CN109858341B - Rapid multi-face detection and tracking method based on embedded system - Google Patents
Rapid multi-face detection and tracking method based on embedded system Download PDFInfo
- Publication number
- CN109858341B CN109858341B CN201811580249.9A CN201811580249A CN109858341B CN 109858341 B CN109858341 B CN 109858341B CN 201811580249 A CN201811580249 A CN 201811580249A CN 109858341 B CN109858341 B CN 109858341B
- Authority
- CN
- China
- Prior art keywords
- face
- image data
- face image
- position information
- embedded system
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a rapid multi-face detection and tracking method based on an embedded system.A camera module and a special coprocessor or a VPU are embedded in the embedded system, the embedded system is initialized and configured, and a global buffer queue containing face corresponding position information is obtained through initialization; acquiring face image data through a camera module, preprocessing the face image data, inputting the preprocessed face image data into a special coprocessor or a VPU (virtual private Unit), obtaining position information corresponding to the face image data by the special coprocessor or the VPU, comparing the position information with position information in a global buffer queue, and judging whether position information matched with the position information exists in the global buffer queue or not; finally, processing the global buffer queue of all the obtained face image data to obtain the face image data with the optimal quality, and outputting and storing the face image data; the invention can effectively improve the processing efficiency of the face image data.
Description
Technical Field
The invention belongs to the technical field of computer vision, and particularly relates to a rapid multi-face detection and tracking method based on an embedded system.
Background
Based on cost and power consumption considerations, the embedded device, especially the monitoring camera, mostly uses a low-cost ARM processor as a core processing platform of the system. Compared with the x86 platform, the ARM processor uses the RISC instruction set, and has high execution efficiency and less power consumption. However, the ARM processor has low main frequency, limited computing resources and weak computing capability, and is difficult to complete complex computing work.
The traditional monitoring camera mainly completes video coding and decoding, transmission and other software level work. With the rapid development of artificial intelligence technology, more and more AI functions, such as face recognition technology, are integrated into the front-end monitoring camera. The face recognition technology based on video streaming requires first parsing the video stream, detecting a face from a video frame, determining face position information, and performing subsequent face tracking and other recognition work according to the initial face position information. The traditional face recognition algorithm has low recognition accuracy, cannot process in real time and cannot meet the actual combat requirements of users.
The problem of low recognition accuracy can be well solved by the face recognition algorithm based on deep learning, however, a deep learning network needs a large amount of computing resources, and the ARM processor is singly used for reasoning work such as face detection and face quality assessment, so that the problems of low processing speed and low processing efficiency can be caused, the real-time performance of the system cannot be guaranteed, the recognition effect is poor, and the user experience is also poor.
Disclosure of Invention
Aiming at the problems of low face recognition accuracy and large recognition calculation amount in the prior art, the invention provides a rapid multi-face detection and tracking method based on an embedded system, and the specific technical scheme of the method is as follows:
a fast multi-face detection and tracking method based on embedded system, the said embedded system includes a camera module used for face detection and face tracking and a special coprocessor used for processing and calculating the face detection data, including the step:
s1, initializing the embedded system, and initializing a global buffer queue for obtaining face data;
s2, setting a detection period with preset duration, and collecting face image data by adopting the camera module;
s3, preprocessing the face image data and transmitting the preprocessed face image data to the special coprocessor, and acquiring the position information of a face corresponding to the face image data by the special coprocessor;
s4, matching the position information with the global buffer queue, judging whether the same position information exists or not, if so, adding the corresponding position information into the corresponding global buffer queue, otherwise, establishing a new global buffer queue corresponding to the position information;
s5, repeatedly using the camera module to collect face image data in the detection period, transmitting the face image data to the special coprocessor after preprocessing operation, executing face tracking by the special coprocessor, and acquiring position information of a face corresponding to the face image data;
s6, repeatedly using the camera module to collect the face image data within the detection period time and repeating the step S5;
s7, before the detection period is finished, processing all the face image data by adopting the special coprocessor, and acquiring the face image data with optimal quality of different faces in the detection period based on a deep learning network;
and S8, detecting face data which can be acquired by the camera module, outputting and storing the face image data which leaves the acquisition range of the camera module from the special coprocessor into a specified database, and deleting the corresponding global buffer queue.
