An infant care camera privacy protection method based on instance segmentation detection

Abstract

This invention relates to a privacy protection method for baby care cameras based on instance segmentation detection, comprising: Step 1, region detection, using an instance segmentation detection algorithm to identify and locate the adult region, the baby region, and the background region; Step 2, background region processing, the identified background region is first reduced in size, then Gaussian blurred, and then enlarged back to its original size to obtain a Gaussian blurred background region, which is then filled back into the original background region. The privacy protection method for baby care cameras based on instance segmentation detection obtained by this invention has the following advantages: the algorithm segments the camera image into adult, baby, and background regions, providing a foundation for subsequent processing; reducing the background region saves computational power required for subsequent processing, improving processing efficiency; and by combining modern image processing technology and artificial intelligence algorithms, it achieves efficient and aesthetically pleasing privacy protection measures.

Description

Technical Field

[0001] This invention relates to the field of camera privacy protection technology, and in particular to a method for protecting the privacy of baby care cameras based on instance segmentation detection. Background Technology

[0002] Infants lack the ability to protect themselves and are highly vulnerable to danger when unsupervised. For example, if an infant accidentally rolls over while sleeping or playing, their face may fall face down. Due to their immature skeletal and muscular development, they cannot easily right themselves, and if parents do not turn them over in time, it can lead to suffocation. Another example is when an infant kicks off their blankets while sleeping; if parents do not cover them up promptly, they may catch a cold. Parents need to expend considerable time and energy caring for their infants to ensure their healthy growth. However, the time parents can spend with their babies is often limited. To monitor their infants' condition at all times, many parents use monitoring devices, but this also carries the risk of privacy breaches. Summary of the Invention

[0003] One objective of this application is to provide a privacy protection method for baby care cameras based on instance segmentation detection, which can reduce computing power requirements while protecting family privacy.

[0004] The technical solution adopted in this application is: a method for protecting the privacy of baby care cameras based on instance segmentation detection, including: Step 1: Detect the camera frame area using an instance segmentation detection algorithm to identify and locate the adult area and the infant area. The remaining part of the camera frame is the background area. Step 2: Background area processing. The identified background area is first reduced in size, then Gaussian blur is applied, and then it is enlarged back to the original size to obtain the Gaussian blurred background area. The Gaussian blurred background area is then filled back into the original background area.

[0005] The algorithm accurately identifies and locates adults and infants in the image, segmenting the camera feed into adult, infant, and background areas to provide a foundation for subsequent processing. Reducing the size of the background area saves computational resources and improves processing efficiency. Applying Gaussian blur to the background protects privacy, creating a semi-transparent frosted effect that effectively hides background details while maintaining overall image harmony. Furthermore, the blurred background reduces the need for edge processing of the infant area, lowering computational requirements and saving resources. By combining modern image processing technology and artificial intelligence algorithms, this efficient and aesthetically pleasing privacy protection measure enhances the user experience and provides greater security.

[0006] In some embodiments of this application, the scaling down process in step two is performed by proportionally reducing the size to 1 / 10 to 1 / 20 of the original size. Scaling down to this range significantly reduces the number of pixels in the background area, lowering the complexity of Gaussian blur calculations. It also ensures that the background area exhibits an ideal blur effect after being enlarged back to the original size, effectively avoiding background loss due to over-scaling or wasted computational power due to under-scaling. Furthermore, proportional scaling ensures that the enlarged background after processing accurately matches the original image size, preventing issues such as edge misalignment or stretching distortion.

[0007] In some embodiments of this application, the Gaussian blur in step two has a convolution kernel of a 5*5 matrix or a 3*3 matrix.

[0008] In some embodiments of this application, a fourth step is also included: periodically updating the background area by repeating the operation of step two within a set time period. This mechanism can effectively cope with dynamic changes in the background environment in the scene. By periodically updating the background area, it can ensure that the blurring process is always based on the latest background state, avoiding the situation where some areas that originally belonged to the background are not blurred due to background changes, or the foreground elements that have changed are mistakenly included in the background blur range.

[0009] Furthermore, the time setting in step four is 1-5 seconds. Updating within this 1-5 second range ensures timely updates to the latest background area without creating a significant visual inconsistency due to the baby's excessive movement.

[0010] In some embodiments of this application, a third step, adult area processing, is also included: when an adult area is detected, the adult area is filled with a corresponding Gaussian blurred background area. Filling the adult area with a Gaussian blurred background area ensures a natural transition between the adult area and the Gaussian blurred background area, thereby effectively protecting adult privacy while maximizing the visual coherence and integrity of the video image. This allows caregivers to clearly observe the baby's activities without disclosing the adult's privacy information when checking on the baby.

