TWI830549B - Objects automatic labeling method and system applying the same - Google Patents

Abstract

An automatic object labeling method includes steps: M continuous image frames are captured at one station of an assembly line. An object detection step is performed which includes selecting a detection image frame that displays an operation including a work piece and a target object from M continuous image frames; and calibrating the position range of the target object in the detection image frame. An Nth retrace image frame is selected from the M consecutive image frames retraced forward from the detection image frame, and a target image of the target object is obtained from the Nth retrace image frame according to the position range. The target image is compared with the M continuous image frames to match at least one other image similar to the target image both are stored as the same labeled data set.

Description

Object automatic marking method and system

The present invention relates to an automatic object labeling method and its system, and in particular, to an automatic object labeling method and its system suitable for artificial intelligence in an assembly line.

Artificial Intelligence (AI) mainly uses computer programs to perform data analysis and information decision-making, and can be applied to production line automation, energy saving, and improvement of operational efficiency. It has now become a development trend in major industries around the world.

The first step in industrial application of artificial intelligence is to define operating procedures and analyze the work content to provide standardized data that computers/machines can learn from. Take the artificial intelligence technology of computer vision as an example. In order for a computer to have the ability to recognize images, it is necessary to first define a specific range (bounding box) on the image screen and give it a standardized annotation. As the "standard answer" for how machine learning identifies different objects. This process of processing and preparing the original data of operating procedures and work content is called data marking.

However, the labeling work seems simple but is quite tedious and requires a lot of manpower and time (consuming more than 80% of the time of the entire artificial intelligence project). And if manual data labeling is used, subjective bias may affect the basis of machine learning standard answers.

Therefore, there is a need to provide an advanced automatic object marking method and system to solve the problems faced by the conventional technology.

One embodiment of this specification discloses a method for automatic object marking, which includes the following steps: first, capture M consecutive frames (frames) at a station of the pipeline. Then an object detection step is performed, including: selecting a detection view frame from M continuous view frames. This detection view frame displays an operation procedure performed on a target object using a work piece. Then mark the position range of the target object in the detection frame. Backtrack from the detection image, select the Nth traceback view frame among the M consecutive view frames, and obtain the marked image of the target object from the Nth traceback viewframe based on this position range. Subsequently, the marked image is compared with M continuous view frames to compare at least one other marked image similar to the marked image; and the marked representative image and other marked images are stored as the same marked data set.

Another embodiment of this specification discloses an automatic object marking system, including an image capture device, an object detection module and a correlation comparison module. The image capture device is used to capture M consecutive view frames at one station of the pipeline. The object detection module is used to perform an object detection step. The object detection steps include: from M A detection view frame is selected from consecutive view frames. This detection view frame displays an operation procedure performed on a target object using a work piece. Then calibrate the position range of the target object in the detection frame; go back from the detection image and select the Nth traceback frame among the M consecutive viewframes; and based on this position range, trace back from the Nth frame Obtain the marked image of the target object. The correlation comparison module is used to compare the marked image with M continuous view frames to compare at least one other marked image that is similar to the marked image; and to store the marked representative image and other marked images as the same mark. data set.

According to the above embodiments, this specification provides a method and system for automatic marking of objects. It uses a two-stage artificial intelligence module to capture and mark objects to be marked in M consecutive view frames at one station of the pipeline. First, an artificial intelligence model is trained and constructed through manual verification. Then, the artificial intelligence model uses the workpiece in the operation program as a benchmark to calibrate the target objects participating in the operation program and determine the location range of the target object. . After extracting the image feature parameters of the target object, an unsupervised learning algorithm (for example, association) is used to compare the image features with the target object in M consecutive view frames based on these image feature parameters. The image of at least one other target object that is similar to the target object is marked with the same tag data.

Data labeling through artificial intelligence modules can significantly reduce the labor cost and time of data labeling. And because data labeling is based on whether the workpiece in the operation program performs the specified operation on the target object is used as an objective judgment criterion, it can also improve the problem of machine learning inaccuracy caused by human subjective bias when manually labeling data.

