METHOD AND DEVICE FOR ANONYMIZING A CAPTURED IMAGE OF AN INDUSTRIAL PLANT

Description

[0001] The invention relates to a method and a device for anonymized image capture in an industrial plant, such as a goods storage, goods transport and / or goods processing plant.

[0002] Image capture processes and devices are often used in industrial plants to monitor and control goods or process flows.

[0003] However, this often involves capturing personal characteristics, which may not be compliant with data protection regulations. These personal characteristics also require storage space and can make subsequent processing of the captured images more complex, for example, when larger image files need to be processed.

[0004] DE 10 2016 223 859 A1 discloses a camera for monitoring a surveillance area. EP 3 723 049 A1 describes a system for anonymizing content to protect privacy. GB 2 596 037 A describes a device for data anonymization. WO 2021 / 243465 A1 describes a system for automatically processing barcodes on personal protective equipment of people in a work area.

[0005] It is therefore an object of the present disclosure to provide an improved method and an improved device for anonymized image acquisition for industrial plants, which in particular simplifies subsequent storage and processing of the images.

[0006] This object is initially achieved by a method for anonymized image acquisition in an industrial plant according to independent claim 1.

[0007] In a first step of the method, an image is captured, for example, using an image capture unit such as a camera or an image sensor, in particular a line or matrix camera. In particular, the image capture unit is a so-called presentation camera. An image is captured regularly, continuously, and / or on an event-related basis, for example, by triggering a light barrier and / or a control element in the industrial plant.

[0008] The image includes an optoelectronically readable code, such as a barcode, a QR code, or a Semacode. The code is attached or applied to an object stored in the industrial facility, such as a product or a package.

[0009] The image further comprises at least one personal characteristic, such as a human hand holding the object, in particular in front of the camera for the purpose of capturing the image. For example, the personal characteristic can also comprise part of a fingerprint. The personal characteristic can also comprise clothing, in particular one or more sleeves, and / or jewelry, such as a ring or a watch. The personal characteristic is a characteristic that allows conclusions to be drawn about the person, in particular a characteristic by which the person can be identified, furthermore in particular uniquely identified.

[0010] In a further step of the process, this personal characteristic is then identified in the image, for example by means of a computing unit, as will be described in detail later.

[0011] In a further step of the process, the previously identified personal characteristic in the image is obscured. In this case, "obscured" specifically means "unrecognizable to the human eye," but can also include "unrecognizable to a processing unit." In particular, the obscuring is irreversible, meaning it cannot be reversed, neither by a human nor by a processing unit.

[0012] The anonymized image capture process makes it possible to process images captured in an industrial facility without taking data protection regulations into account, as the personal characteristics have been made unrecognizable and subsequent inference to a person is no longer possible, for example if the captured image is processed again and / or later for other purposes.

[0013] In particular, the image captured in this way can also be made available to cooperation partners external to the company and / or who are not subject to data protection regulations.

[0014] A particular advantage is that the images contain less information due to the obscured features and therefore require less storage space. Specifically, the original images are deleted, and only the images with the removed or obscured features are saved. Further processing of the images therefore requires less computing power and memory.

[0015] Preferably, but not part of the present invention, the images with the anonymized personal characteristics are used to influence the control of the industrial plant. For example, based on the readable code, a predetermined action of the industrial plant can be triggered, which, for example, results in further processing of a workpiece or the like.

[0016] The method can be further developed by identifying the personal characteristic in the image and segmenting the personal characteristic in the image. Segmentation, in this case, means subdividing the image into related elements and subsequently classifying these subdivisions or segments into individual, particularly predefined classes. Feature extraction algorithms and image preprocessing can also be used. Segmentation allows a statement to be made for certain identified areas, in particular whether these are personal characteristics.

[0017] Known methods can be used for segmentation, such as pixel-, edge-, and / or region-oriented methods. Additionally or alternatively, model-based and / or texture-based methods can be used for segmentation.

[0018] This advanced technology provides a particularly efficient solution. In particular, it ensures that only those areas of the captured image that relate to personal characteristics are processed, i.e., rendered unrecognizable.

[0019] The method can be further developed by identifying and / or segmenting the personal feature in the image using a neural network.

[0020] A neural network, also known as a neural network and / or artificial neural network, is a computer-based artificial intelligence. In particular, identification or segmentation can also be performed using a neural network trained by a deep learning algorithm, so that even newly acquired images that the neural network did not process during training can still be segmented.