[0011] Furthermore, in step three, if there is no corresponding Gaussian blurred background area in the adult area, then a solid color is used to fill the adult area. Solid color filling is the lowest-level fallback mechanism of the program.

[0012] Furthermore, the solid color fill in step three uses gray. Gray blends well into this type of environment, reducing the visual difference between the filled area and the surrounding background, thereby further improving the overall harmony of the image and ensuring that privacy protection is achieved while maintaining the integrity and aesthetics of the video image to the greatest extent possible.

[0013] Furthermore, in step three, if the adult area is located in front of the infant area, the adult area is filled with a corresponding Gaussian blurred background area. This approach avoids the loss of image information caused by adults obscuring the infant, and by filling the background with a blurred background consistent with the surrounding environment, it protects the privacy of adults while maintaining a relatively natural visual presentation of the foreground image where the infant is located. This ensures that caregivers can clearly observe the infant's condition while effectively isolating the privacy of adults.

[0014] Furthermore, in step three, if the adult area has a corresponding Gaussian blurred background area, then the corresponding adult area is filled with the corresponding Gaussian blurred background area, while the remaining adult area is filled with a solid color. This method of filling in different areas further refines the processing logic for the adult area in complex scenes. It maximizes the use of the generated Gaussian blurred background to maintain a natural transition in the image, and through the fallback strategy of solid color filling, it ensures that regardless of the adult area's position in the image or the degree of overlap with the background area, it can effectively shield the adult's privacy while ensuring that the infant area remains clearly visible and does not affect the realization of the caregiving function.

[0015] The present invention provides a privacy protection method for baby care cameras based on instance segmentation detection, which has the following advantages: 1. It can effectively protect the privacy of family members and home background without affecting the real-time monitoring of the baby's status; 2. It achieves automated processing through intelligent algorithms, reducing the need for manual intervention and requiring less computing power; 3. It uses a semi-transparent frosted effect instead of the traditional solid color masking, increasing visual comfort and the overall aesthetics of the image. Attached Figure Description

[0016] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solutions of this disclosure, the following detailed, clear, and complete description of this disclosure is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this disclosure and are not intended to limit it.

[0018] In the description of this application, the use of "first" and "second" is for the purpose of distinguishing technical features only, and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated.

[0019] Those skilled in the art should understand that in the disclosure of this application, the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the above terms should not be construed as limitations on this application.

[0020] like Figure 1 As shown, the baby care camera privacy protection method based on instance segmentation detection according to an exemplary embodiment of this application may include: Step 1: Detect the camera frame area using an instance segmentation detection algorithm to identify and locate the adult area and the infant area. The remaining part of the camera frame is the background area. Step 2: Background area processing. The identified background area is first reduced in size, then Gaussian blur is applied, and then it is enlarged back to the original size to obtain the Gaussian blurred background area. The Gaussian blurred background area is then filled back into the original background area.

[0021] Instance segmentation and detection algorithms, such as Mask R-CNN, are existing technologies that can accurately identify and locate adults and infants in the image, and will not be elaborated upon in this paper. The image is segmented into adult, infant, and background regions to provide a foundation for subsequent processing. Reducing the size of the background region saves computational power and improves processing efficiency. Applying Gaussian blur to the background region protects privacy information in the background environment. Gaussian blur creates a semi-transparent frosted effect, effectively hiding background details while maintaining the overall harmony of the image. Furthermore, the blurred background reduces the processing requirements for the infant region's edges, resulting in lower computational demands and saving computational power. By combining modern image processing technologies and artificial intelligence algorithms, efficient and aesthetically pleasing privacy protection measures are achieved, which not only improves the user experience but also provides users with higher security.

[0022] No privacy protection is applied to the infant area; the original video footage is retained so that caregivers can clearly observe the infant's real-time movements.

[0023] In an exemplary embodiment, the scaling down process in step two involves proportionally reducing the size to 1 / 10 to 1 / 20 of the original size. Scaling down to this range significantly reduces the number of pixels in the background area, lowering the complexity of Gaussian blur calculations. It also ensures that the background area exhibits an ideal blur effect after being enlarged back to the original size. This effectively avoids the waste of computational power caused by excessive or insufficient scaling, and the proportional scaling guarantees that the enlarged background accurately matches the original image size, preventing issues such as edge misalignment or stretching distortion.

[0024] In an exemplary embodiment, the Gaussian blurring in step two uses a 5x5 or 3x3 matrix convolution kernel. Choosing a 5x5 matrix allows for more thorough pixel fusion of the reduced background area, resulting in a softer, more natural blurring effect, particularly suitable for scenes with numerous fine textures or complex lighting variations in the background. A 3x3 matrix, while maintaining a certain level of blur, offers higher computational efficiency and can achieve ideal privacy masking effects even for relatively simple backgrounds with smooth color transitions. In practical applications, the choice can be flexible based on specific scenario requirements. For example, a 3x3 convolution kernel is sufficient when the background is a simple indoor wall, while a 5x5 convolution kernel can better eliminate background details and further enhance the reliability of privacy protection if the background contains rich details such as curtain folds or floor patterns.

[0025] In an exemplary embodiment, a fourth step is also included: periodically updating the background area by repeating the operation of step two on the background area within a set time period. This mechanism can effectively cope with dynamic changes in the background environment in the scene. For example, when a caregiver moves toys near the crib, adjusts the opening and closing of the curtains, or the brightness of the room changes due to changes in the angle of sunlight, by periodically updating the background area, it can ensure that the blurring processing is always based on the latest background state, avoiding the situation where some areas that originally belonged to the background are not blurred due to background changes, or the changed foreground elements are mistakenly included in the background blur range.

[0026] During the update process, to ensure the smoothness of the video stream, a double buffering mechanism can be adopted. That is, when the system processes a new background area, it still uses the background data that has been processed in the previous cycle to synthesize the current video frame and output it. After the new background is processed, it smoothly switches to avoid the phenomenon of the picture being stuttered or flickering, thus balancing the real-time nature of privacy protection and the continuity of video viewing.

[0027] In an exemplary embodiment, the set time in step four is 1-5 seconds. Since infants generally have a small range of movement and move slowly, updating within this 1-5 second timeframe ensures timely updates to the latest background area without creating a significant visual disharmony due to excessive infant movement.

[0028] The set time can be flexibly adjusted according to the dynamic frequency of changes in the actual application scenario. For example, during the nighttime sleep period when the environment around the crib is relatively stable and there is little human activity, the set time can be set to 3-5 seconds to reduce system resource consumption; while in scenarios where the baby is awake and playing during the day, and caregivers frequently enter and exit or adjust room items, the set time can be shortened to 1-2 seconds to ensure that the background area can respond to environmental changes more quickly, thereby further optimizing the balance between system performance and resource consumption while ensuring privacy protection.

[0029] With a short update time, it can also be used in conjunction with the processing of adult areas. When filling adult areas with solid colors, it can quickly update and fill the blurred background areas into the adult areas, ensuring that the picture becomes more harmonious as soon as possible.

[0030] In an exemplary embodiment, a third step, adult area processing, is also included. When an adult area is detected, it is filled with a corresponding Gaussian blurred background area. To ensure the protection of adult privacy, privacy protection processing is also required for the adult area. Based on the position of the adult area in the image, an image of the corresponding position is cropped from the Gaussian blurred background area and filled into the adult area, making the transition between the adult area and the Gaussian blurred background area natural. This effectively protects adult privacy while maintaining the visual continuity and integrity of the video image to the greatest extent, allowing caregivers to clearly observe the baby's activities without disclosing the adult's private information.

[0031] In an exemplary embodiment, in step three, if there is no corresponding Gaussian blurred background area in the adult area, then the adult area is filled with a solid color. When the camera is first turned on and the adult is in front of the camera, that is, when there is no corresponding Gaussian blurred background area in the adult area, it is impossible to capture the image of the corresponding position from the Gaussian blurred background area. At this time, the adult area is filled with a solid color to ensure the protection of the adult's privacy.

[0032] Solid color fill is the lowest-level fallback mechanism of the program. In the event of an error in the background area fill or the adult area fill, solid color fill can be used to provide basic privacy protection to ensure the reliability of privacy protection.

[0033] In an exemplary embodiment, the solid color fill in step three uses gray. In most families, living rooms, bedrooms, and other areas where babies frequently play, background elements such as walls, furniture, or curtains are mostly in neutral tones. Gray can blend well into such environments, reducing the visual difference between the filled area and the surrounding background, thereby further improving the overall harmony of the image and ensuring that privacy is protected while maintaining the integrity and aesthetics of the video image to the greatest extent.

