes, which is equivalent to the recognition speed of 6 employees. In this way, one employee can take care of three machines in actual production, and the operating requirements for employees are also greatly reduced, realizing the control of the detection system by fewer people. In terms of processing algorithm, digital image processing tools such as binary method and affine transformation are fully used. According to the characteristics of the target, the influence of long detection time and high misjudgment rate is comprehensively overcome, and the accuracy requirements of the detection system are finally met. The production capacity has been improved, the cost has been reduced, and the foundation has been laid for future practical and commercialization.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: It is still possible to modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the various embodiments of the present invention. scope.

Claims (5)

1. A fast visual detection method for LED filament power-on, characterized in that, comprising the following steps: S1, collect the power-on image of the LED filament version; S2. Cutting the power-on image of the LED filament version and performing binary threshold processing to remove areas not related to the LED filament and the halo around each luminous point in the image; S3. Carry out grid partitioning of the processed image according to the number of filaments in the LED filament plate and the number of light-emitting points of each LED filament; S4. Set the qualified threshold, calculate the luminous area of the luminous point in each grid, judge the luminous point whose luminous area is within the qualified threshold range as qualified, and judge the others as unqualified; In the step S3, before the image is gridded and partitioned, the steps of offsetting and rotating the image are also included, and the calculation method of the image offset and rotation is as follows: After adjusting the position and angle of the standard product in place in advance, record the position of the four mark points forming a rectangle on the product in the image, as well as the pixel equivalent of the image, and calculate the center position and angle of the rectangle; After collecting the power-on image of the LED filament to be detected, calculate the center position and angle of the rectangle according to the positions of the four mark points in the image, compare it with the data of the standard product, and calculate the offset and rotation . 2. A fast visual detection method for LED filament power-on according to claim 1, characterized in that, before the step S1, the following steps are also included: S0. Put the material box into the material rack, automatically push out a filament plate through the machine and power on; After the step S4, the following steps are also included: S5. Record the position of the unqualified filament, cut off the unqualified filament, and transfer the remaining filament plate to the blanking box. 3. A fast visual detection method for LED filament power-on according to claim 1, characterized in that, in the step S3, the specific method for gridding and partitioning the image is: The image processed in step S2 is searched line by line from top to bottom to find the first bright pixel, and the line where the pixel is located is used as the horizontal baseline of the division. 4. A fast visual detection system for LED filament power-on, characterized in that it comprises: Feeding unit: used for automatic feeding; Image acquisition unit: used to acquire the image of the LED filament plate after power-on; Calculation unit: used to calculate and process the image of the LED filament plate, and judge the qualification; PLC control unit: used to control the servo mechanism to cut off the unqualified filament according to the judgment result of the calculation unit; Unloading unit: used to transfer the processed filament plate to the unloading box for unloading; The calculation unit calculates and processes the image of the LED filament plate, and the specific method for judging the qualification is as follows: The power-on image of the LED filament is cropped and thresholded by the binary method, and the area not related to the LED filament in the image and the halo around each luminous point are removed; The processed image is offset and rotated, and the grid partition is performed according to the number of LED filaments and the number of luminous points of each LED filament; the calculation method of image offset and rotation is: After adjusting the position and angle of the standard product in place in advance, record the position of the four mark points forming a rectangle on the product in the image, as well as the pixel equivalent of the image, and calculate the center position and angle of the rectangle; After collecting the power-on image of the LED filament to be detected, calculate the center position and angle of the rectangle according to the positions of the four mark points in the image, compare it with the data of the standard product, and calculate the offset and rotation ; Set the qualified threshold, calculate the area of the calculated luminous point in each grid, judge the luminous point whose area is within the qualified threshold interval as qualified, and judge the others as unqualified. 5. A fast visual inspection system for LED filament power-on according to claim 4, characterized in that, the specific method for the calculation unit to perform grid partitioning on the image is: The processed image is searched row by row from top to bottom to find the first bright pixel, and the row where the pixel is located is used as the horizontal baseline of the division.

Images