[0021] In particular, identification or segmentation can be carried out using a specially designed computing unit, such as micro-processing units with a high degree of parallelization or programmable logic devices.

[0022] This advanced development provides a particularly powerful solution. In particular, the neural network provides a particularly reliable method for identifying or segmenting personal characteristics.

[0023] The method can be further developed in that the process of making the personal characteristic in the image unrecognizable comprises pixelating the personal characteristic in the image.

[0024] In this case, pixelation means that the original fine-pixel resolution of the personal characteristic is replaced by a much coarser resolution, so that instead of individual, individual sub-characteristics of the personal characteristic being visible, only unrecognizable pixels with a coarse structure are used. This can be achieved using a suitable pixelation algorithm.

[0025] The method can be further developed in that the obscuring of the personal characteristic in the image comprises a removal of the personal characteristic in the image.

[0026] In this case, removal means deleting the pixels relating to the personal characteristic and / or replacing these pixels with single-coloured pixels, such as white or black pixels.

[0027] According to the invention, the method further comprises reading the optoelectronically readable code.

[0028] The optoelectronically readable code is read using suitable reading algorithms. The result of the read code, for example, a serial number, such as a tracking number, or other code content, can then be stored along with the location and / or time of the read.

[0029] The method can be further developed in that the method further comprises storing the image with the personal characteristic made unrecognizable.

[0030] In particular, the storage can be carried out together with the code read out and a location and / or a time of reading out.

[0031] This advanced feature makes it possible to document the passing of a specific waypoint or the storage location of an object. Objects processed or manufactured using the industrial facility can thus be tracked within the facility.

[0032] The method can be further developed in that storing the image comprises transmitting it to a remote storage unit.

[0033] A remote storage unit is, in particular, a storage unit located remotely from the image capture unit and the industrial facility. In particular, the remote storage unit may be a storage unit on which the image with the anonymized personal characteristic is stored for verification purposes. The remote storage unit may, in particular, be located in a different country than the image capture unit or the industrial facility.

[0034] The method is further developed in that the method further comprises cropping the image to a partial area comprising the optoelectronically readable code.

[0035] Cropping, which can also be referred to as masking, means removing image content that does not include the optoelectronically readable code. Cropping can also involve cutting off part of the personal characteristic that was previously or subsequently obscured.

[0036] Cropping can be done across the board, meaning always by a specific cropping factor, and / or dynamically, for example based on an identification of the personal characteristic and / or the optoelectronically readable code.

[0037] If the cropping is performed before the identification or segmentation and / or before the blurring, which is not part of the present invention, less computing power is required for the identification or segmentation and / or blurring, since less image content has to be processed.

[0038] Cropping can occur both before and after transferring to the remote storage device. Alternatively, cropping can also occur without transferring the image to a remote storage device. This reduces storage space on the storage device, both locally and remotely.

[0039] This advanced training requires very little storage space and / or computing power.

[0040] The method can be further developed in that the method further comprises encrypting the captured image, transmitting the captured image to a remote computing unit and decrypting the captured image at the remote computing unit.

[0041] Encryption can be performed using common encryption algorithms. For example, the captured image can be sent to a remote computer via SFTP. Alternatively or additionally, the captured image can be encrypted and decrypted using a symmetric or asymmetric encryption protocol, for example.

[0042] In particular, encryption can take place immediately after the image has been captured, in particular before the personal characteristic is identified and / or rendered unrecognizable, and the thus unprocessed image can be transmitted to a remote processing unit and decrypted there.

[0043] The remote computing unit can be directly connected to the remote storage unit or can be different or remote from it.

[0044] This training prevents anyone from intercepting the captured image with the personal characteristics and drawing unauthorized conclusions about the person.

[0045] The method can be further developed by identifying and / or segmenting the personal feature in the image at the remote computing unit.

[0046] In particular, the remote processing unit can be specially configured to perform the identification or segmentation and, in particular, can also be designed to use a suitable neural network trained specifically for this application. Furthermore, the remote processing unit can be designed to meet specific data protection requirements, for example, through the location and / or certification of the remote processing unit.

[0047] This further development makes it possible for the anonymization to be carried out in a particularly data protection-compliant manner and not in the industrial plant. The aforementioned object is also achieved by a device for anonymized image capture in an industrial plant according to independent claim 10. .