Preferably, the step S4 further includes the steps of: and numbering the global buffer queue corresponding to each person by adopting the special coprocessor.
Preferably, the dedicated coprocessor can be replaced by a VPU.
Preferably, the method further comprises: setting the detection period of a new preset duration, and repeating the steps S2-S8.
Compared with the prior art, the rapid multi-face detection and tracking method based on the embedded system has the beneficial effects that: the VPU or the coprocessor is used for executing the deep learning network reasoning operation with large calculation amount, so that the processing pressure of the original ARM main processor is reduced; the VPU or the coprocessor has low power consumption in operation, the reasoning operation execution speed is high, more human face recognition operation steps can be completed in unit time, more detailed posture and angle information of a person in the process of traveling can be obtained, whether the person is shielded or not can be obtained, the human face tracking is more accurate, and the phenomenon that the position of the human face jumps to cause the drift of the human face is avoided; the human face quality evaluation can carry out all-around judgment on human face posture information, whether the human face is shielded, attribute information such as age and gender and the like of the appeared human face, and can output a human face image with optimal quality or periodically transmit the human face image with optimal quality to the background processing center according to the requirement. The optimal quality face image improves the accuracy of face comparison of the background processing center, reduces the network bandwidth consumption and reduces the construction and operation cost.
Drawings
Fig. 1 is a flow chart illustration of the embedded system-based fast multi-face detection and tracking method in the embodiment of the invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention.
Referring to fig. 1, in an embodiment of the present invention, a fast multi-face detection and tracking method based on an embedded system is provided, where the embedded system includes a camera module for face detection and face tracking and a special coprocessor for processing and calculating face detection data, and includes the steps of:
s1, initializing the embedded system, and initializing a global buffer queue for obtaining face data;
s2, setting a detection period with preset duration, and collecting face image data by adopting a camera module;
s3, preprocessing the face image data and then transmitting the face image data to a special coprocessor, and acquiring the position information of a face corresponding to the face image data by the special coprocessor;
s4, matching the position information with a global buffer queue, judging whether the same position information exists or not, if so, adding the corresponding position information into the corresponding global buffer queue, otherwise, establishing a new global buffer queue corresponding to the position information;
s5, repeatedly using the camera module to collect face image data in the detection period, transmitting the face image data to the special coprocessor after preprocessing operation, executing face tracking by the special coprocessor, and acquiring position information of a face corresponding to the face image data;
s6, repeatedly using the camera module to collect the face image data within the detection period time and repeating the step S5;
s6, before the detection period is finished, processing all face image data by adopting a special coprocessor, and acquiring the face image data with the optimal quality of different faces in the detection period based on a deep learning network;
and S7, detecting face data which can be acquired by the camera module, outputting and storing the face image data which leaves the acquisition range of the camera module from the special coprocessor to a specified database, and deleting the corresponding global buffer queue.
In the specific operation process, the method processes the face image data which can be acquired by all the camera modules in a mode of repeatedly setting the same or different detection periods and repeating the steps S2-S8.
In the embodiment of the invention, in order to process the face image data of the same person more quickly, the invention adopts the special coprocessor to number the global buffer queue corresponding to each person, for example, all the acquired face image data of Zhang III are numbered as I, and subsequently, if all the position information of Zhang III is needed, all the relevant data of the position information of Zhang III acquired by the camera module can be acquired by inputting the number I, so that the processing data of the embedded system can be effectively improved, and the requirement of the embedded system on the memory is reduced.
Preferably, the special coprocessor can be replaced by a VPU in the present invention, and embodiments of the present invention include, but are not limited to, movidius from Intel and other special devices or modules providing deep learning network operation acceleration; the specific choice of the special coprocessor can be determined according to the actual situation, which is not limited and fixed by the present invention.
