Video-based human respiratory rate detection method
Through the combination of OpenCV and Lucas-Kanade methods, the equipment requirements are reduced and detection accuracy is improved, and the problem of expensive and insufficient detection of existing technology equipment is solved, and low-cost and high-precision breathing frequency detection is achieved.
Patent Information
- Application Number
- CN202410976478.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-20
- Publication Date
- 2024-11-19
AI Technical Summary
The prior art equipment requirements are high and expensive when detecting the human respiratory rate, and the two-dimensional measurement area selected is a plane, resulting in insufficient sensitivity and precision in detection.
The Haar cascade classifier of OpenCV is used for face detection, combined with the Lucas-Kanade method to track feature points, use three-dimensional displacement to calculate amplitude, and fuse optical flow algorithm to detect the respiration frequency at the root of the neck to reduce equipment requirements and improve detection accuracy.
It realizes low-cost and accurate breathing frequency detection, with small errors, and improves the sensitivity and accuracy of amplitude detection through the three-dimensional detection area.
Smart Images

Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical field:
[0001] The present invention relates to the technical field of respiratory rate detection. The present invention discloses a method for detecting the respiratory rate of a human body by analyzing the base of the neck region, mainly utilizing face recognition, base of the neck detection, feature point tracking, optical flow algorithm, and combining signal processing technology to realize the function of extracting the respiratory rate from video data. The Lucas-Kanade method in the optical flow algorithm is used to track the displacement of key points in the base of the neck region, and the respiratory rate is determined by analyzing the periodic changes of these displacements. Background technology:
[0002] Source of method: There are many systems and products on the market that can detect human respiratory rate by analyzing videos. They use advanced computer vision technology and algorithms to monitor respiratory rate without contact. However, they have high requirements for equipment and are expensive.
[0003] Differences in methods: There are differences in the implementation methods between the existing system and our system, which are specifically reflected in the following aspects: (1) The existing system has high requirements for equipment and the equipment is expensive. (2) The existing system selects different measurement areas, and the regional amplitude of the existing system is smaller. (3) The area selected by the existing system is two-dimensional, while the area selected by our system is three-dimensional, which can make the amplitude detection more sensitive and accurate.
[0004] Related technologies: OpenCV's Haar cascade classifier is used for face detection, and the goodFeaturesToTrack function is used to detect corner points through the Harris method. The Lucas-Kanade method is used to track feature points and calculate the displacement of feature points between consecutive frames. This system does not require high requirements for shooting equipment. The algorithms and technologies used are existing open source algorithms with strong interpretability and relatively mature technology.
[0005] References: Methods for detecting human respiratory rate based on video, published papers and applied patents, read comprehensively to understand the method of contactless detection of respiratory rate.
[0006] For example, the patent "A respiratory rate detection method based on video signals" provides a respiratory rate detection method based on video signals. The extracted respiratory area at the base of the neck is filtered, and the respiratory movement direction characteristics of the pixels in the respiratory area are extracted; the respiratory area image is decomposed in multi-scale space based on the respiratory movement direction characteristics, and the respiratory phase information is extracted; the extracted respiratory phase information is filtered, and the peak detection is performed to estimate the respiratory rate. "Research on non-contact visual physiological sign parameter detection". The development status of non-contact visual physiological sign parameter detection technology is reviewed, and its hardware requirements, technical points and research status are analyzed from the aspects of equipment selection, face detection, pulse wave signal extraction, and physiological parameter calculation.
[0007] The key is how to combine existing technologies and methods to improve the accuracy of respiratory rate data detected through video, reduce computational pressure, and enhance robustness and tracking stability.
[0008] In summary: the current search has not found: in the method of detecting respiratory rate based on video in this system, the system similar to this system's method has not used the method of integrating feature point tracking and optical flow algorithm to detect respiratory rate in the extraction and processing of the respiratory area at the base of the neck. Summary of the invention:
[0009] The base of the neck region is determined by selecting the face with the largest area, thereby focusing only on the mask of the base of the neck region ROI.
[0010] Select three-dimensional displacement to calculate the amplitude and improve the accuracy of amplitude detection.
[0011] The feature point selection method is fused with the optical flow algorithm, which is suitable for detecting the respiratory frequency at the base of the neck. Specific implementation method:
[0012] Step 1: Integrate into the Yinrunkang health management system.
[0013] Step 2: Receive the video transmitted by the front end.
[0014] Step 3: Read video information.
[0015] Step 4: Face detection and ROI positioning.
[0016] Step 5: Corner point detection and tracking.
[0017] Step 6: Optical flow calculation and coordinate recording.
[0018] Step 7: Peak detection and respiratory rate calculation.
[0019] Step 8: The calculated structure is returned to the front end.
[0020] Step 9: Compare the respiratory rate measured by this system with the measurement results of professional institutions. It can be seen that the respiratory rate measured by this system has a small error with the results of professional institutions, indicating that our measurement method is accurate and reliable. See Table 1 for details. Description of the drawings:
[0021] Figure 1 :System architecture diagram provided in the implementation of the present invention.
[0022] Figure 2 : A system flow chart provided by the implementation of the present invention.
[0023] Figure 3 : Flow chart of the respiratory rate detection method.
[0024] Table 1: Comparison of respiratory rate measurements by this system and those by professional institutions.
Claims
1. Determine the base of neck region by selecting the face with the largest area, thus focusing only on the mask of the base of neck region ROI.
2. Select three-dimensional displacement to calculate the amplitude and improve the accuracy of amplitude detection.
3. The feature point selection method is integrated with the optical flow algorithm, which is suitable for detecting the respiratory frequency at the base of the neck.