[0048] Regarding further developments and their advantages, reference is made to the above explanations on the method for anonymized image acquisition in an industrial plant.

[0049] In particular, the device for anonymized image acquisition in an industrial plant is designed to partially or completely carry out embodiments of the previously described method for anonymized image acquisition in an industrial plant.

[0050] Embodiments of the method and device for anonymized image acquisition in an industrial plant will now be described in detail in connection with the following figures, wherein Fig. 1 shows a schematic view of an embodiment of a device for anonymized image capture in an industrial plant; Fig. 2 shows a flowchart of an embodiment of a method for anonymized image capture in an industrial plant; and Fig. 3a shows individual steps of an applied method for anonymized image capture in an industrial plant. The same reference symbols indicate the same or similar features.

[0051] Fig. 1 shows a schematic view of an embodiment of a device 1 for anonymized image capture in an industrial plant.

[0052] The device 1 initially comprises a local unit 10 and a unit 20 remote from the local unit 10. In particular, the remote unit 20 is shown as an example as a cloud service that contains parts of the device 1 or partial steps of a method 100, also as in connection with the following Fig. 2 described, can be carried out.

[0053] The local unit 10 comprises a local storage unit 11, a local processing unit 12, and an image acquisition unit 13. The local unit 10 also comprises a wireless and / or wired communication unit, which is not shown for reasons of clarity. Local, in this case, means located in or at the industrial plant or part of the industrial plant.

[0054] The remote unit 20 comprises a remote storage unit 21 and a remote processing unit 22. Additionally, the remote unit 20 comprises a wireless and / or wired communication unit, which is not shown for reasons of clarity. "Remote" in this case means spatially separated from the local unit 10 and, in particular, not located in the industrial plant.

[0055] In particular, the remote unit is one or more servers located at a remote location and forming a cloud service. For clarity, the remote storage unit 21 and the remote computing unit 22 are depicted as belonging to the same cloud service.

[0056] Alternatively, the storage unit 21 is provided by a first cloud service and the computing unit is provided by a second, different cloud service, wherein the first cloud service and the second cloud service are connected via a wireless and / or wired communication unit.

[0057] Furthermore, in particular, the remote unit 20 comprises at least no image acquisition unit or the local image acquisition unit 13 can be the only local unit arranged in the industrial plant besides a communication unit.

[0058] The device 1 is designed in particular to carry out the following in connection with Fig. 2 to carry out the procedures described.

[0059] Fig. 2 shows a flowchart of an embodiment of a method 100 for anonymized image acquisition in an industrial plant.

[0060] In a first step 101 of the method 100, an image is captured which includes an optoelectronically readable code and a personal feature.

[0061] In a subsequent step 102, the captured image is encrypted.

[0062] In a further step 103, the encrypted image is transmitted to a remote computing unit.

[0063] In a subsequent step 104, the encrypted image is decrypted at the remote processing unit.

[0064] In a further step 105, the personal feature in the decrypted image is identified by the remote processing unit by segmenting using a neural network.

[0065] In In a subsequent step 106, the personal feature in the image is made unrecognizable by pixelating the personal feature in the image and / or by removing the personal feature from the image.

[0066] In a further step 107, the image is cropped to a partial area comprising the optoelectronically readable code.

[0067] In a further step 108, the optoelectronically readable code is read out.

[0068] In a subsequent step 109, the image with the obscured personal characteristic is transferred to a remote storage unit and stored there in a final step 110. The method 100 can then begin again with step 101.

[0069] Fig. 3a-d show a process for anonymized image capture in an industrial plant. In a first step, an image 1000 is captured, which includes both a personal characteristic 1100 and an optoelectronically readable code 1200, as shown in Fig. 3a shown.

[0070] In the example shown, the personal characteristic 1100 consists of two hands, whereby a part of the garment, namely the sleeve, and a ring are also shown, which form the personal characteristic 1100 by way of example.

[0071] The optoelectronically readable code 1200 is in the present case designed as a QR code and is applied to an object 1300, which is shown here as an example as a rectangular sheet of paper.

[0072] The Fig. 3a Figure 1000 shown describes this in connection with Fig. 2 described in the first method step 101, as this occurs in particular when presenting the object 1300 with the help of the hands of a person in front of a presentation camera.

[0073] In a further step, the personal characteristic 1100 is then identified in the image 1000 using a segmentation algorithm executed by a neural network, as described in connection with the Fig. 2 shown method step 105 was described.

[0074] The result of this step 105 is in Fig. 3b shown, in which only the personal characteristic 1100 is contained in the image 1000. However, this is shown solely for the purpose of understanding the identification or segmentation. In fact, the object 1300 with the optoelectronically readable code 1200 is still contained in the image 1000 or represents the Fig. 3b The image shown does not represent all available image information, but only the result of the identification or segmentation.

[0075] In a further step, the personal characteristic 1100 contained in the image 1000 is made unrecognizable. This is Fig. 3c with a pixelation of the personal characteristic 1300. Alternatively, in Fig. 3d a removal or deletion of the personal characteristic 1100. This obscuring corresponds to the one in connection with the Fig. 2 described process step 106.

[0076] In addition, the Fig. 3d shown image 1000, the image section is cropped in such a way that further information not required for reading the optoelectronic code 1200, such as the background and parts of the personal characteristic, have been removed, so that the image 1000 is as in Fig. 3d shown comprises only a partial area of ​​the image 1000 on which the optoelectronically readable code 1200 is depicted. This in Fig. 3d The image shown in Figure 1000 corresponds to the one used in connection with the Fig. 2 image generated in process step 107 described above.

[0077] The previously described embodiments enable efficient solutions that lead to lower memory load and lower computing load, while at the same time contributing to the data protection-compliant processing of objects by persons in an industrial facility, such as a goods storage, goods transport and / or goods processing facility. Bezugszeichenliste

[0078] 1 Device 10 Local unit 11 Local storage unit 12 Local processing unit 13 Image acquisition unit 20 Remote unit 21 Remote storage unit 22 Remote processing unit 100 Method 101 Method step 102 Method step 103 Method step 104 Method step 105 Method step 106 Method step 107 Method step 108 Method step 109 Method step 110 Method step 1000 Captured image 1100 Personal characteristic 1200 Optoelectronically readable code 1300 Object

Claims

1. A method (100) for anonymizing an acquired image of an industrial plant, comprising the steps: - acquiring (101) an image (1000) by means of an image acquisition unit (13), wherein the image comprises an industrial plant, an optoelectronically readable code (1200) and a personal feature (1100); - identifying (105) the personal feature (1100) in the image (1000); - obscuring (106) the personal feature (1100) in the image (1000); - then cropping the image (1000) to a part region comprising the optoelectronically readable code (1200) such that further information not necessary for the reading out of the optoelectronically readable code (1200) is removed; and - reading out (108) the optoelectronically readable code (1200).

2. A method (100) according to the preceding claim 1, wherein the identifying (105) comprises segmenting the personal feature (1100) in the image (1000).

3. A method (100) according to the preceding claim 2, wherein the identifying (105) and / or the segmenting takes / take place by means of a neural network.

4. A method (100) according to any one of the preceding claims, wherein the obscuring (106) comprises pixelating the personal feature (1100) in the image (1000).

5. A method (100) according to any one of the preceding claims, wherein the obscuring (106) comprises removing the personal feature (1100) from the image (1000).

6. A method (100) according to any one of the preceding claims, further comprising the step: - storing (110) the image with the obscured personal feature (1100).

7. A method (100) according to the preceding claim 6, wherein the storing (110) of the image comprises a transmission (109) to a remote memory unit (21).

8. A method (100) according to any one of the preceding claims, further comprising the steps: - encrypting (102) the acquired image (1000); - transmitting (103) the acquired image (1000) to a remote processing unit (22); and - decrypting (104) the acquired image (1000) at the remote processing unit (22).

9. A method (100) according to the preceding claim 8, wherein the identifying (105) and / or the segmenting takes / take place at the remote processing unit (22).

10. An apparatus (10) for anonymizing an acquired image of an industrial plant, comprising: - an image acquisition unit (13) which is configured to acquire an image (1000), wherein the image (1000) comprises an industrial plant, an optoelectronically readable code (1200) and a personal feature (1100); - a memory unit (11, 21) which is configured to store the acquired image (1000); and - a processing unit (12, 22) which is configured to identify the personal feature (1100) in the image (1000) and to obscure the personal feature (1100) in the image (1000), wherein the apparatus (10) is further configured to then crop the image (1000) to a part region comprising the optoelectronically readable code (1200) such that further information not necessary for the reading out of the optoelectronically readable code (1200) is removed, and to read out the optoelectronically readable code (1200).