Compared with the prior art, the rapid multi-face detection and tracking method based on the embedded system has the beneficial effects that: the VPU or the coprocessor is used for executing the deep learning network reasoning operation with large calculation amount, so that the processing pressure of the original ARM main processor is reduced; the VPU or the coprocessor has low power consumption in operation, the reasoning operation execution speed is high, more human face recognition operation steps can be completed in unit time, more detailed posture and angle information of a person in the process of traveling can be obtained, whether the person is shielded or not can be obtained, the human face tracking is more accurate, and the phenomenon that the position of the human face jumps to cause the drift of the human face is avoided; the human face quality evaluation can carry out all-around judgment on human face posture information, whether the human face is shielded, attribute information such as age and gender and the like of the appeared human face, and can output a human face image with optimal quality or periodically transmit the human face image with optimal quality to the background processing center according to the requirement. The optimal quality face image improves the accuracy of face comparison of the background processing center, reduces the network bandwidth consumption and reduces the construction and operation cost.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications may be made to the embodiments described in the foregoing detailed description, or equivalent changes may be made in some of the features of the embodiments described above. All equivalent structures made by using the contents of the specification and the attached drawings of the invention can be directly or indirectly applied to other related technical fields, and are also within the protection scope of the patent of the invention.
Claims (4)
1. A fast multi-face detection and tracking method based on embedded system is characterized in that the embedded system comprises a camera module for face detection and face tracking and a special coprocessor for processing and calculating face detection data, comprising the following steps:
s1, initializing the embedded system, and initializing a global buffer queue for obtaining face data;
s2, setting a detection period with preset duration, and collecting face image data by adopting the camera module;
s3, preprocessing the face image data and transmitting the preprocessed face image data to the special coprocessor, and acquiring the position information of a face corresponding to the face image data by the special coprocessor;
s4, matching the position information with the global buffer queue, judging whether the same position information exists or not, if so, adding the corresponding position information into the corresponding global buffer queue, otherwise, establishing a new global buffer queue corresponding to the position information;
s5, repeatedly using the camera module to collect face image data in the detection period, transmitting the face image data to the special coprocessor after preprocessing operation, executing face tracking by the special coprocessor, and acquiring position information of a face corresponding to the face image data;
s6, repeatedly using the camera module to collect the face image data within the detection period time and repeating the step S5;
s7, before the detection period is finished, processing all the face image data by adopting the special coprocessor, and acquiring the face image data with optimal quality of different faces in the detection period based on a deep learning network;
and S8, detecting face data which can be acquired by the camera module, outputting and storing the face image data which leaves the acquisition range of the camera module from the special coprocessor into a specified database, and deleting the corresponding global buffer queue.
2. The embedded system based fast multi-face detection and tracking method according to claim 1, wherein the step S4 further comprises the steps of: and numbering the global buffer queue corresponding to each person by adopting the special coprocessor.
3. The embedded system based fast multi-face detection and tracking method of claim 1, wherein the dedicated coprocessor can be replaced by a VPU.
4. The embedded system based fast multi-face detection and tracking method of claim 1, wherein the method further comprises: setting the detection period of a new preset duration, and repeating the steps S2-S8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811580249.9A CN109858341B (en) | 2018-12-24 | 2018-12-24 | Rapid multi-face detection and tracking method based on embedded system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811580249.9A CN109858341B (en) | 2018-12-24 | 2018-12-24 | Rapid multi-face detection and tracking method based on embedded system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109858341A CN109858341A (en) | 2019-06-07 |
| CN109858341B true CN109858341B (en) | 2021-05-25 |
Family
ID=66892022