[0034] In an exemplary embodiment, in step three, if the adult area is in front of the infant area, the adult area is filled with the corresponding Gaussian blurred background area. When the adult area is in front of the infant area, meaning the adult and infant areas overlap and the adult area is closer to the foreground of the camera, the adult area may partially or completely obscure the infant area. In this case, the adult area is still filled with the Gaussian blurred background area. If there is no Gaussian blurred background area in the overlapping area, a solid color is used to fill the adult area.

[0035] This does not conflict with the principle of not obscuring the baby. In the actual video footage, the baby is already obscured by the adult, reflecting the authenticity of the image. This approach avoids the loss of image information caused by the adult obscuring the baby, and by using a blurred background that matches the style of the surrounding environment, it protects the adult's privacy while maintaining a relatively natural visual presentation of the foreground where the baby is located. This ensures that the caregiver can clearly observe the baby's condition while effectively isolating the adult's privacy.

[0036] In an exemplary embodiment, in step three, if the adult area has a corresponding Gaussian blurred background area, then the corresponding adult area is filled with the corresponding Gaussian blurred background area, while the remaining adult area is filled with a solid color. For example, if the upper half of the adult area has a corresponding Gaussian blurred background area, but the lower half does not, then the upper half will be filled with the Gaussian blurred background at that location, maintaining visual continuity with the surrounding processed background. The lower half will be uniformly filled with a preset solid color, ensuring that the entire adult area is completely covered and preventing the leakage of privacy information. This method of filling in different areas further refines the processing logic for the adult area in complex scenes. It maximizes the use of the generated Gaussian blurred background to maintain a natural transition in the image, and through the fallback strategy of solid color filling, it ensures that regardless of the adult area's position in the image or the degree of overlap with the background area, it can effectively shield the adult's privacy while ensuring that the infant area is always clearly visible and does not affect the realization of the caregiving function. This situation tends to occur when the adult region and the infant region partially overlap. In other words, the corresponding position of the infant region does not have a Gaussian blurred background region. However, when the adult region and the infant region overlap and the adult region is in front, the overlapping adult region does not have a corresponding Gaussian blurred background region, while the non-overlapping part does.

[0037] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for protecting the privacy of baby care cameras based on instance segmentation detection, characterized in that, include: Step 1: Detect the camera frame area using an instance segmentation detection algorithm to identify and locate the adult area and the infant area. The remaining part of the camera frame is the background area. Step 2: Background area processing. The identified background area is first reduced in size, then Gaussian blur is applied, and then it is enlarged back to the original size to obtain the Gaussian blurred background area. The Gaussian blurred background area is then filled back into the original background area.

2. The method for protecting the privacy of a baby care camera based on instance segmentation detection according to claim 1, characterized in that, The reduction process in step two involves proportionally reducing the size to 1 / 10 to 1 / 20 of the original size.

3. The method for protecting the privacy of a baby care camera based on instance segmentation detection according to claim 1, characterized in that, The Gaussian blur in step two has a convolution kernel that is either a 5x5 matrix or a 3x3 matrix.

4. The method for protecting the privacy of a baby care camera based on instance segmentation detection according to claim 1, characterized in that, It also includes step four, which involves periodically updating the background area by repeating step two within a set time period.

5. A method for protecting the privacy of a baby care camera based on instance segmentation detection according to claim 4, characterized in that, The time set in step four is 1-5 seconds.

6. The method for protecting the privacy of a baby care camera based on instance segmentation detection according to claim 1, characterized in that, It also includes step three, adult region processing, where when an adult region is detected, the adult region is filled with the corresponding Gaussian blurred background region.

7. A method for protecting the privacy of a baby care camera based on instance segmentation detection according to claim 6, characterized in that, In step three, if there is no corresponding Gaussian blurred background area in the adult area, then the adult area is filled with a solid color.

8. A method for protecting the privacy of a baby care camera based on instance segmentation detection according to claim 7, characterized in that, The solid color fill in step three uses gray fill.

9. A method for protecting the privacy of a baby care camera based on instance segmentation detection according to claim 6, characterized in that, In step three, if the adult area is located in front of the infant area, the adult area will be filled with the corresponding Gaussian blurred background area.

10. A method for protecting the privacy of a baby care camera based on instance segmentation detection according to claim 6, characterized in that, In step three, if the adult area has a corresponding Gaussian blurred background area, then the adult area at the corresponding position is filled with the corresponding Gaussian blurred background area, and the remaining adult area is filled with a solid color.

Images