10: Object automatic marking system

11:Image capture device

12:Object detection module

13: Association comparison module

14:Database

14a: Marked data set

20: Assembly line

21:Site

22: gloves

22a: Workpiece block

22a’: Backtracking workpiece block

23:Object to be marked

23a: Block of objects to be marked

24:Target object

S21, S22, S23, S241, S242, S25, S26, S27: steps

K ₁ -K ₅ , F _t+1 -F _m : continuous view frame

F ₀ -Ft: training view frame

FD: Detection frame

S, S’: position range of the target object

P ₁ -P _n : characteristic parameters

FN: Lookback view frame

L: field of view length

V: horizontal movement speed

In order to have a better understanding of the above and other aspects of this specification, embodiments are cited below and described in detail with the accompanying drawings: Figure 1 shows an automatic object marking system according to an embodiment of this specification. The functional configuration schematic diagram; Figure 2 is a flow chart illustrating a method for automatic marking of objects using the automatic object marking system in Figure 1, according to an embodiment of this specification; and Figure 3 is a flow chart of a method for automatic marking of objects according to an embodiment of this specification. A schematic diagram showing partial view frame changes when executing the automatic object marking method described in Figure 2.

This instruction manual provides an automatic object labeling method and its system, which can significantly reduce the labor cost and time of data labeling, and at the same time improve the problem of machine learning inaccuracy due to subjective deviation when manual data labeling is used. In order to make the above-mentioned embodiments and other objects, features and advantages of this specification more clearly understandable, a plurality of preferred embodiments are enumerated below and described in detail with reference to the accompanying drawings.

However, it must be noted that these specific implementation examples and methods are not intended to limit the present invention. The invention may still be implemented using other features, components, methods and parameters. The preferred embodiments are proposed only to illustrate the technical features of the present invention and are not intended to limit the patentable scope of the present invention. Those with ordinary knowledge in this technical field will be able to make various adjustments based on the description of the following specification without departing from the spirit of the present invention. Make equal modifications and changes within the scope of God. In different embodiments and drawings, the same components will be represented by the same component symbols.

Please refer to Figures 1 to 3. Figure 1 is a schematic functional configuration diagram of an automatic object marking system 10 according to an embodiment of this specification. Figure 2 is a flowchart illustrating a method for automatic marking of objects using the automatic object marking system 10 of Figure 1 according to an embodiment of this specification. Figure 3 is a schematic diagram of partial view frame changes when executing the automatic object marking method described in Figure 2 according to an embodiment of this specification.

In some embodiments of this specification, the automatic object marking system 10 is configured in a pipeline 20 . The operating procedures performed at the station 21 of the assembly line 20 may include using workpieces to process, select, classify, print, mark or otherwise process a target object. For example, in this embodiment, the assembly line 20 may include a conveyor belt. The operating procedures performed at the station 21 include a picking worker wearing gloves 22 (i.e., workpieces) to select a specific type of plastic product (i.e., the object to be marked 23) from a variety of different plastic products (i.e., the object to be marked 23) on the conveyor belt. Target object 24).

As shown in Figure 1, the automatic object marking system 10 includes an image capture device 11, an object detection module 12, an association comparison module 13 and a database 14. The image capture device 11 is installed on a station 21 of the assembly line 20 and is used to capture M continuous view frames F for the operation procedures performed on the station 21 (as described in step S21 in Figure 2). ₀ -F _m (M is a positive integer greater than 1). In some embodiments of this specification, the image capturing device 11 may include at least one digital camera or video camera.

In this embodiment, when the image capture device 11 captures a variety of different plastic products (i.e., objects to be marked) 23 passing through the conveyor belt at the station 21 of the assembly line 20 for about 4.5 minutes, approximately 8100 continuous view frames ( That is, M equals 8100). The frame rate (FPS) of the M consecutive view frames F ₀ -F _m captured by the image capturing device 11 is 30 frames per second (FPS).

The object detection module 12 is used to perform an object detection step. First, as described in step S22 of Figure 2, the object detection module 12 is used to separate out a plurality of object blocks 23a and/or to be marked in each of the M continuous view frames F ₀ -F _m . or a work piece block 22a.

For example, in some embodiments of this specification, the step S22 of using the object detection module 12 to distinguish a plurality of object blocks 23a to be marked and/or a workpiece block 22a may include using a fast area convolutional neural network. (Faster Region-based Convolutional Neural Networks, Faster R-CNN), YOLOv4 and other object detection artificial intelligence models (but not limited to this). For example, in some embodiments, each of the M continuous view frames F ₀ -F _m can be used with an object detection artificial intelligence model, and then in each of the M continuous view frames F ₀ -F _m A plurality of object blocks 23a to be marked are respectively separated to select the plastic products (ie, the objects to be marked) 23 displayed in the corresponding view frame.

In this embodiment, as shown in Figure 3, gloves 22 worn by the picking workers appear in M continuous view frames F ₀ -F _m including K continuous operation view frames K ₀ -K ₆ . View frames K ₀ -K ₂ show the process of the picking worker's gloves 22 moving towards the target object 24 ; view frame K ₃ shows the picking worker's gloves 22 starting to grab the target object 24 ; view frames K ₄ -K ₆ show the picking worker The glove 22 has grabbed the target object 24 and moved outside the viewing frame. When the content displayed in the corresponding view frame (for example, K continuous operation view frames K ₀ -K ₆ ) includes gloves 22 (ie, workpieces) worn by the picking workers, the object detection module 12 can not only distinguish In addition to the plastic product (i.e., the object to be marked) 23, a workpiece area 22a of the framed gloves 22 (i.e., the workpiece) is also separated in the view frame.

In some embodiments of this specification, step S22 of using the object detection module 12 to separate out a plurality of object blocks 23a and/or a workpiece block 22a includes the following sub-steps. First, calculate K consecutive In each of the operation view frames K ₀ -K ₆ , the intersection rate over objects (Intersection rate over objects) of the workpiece block 22 a and each to-be-marked object block 23 a (including the target object 24 ) in the corresponding view frame . After that, the one with the largest intersection ratio (for example, the operation view frames K ₃ -K ₆ ) is selected from the K continuous operation view frames K ₀ -K ₆ to determine whether the workpiece (glove 22 ) performs an operation on the target object 24 item-specific operating procedures. For example, in some embodiments of this specification, when the number of frames (K ₃ -K ₆ ) with the largest intersection ratio in the K continuous operation view frames K ₀ -K ₆ is greater than H (for example, 4), it can be confirmed A specific type of plastic product (target object 24) has been grasped by the glove (workpiece block 22a). Where H is a positive integer greater than 2 (H>2). In this embodiment, the number of H may be 3.

In this embodiment, the maximum intersection ratio between the workpiece block 22a and each object block 23a to be marked in the corresponding view frame (continuous operation view frame K ₀ -K ₆ ) is the ratio between the workpiece block 22a and the corresponding view frame. The maximum overlapping area ratio of each object block 23a to be marked in the frame (continuous operation view frame K ₀ -K ₆ ), such as equation (1):

In some embodiments of this specification, as described in step S23 of Figure 2: multiple view frames among M consecutive view frames F ₀ -F _m can be optionally selected as training view frames F ₀ - F _t (as shown in Figure 3), and through manual verification, train and construct an artificial intelligence model (i.e., object detection module 12) to perform segmentation of a plurality of object blocks 23a to be marked. and/or step S22 of a workpiece block 22a. For example, in this embodiment, machine learning (ML) based on convolutional neural network algorithm can be used to construct the artificial intelligence model (object detection module 12). At the same time, 120 view frames within the first minute of 8100 continuous view frames (ie, M equals 8100) captured by the image capture device 11 for 4.5 minutes are selected as the training view frames F ₀ -F _t , and are manually verified. Training artificial intelligence model (object detection module 12). In some embodiments of this specification, the artificial intelligence model suitable for constructing the object detection module 12 may be, for example, Faster Region-based Convolutional Neural Networks (Faster R-CNN), YOLOv4, etc. Detect artificial intelligence models or a combination of the above.

Then, as described in step S241 in Figure 2: select a detection view frame FD from M consecutive view frames F ₀ -F _m . For example, in some embodiments of this specification, the so-called detection frame FD is the frame from which it can initially be determined that the picking worker's gloves 22 (workpieces) have grasped a specific type of plastic product (target object) 24 . In this embodiment, the detected view frame FD is from the M continuous view frames F ₀ -F _m shown in Figure 3, excluding other view frames after the training view frames F ₀ -Ft (for example, , continuous view frames F _t+1 -F _m ), and then continuously operate the view frames K ₃ -K ₆ from the K (K is a positive integer greater than 1) containing the work piece block 22a (glove 22 appears) , select the first operation view frame (for example, the operation view frame K ₃ ) among K ₃ -K ₆ with the largest intersection ratio as the detection view frame FD.

At the same time, as described in step S242 of FIG. 2 , the position range S of the target object 24 is calibrated in the detection frame FD. In some embodiments of this specification, in detecting the view frame The step of demarcating the position range S of the target object 24 in FD includes the following steps: among the plurality of object blocks 23a to be marked in the detection frame FD, select the area with the largest intersection ratio with the workpiece block 22a. Block selection range. The object block 23a to be marked in the block selection range can be marked as the target object 24; at the same time, the position range S of the target object 24 in the detection view frame FD can be located.

Then, as described in step S25 of Figure 2, trace back from the detection image frame FD, and select the Nth traceback frame FN (for example, the operation frame) among the K consecutive operation frames K ₀ -K ₆ K ₀ ); and according to the position range S (and multiple relative position ranges S') of the target object 24 in the detection frame FD, obtain the mark image of the target object 24 from the Nth lookback frame FN. The step of backtracking from the detection image frame FD includes the following sub-steps: First, according to the horizontal movement speed V of the pipeline and the field of view length L of the detection frame FD, the estimated position range S corresponds to the detection frame FD. A plurality of relative position ranges S′ of the position range S of the target object 24 . Next, select the Nth traceback view frame FN.

The sub-step of estimating the relative position range S' includes: dividing the horizontal movement speed V of the assembly line by the field of view length L (V/L) of the detection frame FD, the distance between the target object 24 entering and moving out of the detection frame FD can be calculated. Time required for field of view. Then from the frame rates FPS of M continuous view frames F ₀ -F _m , the displacement distance of each continuous view frame (continuous operation view frame K ₀ -K ₆ ) can be estimated, and then from the detection view frame FD (such as , the operation view frame K ₃ ) goes back N consecutive view frames (for example, in Figure 3, N=3) (for example, the operation view frame K ₀ ). And it is calculated that the position range S in the detection view frame FD (for example, the operation view frame K ₃ ) corresponds to the relative position range S' of the lookback view frame (for example, the operation view frame K ₀ ).

In some embodiments of this specification, the number of N can be obtained by calculating the Gaussian function of the frame rate (FPS/6) of one-sixth of the M continuous view frames F ₀ -F _m : N=[FPS/6 ]. For example, in this embodiment, the horizontal movement speed V is defined as the number of view frames from when the target object 24 enters the field of view to when it leaves the field of view (for example, V=58). Coupled with the size of the detection frame FD (length × width, h × w), the displacement distance of each continuous frame (continuous operation frame K ₀ -K ₆ ) can be estimated to be approximately several pixels (pixels) Length of quantity: pixels=int(w/V)×int( FPS /6)=int(1920/58)×int(30/6)=33×5

It can be converted to the relative position range S' of the position range S in the detection view frame FD (for example, the operation view frame K ₃ ) in each view frame corresponding to the backtracking (the operation view frame K ₀ -K ₂ ); at the same time The workpiece block 22a in the detection view frame FD can also be converted into the backtracking workpiece block 22a' in each view frame (operation view frame K ₀ -K ₂ ) corresponding to the backtracking.

The sub-step of selecting the Nth traceback view frame FN includes: among the K continuous operation view frames K ₀ -K ₆ , select a target object 24 that is not in the workpiece block 22 a (glove 22 ) except for the detection view frame FD. ) obscures the operation view frame as the Nth lookback view frame FN (for example, the operation view frame K ₀ ). At this time, the intersection ratio between the workpiece block 22a and the relative position range S' is 0; the intersection ratio between the traceback workpiece block 22a' and the relative position range S' is still maximum. The image selected in the relative position range S′ in the Nth traceback view frame FN is the marked image of the target object 24 .

Subsequently, as described in step S26 of Figure 2, the marked image is compared with M continuous view frames F ₀ -F _m to compare at least one other marked image similar to the marked image. In some embodiments of this specification, the step of comparing the marked image with the M continuous view frames F ₀ -F _m includes an image correlation comparison technology using an unsupervised learning algorithm. In this embodiment, the image correlation comparison technology includes the following steps: first, extract a plurality of feature parameters (feature extraction) P1-Pn of the marked image; import these feature parameters _P1 - _Pn into the correlation comparison module 13 , to compare at least one other marked image that is similar to the marked image.

As described in step S27 of Figure 2, the mark representative image and other mark images are transmitted to the database 14 in the object automatic marking system 10 and stored as the same mark data set 14a, thereby completing an automated marking of the target object 24.

In some embodiments of the present description, different operating procedures may also be used at the same station 21 or different stations (not shown) of the pipeline 20 to repeat steps S21 to S25 for other types of items to be marked in the pipeline 20 The object proceeds to the next object marking process. For example, in some embodiments of this specification, picking workers can wear gloves (not shown) of different colors or styles as workpieces, and grab them from the same station 21 or different stations (not shown) of the assembly line 20 Other types of plastic products (objects to be marked) 23 in the conveyor belt serve as target objects (not shown) in the next object marking process. And repeat steps S21 to step S25 to mark another type of plastic product in the assembly line 20 .

Comparing the automatic object labeling method provided by the embodiments of this specification with the traditional manual labeling operation, it can be found that the labeling efficiency of the automatic object labeling method provided by the embodiments of this specification is 68 times that of the manual labeling operation; selection detection The accuracy rate of view frame FD (judging that the glove 22 has grasped the specific target object 24) is as high as 86.73%; and the accuracy rate (recall) of object image comparison is as high as 92.68%. It can be seen that using the automatic object labeling method provided by the embodiments of this specification can greatly increase the data labeling efficiency of the artificial intelligence system and reduce the labor cost and time of data labeling. Furthermore, since this manual The automatic object marking method provided by the embodiment uses whether the workpiece in the operation program performs a specified operation on the target object (the glove 22 grabs the specific target object 24) as an objective judgment criterion. Therefore, it can also improve the manual data marking. At this time, the problem of machine learning inaccuracy caused by human subjective bias.

According to the above embodiments, this specification provides a method and system for automatic marking of objects. It uses a two-stage artificial intelligence module to capture and mark objects to be marked in M consecutive view frames at one station of the pipeline. First, an artificial intelligence model is trained and constructed through manual verification. Then, the artificial intelligence model uses the workpiece in the operation program as a benchmark to calibrate the target objects participating in the operation program and determine the location range of the target object. . After extracting the image feature parameters of the target object, an unsupervised learning algorithm (for example, correlation comparison) is used to compare at least one other image feature similar to the target object in M continuous view frames based on these image feature parameters. The image of the target object and tag both with the same tag data.

Data labeling through artificial intelligence modules can significantly reduce the labor cost and time of data labeling. And since data labeling is based on whether the workpiece in the operation program performs the specified operation on the target object as the objective judgment standard, it can also improve the machine learning inaccuracy caused by human subjective bias when the conventional technology uses manual data labeling. problem.

Although the present invention has been disclosed above in terms of preferred embodiments, they are not intended to limit the present invention. Anyone with ordinary knowledge in the technical field may make some modifications and modifications without departing from the spirit and scope of the present invention. , therefore, the protection scope of the present invention shall be subject to the scope of the appended patent application.

S21, S22, S23, S241, S242, S25, S26, S27: steps

Claims (18)

An object automatic marking method includes: using an image capturing device to capture M continuous frames at a station of an assembly line; using an object detection module to perform an object detection step, It includes: selecting a detection frame from the M continuous frames, the detection frame showing an operation procedure performed on a target object using a workpiece; and marking the detection frame in the detection frame. A position range of the target object; using a correlation comparison module to perform a correlation comparison step, including: tracing back from the detection view frame, and selecting an Nth traceback view frame among the M consecutive view frames; According to the position range, a mark image of the target object is obtained from the Nth lookback frame; the mark image is compared with the M consecutive view frames to compare at least one mark image that is similar to the mark image. other tagged images; and storing the tagged representative image and the at least one other tagged image as the same tagged data set. The object automatic marking method as described in claim 1, wherein the step of using the object detection module to select the detection frame from the M continuous view frames includes: Use an object detection module to separate out a plurality of object blocks to be marked and/or a workpiece block in each of the M continuous view frames; select from the M continuous view frames containing K consecutive operation view frames of the workpiece block; and among the first of the K consecutive operation view frames, the workpiece block goes back to the Nth traceback viewframe and the Nth traceback viewframe A plurality of intersection rates (Intersection rates over objects) of a plurality of object blocks to be marked are searched, and the one with the largest intersection rate is searched; among them, the first one of the K consecutive operating view frames is used as the detection view frame. The object automatic labeling method described in claim 2, wherein the object detection step further includes: using multiple training view frames in the M continuous view frames to train and construct an artificial intelligence ( Artificial Intelligence (AI) model; and using the output results of the artificial intelligence model to calculate the plurality of intersection ratios. The automatic marking method of objects as described in request item 2, wherein the maximum intersection ratio is the plural number in the detection view frame, the workpiece block goes back to the Nth traceback viewframe, and the Nth traceback viewframe A maximum overlapping area ratio of the object blocks to be marked. The object automatic marking method as described in claim 2, wherein the object with the maximum intersection ratio in the detection view frame is the target object. The object automatic marking method described in claim 2, wherein the step of selecting the Nth traceback view frame using the association comparison module includes selecting the target object in the detection view frame that is not blocked by the workpiece The occluder serves as the Nth lookback frame. The object automatic marking method described in claim 2, wherein the step of using the object detection module to mark the position range of the target object in the detection frame includes: selecting the target object in the detection frame A frame selection range for the one with the maximum intersection ratio among a plurality of object blocks to be marked; and based on a horizontal movement speed of the pipeline and a field of view length of the view frame, it is estimated that the frame selection range is in each of the K consecutive Manipulate multiple relative position ranges within the viewframe. The automatic marking method for objects as described in claim 1, wherein the assembly line is a conveyor belt; and the operating procedure uses a glove as the workpiece to select the target object. The automatic object marking method as described in claim 1, wherein the step of comparing the marked image with the M continuous view frames includes using an unsupervised learning algorithm, including: Extract a plurality of feature parameters (feature extraction) of the marked image; and input the plurality of feature parameters into an association model to compare the at least one other marked image in the M continuous view frames. An automatic object marking system includes: an image capture device used to capture M continuous view frames at a station of a pipeline; an object detection module used to perform an object detection step, including: from the Select a detection view frame from M consecutive view frames, wherein the detection view frame displays an operation procedure performed on a target object using a workpiece; mark a position of the target object in the detection view frame Range; look back from the detection image, select an Nth lookback frame among the M continuous viewframes; and obtain a mark of the target object from the Nth lookback frame based on the position range image; and a correlation comparison module, including a step for performing a correlation comparison, including: comparing the marked image with the M continuous view frames to compare at least one other image that is similar to the marked image. Mark the image; and store the marked representative image and the at least one other marked image as the same marked data set. The object automatic marking system as described in claim 10, wherein the step of selecting the detection view frame from the M continuous view frames includes: Separate a plurality of object blocks to be marked and/or a workpiece block in each of the M continuous view frames; select K blocks containing the workpiece block from the M continuous view frames. Continuously operate viewframes; and calculate the first of the K consecutively operated viewframes, the workpiece block going back to the Nth traceback viewframe and the plurality of object blocks to be marked in the Nth traceback viewframe A plurality of intersection ratios, and search for the largest intersection ratio among them; among them, the first one of the K consecutive operation view frames is used as the detection view frame. The automatic object marking system as described in claim 11, wherein the object detection step further includes: using multiple training view frames in the M continuous view frames to train and construct an artificial intelligence model in a manual verification manner. ; and using the artificial intelligence model output result to calculate the complex intersection proportions. The object automatic marking system as described in request item 11, wherein the maximum intersection ratio is the detection view frame, and the workpiece block goes back to the Nth traceback viewframe and a plurality of the Nth traceback viewframe. A maximum overlapping area ratio of the object area to be marked. The object automatic marking system as described in claim 11, wherein the object with the maximum intersection ratio in the detection frame is the target object. The object automatic marking system as described in claim 11, wherein the step of selecting the Nth lookback frame includes selecting the target object that is not obscured by the workpiece block in the detection frame as the Nth A lookback viewframe. The automatic object marking system as described in claim 11, wherein the step of marking the position range of the target object in the detection view frame includes: selecting the plurality of object blocks to be marked in the detection view frame A frame selection range that has the maximum intersection ratio; and according to a horizontal movement speed of the pipeline and a field of view length of the view frame, estimate the frame selection range in each of the K consecutive operation view frames. relative position range. The automatic marking system for objects as described in claim 10, wherein the assembly line is a conveyor belt; and the operating procedure uses a glove as the workpiece to select the target object. The automatic object marking system of claim 10, wherein the step of comparing the marked image with the M continuous view frames includes using an unsupervised learning algorithm, including: extracting a plurality of characteristic parameters of the marked image ; and inputting the plurality of feature parameters into the correlation comparison model, and comparing the at least one other marked image in the M continuous view frames.

Images