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811580249.9A Active CN109858341B (en) | 2018-12-24 | 2018-12-24 | Rapid multi-face detection and tracking method based on embedded system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109858341B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110942545B (en) * | 2019-11-20 | 2022-04-05 | 中船重工(武汉)凌久高科有限公司 | Dense person entrance guard control system and method based on face recognition and video fence |
| CN111368732B (en) * | 2020-03-04 | 2023-09-01 | 阿波罗智联(北京)科技有限公司 | Method and device for detecting lane lines |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8649563B2 (en) * | 2011-05-09 | 2014-02-11 | Google Inc. | Object tracking |
| WO2015005964A1 (en) * | 2013-07-08 | 2015-01-15 | Augmented Reality Lab LLC | System and method for computer vision item recognition and target tracking |
| CN106228628A (en) * | 2016-07-15 | 2016-12-14 | 腾讯科技（深圳）有限公司 | System, the method and apparatus of registering based on recognition of face |
| CN106598226A (en) * | 2016-11-16 | 2017-04-26 | 天津大学 | UAV (Unmanned Aerial Vehicle) man-machine interaction method based on binocular vision and deep learning |
| CN107944401A (en) * | 2017-11-29 | 2018-04-20 | 合肥寰景信息技术有限公司 | The embedded device for tracking and analyzing with multiple faces dynamic |
| CN207558028U (en) * | 2017-10-03 | 2018-06-29 | 杭州强文智能科技有限公司 | Intelligent bus device of boarding |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8995715B2 (en) * | 2010-10-26 | 2015-03-31 | Fotonation Limited | Face or other object detection including template matching |
-
2018
- 2018-12-24 CN CN201811580249.9A patent/CN109858341B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8649563B2 (en) * | 2011-05-09 | 2014-02-11 | Google Inc. | Object tracking |
| WO2015005964A1 (en) * | 2013-07-08 | 2015-01-15 | Augmented Reality Lab LLC | System and method for computer vision item recognition and target tracking |
| CN106228628A (en) * | 2016-07-15 | 2016-12-14 | 腾讯科技（深圳）有限公司 | System, the method and apparatus of registering based on recognition of face |
| CN106598226A (en) * | 2016-11-16 | 2017-04-26 | 天津大学 | UAV (Unmanned Aerial Vehicle) man-machine interaction method based on binocular vision and deep learning |
| CN207558028U (en) * | 2017-10-03 | 2018-06-29 | 杭州强文智能科技有限公司 | Intelligent bus device of boarding |
| CN107944401A (en) * | 2017-11-29 | 2018-04-20 | 合肥寰景信息技术有限公司 | The embedded device for tracking and analyzing with multiple faces dynamic |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109858341A (en) | 2019-06-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764048B (en) | Face key point detection method and device | |
| CN106875422B (en) | Face tracking method and device | |
| CN111062239A (en) | Human body target detection method and device, computer equipment and storage medium | |
| CN112862874B (en) | Point cloud data matching method and device, electronic equipment and computer storage medium | |
| CN109993137A (en) | A kind of fast face antidote based on convolutional neural networks | |
| CN109325456A (en) | Target identification method, device, target identification equipment and storage medium | |
| CN109858341B (en) | Rapid multi-face detection and tracking method based on embedded system | |
| CN104616002A (en) | Facial recognition equipment used for judging age groups | |
| WO2022242127A1 (en) | Image feature extraction method and apparatus, and electronic device and storage medium | |
| CN114886404B (en) | Electronic equipment, device and storage medium | |
| CN115393597A (en) | Semantic segmentation method and device based on pulse neural network and laser radar point cloud | |
| CN113379045A (en) | Data enhancement method and device | |
| CN114783597B (en) | Method and device for diagnosing multi-class diseases, electronic equipment and storage medium | |
| Lu et al. | Dynamic offloading on a hybrid edge–cloud architecture for multiobject tracking | |
| CN111127355A (en) | Method for finely complementing defective light flow graph and application thereof | |
| CN111126456A (en) | Processing method, device and equipment of neural network model and storage medium | |
| CN115984093A (en) | Depth estimation method based on infrared image, electronic device and storage medium | |
| CN113807330B (en) | Three-dimensional sight estimation method and device for resource-constrained scene | |
| CN115311723A (en) | Living body detection method, living body detection device and computer-readable storage medium | |
| CN113222989A (en) | Image grading method and device, storage medium and electronic equipment | |
| CN112053386A (en) | Target tracking method based on depth convolution characteristic self-adaptive integration | |
| CN111611872A (en) | Novel binocular vision vehicle detection method and system | |
| Li et al. | An FPGA-based tree crown detection approach for remote sensing images | |
| CN103035012A (en) | Target tracking method and device | |
| CN113255408B (en) | Behavior recognition method, behavior recognition device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: Room 108, 1st floor, 30 Building, No. 5 Jiangtai Road, Chaoyang District, Beijing, 100020 Applicant after: Beijing Pengsi Technology Co., Ltd Address before: 716, 6th Floor, No. 2 Building, Xinfang Road, Chaoyang District, Beijing 100020 Applicant before: Beijing Pengsi Intelligent Technology Co., Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |