A tracking and identification method and system for specific trades

Abstract

The application relates to the technical field of image processing, and discloses a tracking identification method and system for specific types of work. The method divides the work type detection task into face detection and work type wearing feature detection, identifies whether the worker is within the jurisdiction through face detection, then identifies the work type of the worker through work type wearing feature identification, matches and judges whether corresponding safety protection is completed to generate image identification information. Then, the PWM signal required for the face to be located at the center of the image is calculated according to the image identification information, the gimbal steering engine is further driven according to the PWM signal, new image information is acquired for a new round of analysis, so that the face is always located at the center of the image, and tracking analysis of the whole detection process is realized. In this way, the face can be tracked and identified in a complex environment, and whether the worker meets the safety protection requirements of the work type is detected through the wearing feature.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a tracking identification method and system for specific work types. BACKGROUND

[0002] In recent years, as China pays more and more attention to labor safety, ensuring that specific workers do safety protection according to the requirements of work types has become an important task. In reality, manual supervision is often used to implement this task, but this is low in efficiency and increases operating costs, and also brings potential production hazards.

[0003] In addition, there are some recognition and detection systems in the prior art. The existing detection and recognition systems have certain limitations in function and hardware. For example, the system described in Publication No. CN221261707U only meets the requirement of face recognition and cannot detect work type wearing features, and the camera detects the target at a fixed angle, which requires the user to actively adjust their posture to adapt to the lens. Poor posture can further reduce recognition accuracy, and the adaptability to different lighting conditions in image processing is low. The system described in Publication No. CN118118785A can actively track and identify objects, but it cannot detect work type wearing features and cannot identify specific faces. In addition, it is easily disturbed when there are many people, and cannot distinguish between primary and secondary objects, resulting in inaccurate positioning. Publication No. CN116543327A designs a program to recognize faces and detect work types through a computer, but standard computers are difficult to carry and have high costs, which is not conducive to large-scale application. This requires the design of a lightweight embedded intelligent device that can run offline. In addition, most control methods in existing engineering use conventional PID controllers. Although the PID can meet certain steady-state and transient performance of the system through the functions of PID, there are still certain limitations. For example, the integral term in the conventional PID has a positive effect on the elimination of steady-state error, but the integral term has a negative impact on the transient performance of the system, increasing overshoot and causing certain hysteresis. In high-speed industrial sites and densely populated access control sites, target tracking is not in place and not timely, which can cause target loss, target misjudgment, and reduced recognition accuracy.

[0004] Therefore, it is difficult to track and identify faces in complex environments while detecting whether workers meet the safety protection requirements of work types through wearing features in the prior art. SUMMARY

[0005] The present application provides a tracking identification method and system for specific work types, which solves the technical problem that it is difficult to track and identify faces in complex environments while detecting whether workers meet the safety protection requirements of work types through wearing features in the prior art.

[0006] In order to achieve the above object, the present application is realized by the following technical scheme:

[0007] In a first aspect, the present application provides a tracking identification method for specific types of work, comprising:

[0008] S1: acquiring image information, detecting the target face and work type wearing features corresponding to the image information;

[0009] S2: performing face matching on the target face and work type wearing features with a database;

[0010] S3: judging whether the target face has done safety protection according to the work type requirements according to the matching result; and generating image recognition information;

[0011] S4: calculating the PWM signal required for the face to be in the center of the image according to the image recognition information;

[0012] S5: controlling the steering engine driving module to drive the gimbal steering engine according to the PWM signal to obtain new image information for tracking.

[0013] In a second aspect, the present application provides a tracking identification system for specific types of work, comprising:

[0014] An image processor is configured to acquire image information, detect the target face and work type wearing features corresponding to the image information, perform face matching on the target face and work type wearing features with a database, judge whether the target face has done safety protection according to the work type requirements according to the matching result, and generate image recognition information.

[0015] A main controller is configured to calculate the PWM signal required for the face to be in the center of the image according to the image recognition information.

[0016] A steering engine driving module is configured to drive the gimbal steering engine according to the PWM signal to obtain new image information for tracking.

[0017] In a third aspect, the present application provides a tracking identification system for specific types of work, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to realize the steps of the above method.

[0018] Advantages:

[0019] The tracking identification method for specific types of work provided by the application divides the work type detection task into face detection and work type wearing feature detection, identifies whether it is a worker within the jurisdiction through face detection, then identifies the work type to which the worker belongs through work type wearing feature recognition, matches and judges whether the corresponding safety protection is completed to generate image recognition information. Then, the PWM signal required to make the face in the center of the image is calculated according to the image recognition information, and the gimbal steering engine is further driven according to the PWM signal, so as to obtain new image information for a new round of analysis, so that the face is always in the center of the image, and the tracking analysis of the whole detection process is realized. In this way, through the combination of face detection, work type wearing feature detection, face matching and the way of calculating the PWM signal required to make the face in the center of the image, the face can be tracked and recognized in a complex environment, and at the same time, it is detected whether the worker meets the safety protection requirements of the work type.

[0020] In a further technical solution, the integral term is modified, an integral switching mechanism and a time regularization mechanism are introduced, so that the motion executor has good steady-state performance and transient performance. The control method is better than the conventional PID, so that the gimbal tracking has good adaptive ability and anti-interference ability.

[0021] In a further technical solution, real-time, efficient and reliable communication methods are used for data transmission between modules, real-time data transmission is performed between the system and the upper computer, and the system as a whole can operate independently without the upper computer. The man-machine interaction is simple and friendly, the information is complete and the function is complete, and voice broadcast can be performed. BRIEF DESCRIPTION OF DRAWINGS

[0022] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application, and are incorporated in and constitute a part of this application. The schematic embodiments of the application and their descriptions are used to explain the application, and do not constitute an improper limitation on the application. In the drawings:

[0023] Figure 1 is one of the flowcharts of the tracking identification method for specific types of work in the preferred embodiment of the application;

[0024] Figure 2 is the second flowchart of the tracking identification method for specific types of work in the preferred embodiment of the application;

[0025] Figure 3 is a schematic diagram of the overall hardware composition of the gimbal in the preferred embodiment of the application;

[0026] Figure 4 is a system hardware relationship diagram in the preferred embodiment of the application;

[0027] Figure 5 is a switching mechanism schematic diagram of the nonlinear integrator in the preferred embodiment of the application;

[0028] Figure 6 is a schematic diagram of an interactive interface in a preferred embodiment of the present application. DETAILED DESCRIPTION

[0029] The technical solutions of the present application will be described clearly and completely below. Obviously, the described embodiments are only some of the embodiments of the present application, but not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0030] It should be understood that the tracking and identification method for a specific type of work provided in the present application can be applied to the identification and detection of a specific type of work, such as the identification and detection of peasant workers, and the like, which are only examples and are not limited.

[0031] Referring to Figures 1-2 , the present application provides a tracking and identification method for a specific type of work, comprising:

[0032] S1: acquiring image information, detecting a target face and a type of work wearing feature corresponding to the image information;

[0033] S2: performing face matching on the target face and the type of work wearing feature with a database;

[0034] S3: judging whether the target face has completed safety protection according to the matching result; and generating image identification information;

[0035] S4: calculating a PWM signal required for the face to be in the center of the image according to the image identification information;

[0036] S5: controlling a steering engine driving module to drive a gimbal steering engine according to the PWM signal to obtain new image information for tracking.

[0037] The tracking and identification method for a specific type of work described above divides the type of work detection task into face detection and type of work wearing feature detection, identifies whether it is a worker within the jurisdiction through face detection, then identifies the type of work to which the worker belongs through the type of work wearing feature, performs matching and judges whether the corresponding safety protection is completed to generate image identification information. Then, the PWM signal required for the face to be in the center of the image is calculated according to the image identification information, and the gimbal steering engine is further driven according to the PWM signal, so as to obtain new image information for a new round of analysis, so that the face is always in the center of the image, and tracking analysis of the entire detection process is realized. In this way, through the combination of face detection, type of work wearing feature detection, face matching, and calculation of the PWM signal required for the face to be in the center of the image, the face can be tracked and identified in a complex environment, and at the same time, it can be detected whether the worker meets the safety protection requirements of the type of work through the wearing feature.

[0038] It is worth pointing out that the tracking identification method for the specific work type described above can be applied to an adaptive tracking identification system for a specific work type, and the system components include a main controller, an image processor, and a motion executor.

[0039] The overall hardware composition of the holder is shown in Figure 3 , including an OpenArt mini embedded image processor 1, a 3D printing support 2, a holder metal kit 3, and a digital rudder 4.

[0040] The overall composition of the system is shown in Figure 4 , and the specific connection relationship is: PA0 and PA1 of the TM4C1294XL development board are connected with TX and RX of the OpenArt mini respectively; the signal line of the upper rudder is connected with PK4 of the TM4C1294XL, the lower rudder is connected with PK5 of the TM4C1294XL, and the 5V and GND of the two rudders are connected to the rudder drive board; PB5, PN4, PN5, PK0 and PK1 of the TM4C1294XL development board are connected with A1, A2, A3, A4 and A5 of the MP3 voice broadcast module respectively, 5V and GND of the MP3 voice broadcast are connected to the TM4C1294XL, the Speaker pin of the MP3 voice broadcast is connected with the small power loudspeaker; the 2S lithium battery is connected with the positive and negative poles of the rudder drive board, and the 5V and GND of the rudder drive board are connected to the TM4C1294XL development board and the OpenArt mini.

[0041] The communication mode between each module is set as follows:

[0042] 1) UART is used for serial data asynchronous transmission between the TM4C1294XL development board and the OpenArt mini.

[0043] 2) The TM4C1294XL development board and the voice broadcast module use GPIO mode for data transmission, and the transmission form uses IO coding.

[0044] 3) TFT and touch screen transmit data to the TM4C1294XL development board through EPI.

[0045] 4) The TM4C1294XL development board transmits control quantity to the holder through PWM signal.

[0046] It is worth pointing out that the image processing method in the present application solves the three tasks of face detection, face matching and work type feature recognition, and it is worth noting that face detection is to identify the target that meets the face biological characteristics, and face matching is to identify a specific face on the basis of detecting the face.

[0047] For face detection and workwear feature recognition, a connected component algorithm based on RGB model is adopted, and the steps are as follows:

[0048] 1) First, the threshold editor is used to filter out the threshold of workwear features (mask, safety helmet, etc.) under the RGB model. The threshold range with better segmentation effect is recorded as the target threshold zone.

[0049] 2) A label matrix with the same size as the original image is created to store the connected component label of each pixel. All elements in the initial label matrix are set to 0.

[0050] 3) In the RGB model, if the threshold of R, G, and B channels of a pixel meets the target threshold zone, it is called a foreground pixel under the RGB model, otherwise it is a background pixel.

[0051] 4) Traverse the image pixel by pixel I , check the neighbor pixels of each foreground pixel based on the 4-neighborhood model:

[0052] For 4-neighborhood, check pixels x-1, y ) and x, y+1 ). Let the label set of the neighborhood pixels be S , then:

[0053] ;

[0054] In the formula, L denotes the image matrix.

[0055] According to the case of S, the processing is as follows:

[0056] If S is empty (i.e. all neighbors are background pixels), assign a new label l to L ( x , y ):

[0057] ;

[0058] If S is not empty, select the smallest label :

[0059] ;

[0060] 5) Perform a second traversal of the entire image, and use the union-find set structure to record and merge equivalent labels. Each equivalence class represents a set of mutually equivalent labels, and for each equivalence class, find its representative label (typically the smallest label value), and update the labels in the class to .

[0061] 6) Further, the connected domain under certain RGB threshold is outputted, and the obtained connected domain is further subjected to area judgment to distinguish the primary and secondary objects and exclude interference, so as to finally obtain the connected domain belonging to the face and the work type wearing feature; in addition, the perimeter, area and geometric center of the connected domain can be extracted.

[0062] In the embodiment, the RGB threshold refers to the RGB threshold range belonging to the face and the work type feature, which is the image attribute of the face and the work type feature. Then the connected domain is found through the threshold range. The connected domain actually refers to the area belonging to the face and the work type wearing feature in the image.

[0063] For face matching, LBP is used for face image texture description and feature extraction, and the steps are as follows:

[0064] 1) The collected image is converted from a color image to a gray image.

[0065] 2) A comparison image library is made through the continuous shooting function, 20-100 images are made and stored in a memory card, and the shooting angles of each image are required to be different, so as to achieve the purpose of data augmentation.

[0066] 3) For each pixel I of the image I ( x, y ), an 8-neighborhood method is used to compare the gray values of the center pixel and the surrounding pixels. If the gray value of the surrounding pixel is greater than or equal to the gray value of the center pixel, it is recorded as 1, otherwise it is recorded as 0. The top-left pixel is recorded as the 0th pixel, and the numbering is performed clockwise. I ( x, y ) LBP feature value is expressed by the following formula:

[0067] ;

[0068] Wherein, P is the number of field pixels. Here, the value is 8, is the gray value of the th pixel around the pixel i .

[0069] 4) After obtaining the feature map of the current image about LBP value, the error is calculated with each picture of the pre-acquired target face library and the error sum of squares is obtained, the error sum of squares obtained by comparing each picture in the library is accumulated, and compared with the set threshold, if less than the set threshold, it is considered that the face matching is correct. Correct matching means that the face in the current image is the face described by the target face library. In other words, there is a total library, the total library includes a sub-library, each sub-library represents a specific face, and the current image is compared with these sub-libraries respectively, and in the comparison with a certain sub-library: the feature map is compared with each picture of the current sub-library to calculate the error and obtain the error sum of squares, the accumulated sum of the error sum of squares obtained by comparing each picture in the library is calculated, and the accumulated sum is compared with the set threshold, and the accumulated sum is less than the set threshold, then it is considered that the face in the current image matches the face represented by the current sub-library.

[0070] The method is a relatively lightweight and efficient recognition method, and is not sensitive to illumination changes.

[0071] 5) Further, OpenART mini transmits information to TM4C1294XL master through UART, including the geometric center coordinates (X, Y) of the target, whether a face is detected, whether it is a target person, and whether the target person has done safety protection according to the requirements of the type of work; OpenART mini marks the geometric center of the recognized object with a cross in the image transmitted to the PC end, and marks the pixel range of the recognized object with an external rectangle.

[0072] 6) Finally, save the program file to OpenART mini, complete the burning, and run offline.

[0073] Further, the PWM signal required for the face to be located at the center of the image is calculated according to the image recognition information, specifically as follows:

[0074] In this step, the switching mechanism of the shunt nonlinear integrator is as shown in the accompanying Figure 5

[0075] Preferably, in order to make the tracking control process have strong adaptability and anti-interference ability, a shunt nonlinear integrator is introduced on the basis of the position type PID. The input of the system is the actual position of the recognized object in the image, and the output is the PWM required for controlling the rudder. The control target of the gimbal is to control the up and down rudders using the control amount calculated by the master control, so that the rudders deflect in two directions in parallel to reduce the error, and finally the center of the camera lens is aligned with the geometric center of the face.

[0076] In a discrete system, the position type PID is:

[0077] ;

[0078] ​ ;

[0079] in, This is the proportionality coefficient. The integral coefficient is... The differential coefficients are... for k The amount of control at any given moment for k Time error, e ( i ) represents intermediate computational quantities, indicating that in i The error between the target position and the actual position at any given time. To identify the target location of the geometric center of the object in the image, To identify the actual location of the geometric center of an object in the image, ,but k The time error is the difference between the target position and the actual position of the object being identified in the X / Y direction of the image sensor.

[0080] While the linear integrator in the conventional positional PID controller described above can improve the system's steady-state performance, it also introduces phase lag and overshoot. Therefore, it is necessary to improve the linear integrator to incorporate features from the positional PID controller. k The integral term at time t is denoted as The control law of the shunt nonlinear integrator can then be rewritten as:

[0081] ;

[0082] ;

[0083] In the formula, I S ( k ) is in k Nonlinear integral term of the branch at time step ， For control inputs with branched nonlinear integral terms, For logical "OR" operator, For logical AND operator, The time elapsed since the most recent switch. Minimum handover time interval This is the tilt factor.

[0084] Combined with appendix Figure 5 The above controller first considers and The symbolic relation, in When crossing the 0 boundary, the integral direction is judged and changed to alleviate the overshoot phenomenon and accelerate convergence. However, the switching mechanism has the cost that when the system tends to be stable, small disturbances make the 0 boundary frequently, causing the integrator to frequently switch. Therefore, a tilt factor is introduced to make the switching boundary deviate from the 0 boundary and tilt a certain amount, when stable within the specified range of the switching boundary, the integrator does not switch. Then, a time regularization mechanism is introduced, the time length from the last switching to the current time, the minimum switching time interval must be satisfied before switching , and the rest of the case does not switch. This further avoids the integrator from frequently switching in a limited time.

[0085] In summary, the control method deployment includes: the actual position of the target in the X-axis direction of the picture, the actual position of the target in the Y-axis direction of the picture; let the control amount in the X-axis direction of the picture, the control amount in the Y-axis direction of the picture. Two sets of position type PID parallel control of the upper and lower rudders are used, which are of the same structure and have shunt nonlinear integrators, so that , the input of the main controller is compared with the reference value, the main control calculates the current , , according to the actual physical meaning of PID, the parameter range is gradually narrowed through the control variable method and the bisection method, and then the best PID parameters are selected according to the overshoot, adjustment time and other indicators of the control waveform, and then is calculated, which is the required PWM signal. Among them, the PID control input is the PWM signal.

[0086] The system control can guarantee certain steady-state performance and good transient performance, has strong self-adaptability and anti-interference ability, and finally the gimbal achieves the control target in the direction, that is, the camera center is aligned with the picture center.

[0087] In addition, the tracking identification method for specific types of work provided by the application also has a man-machine interaction function, which is specifically as follows:

[0088] In an example, in combination with Figure 2 and Figure 6The rows are described as follows: in the human-computer interaction interface, row 1 is a target position coordinate, row 2 is a position error, row 3 is a PWM output value, row 9 is a prompt, rows 4, 5 and 6 are state flag bits, wherein the state flag bits include whether a face is detected, whether a workwear feature is detected, and whether a face is matched; the interrupt program continuously scans the touch screen while performing ADC conversion to obtain touch point coordinates, virtual buttons are set in a specific area, and rows 7 and 8 are respectively used to control the gimbal to reset and unlock.

[0089] The voice broadcast module adopts an MP3 voice broadcast module and is triggered through an IO. According to information transmitted by the main controller, the recognition results of the face and the workwear feature are output and voice broadcast is performed. The broadcast content includes the following: an initial broadcast, a face recognition result broadcast after the recognition is completed, and a safety recognition broadcast after the face information is determined and the work type is matched. In addition, the voice broadcast content can be freely modified, and the volume can be adjusted.

[0090] It is worth emphasizing that the system supports various types of single-chip microcomputers and peripheral modules. The structure is simple and efficient, and the system mechanical structure can be built using ordinary metal parts. Thanks to the control method, a small torque and low frequency steering engine can meet the system task. The hardware drive requirement is low, and no additional drive module is needed. Except for the steering engine drive module, only general signal lines are needed to complete the cooperation between modules.

[0091] The system takes OpenART mini as the system core, supports replacing the main controller and peripherals with different configurations to further adjust the system development cost, and responds to development needs. Thanks to the system program packaging and compatibility, in another embedded system equipped with OpenART mini, the corresponding program blocks of each module are transplanted and the hardware interface is modified, and each module is connected according to the system architecture of the application, so that the system can be used.

[0092] The application also provides a tracking and identification system for a specific work type, which comprises:

[0093] An image processor is configured to acquire image information, detect a target face and a workwear feature corresponding to the image information, perform face matching on the target face and the workwear feature with a database, determine whether the target face is well protected according to a work type requirement according to a matching result, and generate image recognition information.

[0094] A main controller is configured to calculate a PWM signal required for the face to be located at the center of an image according to the image recognition information.

[0095] A steering engine drive module is configured to drive a gimbal steering engine according to the PWM signal to obtain new image information for tracking.

[0096] In an example, the main controller can adopt the TM4C1294XL development board of TI. The processing information transmitted by the image processor is received, the PWM required for the face to be in the center of the image is calculated according to the image information, and is output to the servo drive module to drive the gimbal servo; the image information and the detection result are displayed on the touch screen, and the user resets the gimbal and unlocks the gimbal through the virtual buttons on the screen; the voice module is controlled based on the image detection result.

[0097] The image processor can adopt the OpenART mini of SeekFree. The face is located through a classic image processing method, if the face is recognized, the face is matched through the face database in the memory card and the recognition result is output; at the same time, it is also judged whether the target person has done safety protection according to the requirements of the type of work; the image recognition information is transmitted to the main control. The image processor can be connected to a computer for real-time image stream transmission, and the computer terminal can be called online to output information through a serial port; the geometric center of the recognition object in the transmitted image is marked with a cross, and the pixel range of the recognition object is marked with a rectangle; the image processor has an image acquisition function, which can automatically take photos with light compensation, continuous shooting and image storage.

[0098] The voice broadcast module can adopt an MP3 voice playback module. The recognition results of the face and the type of work wearing features are output and voice broadcast according to the information transmitted by the main control; the voice broadcast content can be freely modified, and the volume can be adjusted.

[0099] The servo drive module: convert 7.4V voltage to 5V voltage for servo and mainboard power supply, and connect the PWM signal pin output to the servo connected to the main control.

[0100] The gimbal: two digital servos are used and placed orthogonally.

[0101] Among them, UART is used for serial data asynchronous transmission between the main controller and the image processor; the main controller and the voice broadcast module are connected through GPIO for data transmission, the transmission form adopts IO coding, which can represent multiple situations with limited pins, and the transmission efficiency is high; the TFT and the touch screen transmit data to the main controller through EPI.

[0102] The above system has strong compatibility, and the software and hardware are designed in a modular way, which ensures low coupling between modules, provides support for users to adjust the system, and also reduces the adverse effects of single function failure on the system. The system is compatible with most embedded chips as the main control, and the codes of each module are independently encapsulated and have perfect interfaces. By transplanting the corresponding program blocks of each module and modifying the hardware interface, the system can be used in another embedded system with OpenART mini, so it is easier for users to replace hardware configurations to further adjust the system development cost to meet different development needs.

[0103] The application further provides a tracking and identification system for a specific type of work, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps of the above method when executing the computer program. The tracking and identification system for a specific type of work can implement each embodiment of the above tracking and identification method for a specific type of work and achieve the same beneficial effects, which will not be described here.

[0104] To sum up, the tracking and identification system for a specific type of work provided by the application can track a target, collect and match face information, and then detect whether a worker has done safety protection corresponding to the type of work. The system is composed of four main parts, i.e., a main controller, an image processor, a motion executor, and a man-machine interaction. A user resets and unlocks the system through a touch-type man-machine interaction. The image processor OpenART mini completes the recognition of a face and a type of work through a visual processing method. After receiving image information, the main controller TM4C1294XL adopts a position-type PID combined with a shunt nonlinear integrator to control the motion of a gimbal servo motor, so as to improve the transient performance and steady-state performance of the system and to improve the adaptive ability and anti-interference ability of the system. Finally, the system autonomously completes real-time and accurate tracking and positioning, efficiently and accurately recognizes face information and a type of work wearing feature, and instructs a user through a voice module.

[0105] The above only describes the preferred embodiments of the application and is not used to limit the application. For those skilled in the art, the application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A tracking identification method for a specific job, characterized by, The method comprises: S1: acquiring image information, detecting a target face and work type wearing feature corresponding to the image information; S2: performing face matching on the target face and work type wearing feature and a database; S3: judging whether the target face is well protected according to the work type requirement according to the matching result; and generating image recognition information; S4: calculating a PWM signal required for the face to be in the image center according to the image recognition information; S5: controlling a steering engine driving module to drive a gimbal steering engine to track new image information acquired as a target according to the PWM signal; The image recognition information comprises a geometric center coordinate of the target, whether the face is detected, whether it is the target face, and whether the target face is well protected according to the work type requirement; The S4 comprises: inputting the geometric center coordinate of the target into a main controller to calculate a corresponding PWM signal as follows: The S1 comprises: ； ； wherein, is a proportional coefficient, is an integral coefficient, is a differential coefficient, is a control amount at time k, is is an error at time k, denotes an error between a target position and an actual position of the recognition object at time i, is a target position of a geometric center of the recognition object in the image, is an actual position of the geometric center of the recognition object in the image, then is an error at time k, which is a difference between a target position and an actual position of the recognition object in the X / Y direction of the image sensor; The above position-type PID is modified as k The integral term at time t is denoted as , The control law with the shunt nonlinear integrator is as follows: ； ； In the formula, I S k is the shunt nonlinear integral term at the time instant k ， is the control input with the shunt nonlinear integral term, is a logical OR operator, is a logical AND operator, is the time length from the last switching to the current time instant, is the minimum switching time interval, is a tilt factor;​​ Record The actual position of the target in the X-axis direction of the picture, The actual position of the target in the Y-axis direction of the picture; Record The control amount in the X-axis direction of the picture, The control amount in the Y-axis direction of the picture; using two sets of position PID parallel control of the same structure with shunt nonlinear integrator, the upper and lower steering gears, , The input main controller and compared with the reference value, the main controller calculates the current , , according to the actual physical meaning of PID parameter debugging, through the control variable method, dichotomy gradually narrowing the parameter range, and then according to the control waveform overshoot, adjustment time to select the best PID parameters, and then calculate The required PWM signal.

2. The tracking identification method for specific job according to claim 1, characterized in that, S12: in an RGB model, if the RGB three channels of a pixel respectively meet a set threshold range, the pixel is regarded as a foreground pixel under the RGB model, otherwise the pixel is regarded as a background pixel; S11: create a label matrix of the same size as the image information, all elements in the initial label matrix are set to 0, and the label matrix is used to store the connected domain label of each pixel of the image I a label matrix of the same size as the image, all elements in the initial label matrix are set to 0, and the label matrix is used to store the connected domain label of each pixel of the image S15: outputting a connected domain under a set RGB threshold range, and performing area judgment on the obtained connected domain to finally obtain a connected domain to which a face and a work type wearing feature belong, wherein the set RGB threshold is determined according to the face and the work type wearing feature. S13: iterate through each pixel to finally iterate through the image I check neighbor pixels of each foreground pixel based on 4-neighborhood S14: traversing the image I performing a second traversal, using a union-find structure to record and merge equivalent labels, each equivalence class representing a group of mutually equivalent labels, finding the smallest label value for each equivalence class as a representative label updating the labels in the class to ; The S2 comprises:

3. The tracking identification method for specific job according to claim 1, characterized in that, S21: performing gray scale processing on the real-time acquired image; x, y S22: For each pixel in the image I x, y ), using the 8-neighborhood method, compare it with the gray value of the surrounding pixels, if the gray value of the surrounding pixels is greater than or equal to the gray value of the center pixel, it is recorded as 1, otherwise it is recorded as 0, record the top left corner pixel as the 0th pixel, and number clockwise, calculate the LBP feature value of I S23: obtaining a feature map of the current image about the LBP value, calculating an error sum of squares by comparing the feature map with a pre-acquired target face library, and calculating an accumulation sum of the error sum of squares obtained by comparing each picture in the library, comparing the accumulation sum with a set threshold, and considering that the face in the current image is correctly matched with the face in the library when the accumulation sum is less than the set threshold. ) that satisfies the following relationship;​​ ； wherein, P is the number of field pixels, is the gray value of the pixel surrounding the first i pixel; The S3 comprises:

4. The tracking identification method for specific job according to claim 1, characterized in that, Taking a mask, a safety helmet and a protective clothing as recognition features, and recognizing whether the target face is well protected according to the work type requirement based on the recognition features; and generating image recognition information. The method further comprises:

5. The tracking identification method for specific job according to claim 1, characterized in that, Performing voice broadcast according to the image recognition information. The method comprises:

6. A tracking and identification system for a specific job, characterized in that, An image processor is configured to acquire image information, and detect a target face and work type wearing feature corresponding to the image information; Perform face matching on the target face and work type wearing feature and a database, judge whether the target face is well protected according to the work type requirement according to a matching result, and generate image recognition information; A main controller is configured to calculate a PWM signal required for the face to be in the image center according to the image recognition information; A steering engine driving module drives a gimbal steering engine according to the PWM signal to track new image information acquired as a target; The image recognition information comprises a geometric center coordinate of the target, whether the face is detected, whether it is the target face, and whether the target face is well protected according to the work type requirement; The main controller is configured to calculate a PWM signal required for the face to be in the image center according to the image recognition information, and comprises: inputting the geometric center coordinate of the target into a main controller to calculate a corresponding PWM signal as follows: ​ The set position type PID is as follows: ； ； wherein, is a proportional coefficient, is an integral coefficient, is a differential coefficient, is a control amount at time k, is is an error at time k, denotes an error between a target position and an actual position of the recognition object at time i, is a target position of a geometric center of the recognition object in the image, is an actual position of the geometric center of the recognition object in the image, then is an error at time k, which is a difference between a target position and an actual position of the recognition object in the X / Y direction of the image sensor; The above position-type PID is modified as k The integral term at time t is denoted as , The control law with the shunt nonlinear integrator is as follows: ； ； In the formula, I S k is the shunt nonlinear integral term at the time instant k ， is the control input with the shunt nonlinear integral term, is a logical OR operator, is a logical AND operator, is the time length from the last switching to the current time instant, is the minimum switching time interval, is a tilt factor;​​ Record The actual position of the target in the X-axis direction of the picture, The actual position of the target in the Y-axis direction of the picture; Record The control amount in the X-axis direction of the picture, The control amount in the Y-axis direction of the picture; using two sets of position type PID parallel control of the upper and lower steering gears with shunt nonlinear integrators of the same structure, , The input main controller and compared with the reference value, the main controller calculates the current , , according to the actual physical meaning of PID parameter debugging, through the control variable method, dichotomy gradually narrowing the parameter range, and then according to the control waveform overshoot, adjustment time to select the best PID parameters, and then calculate The required PWM signal.

7. The tracking identification system for specific trades as claimed in claim 6 wherein, Also include: A voice broadcast module for voice broadcast according to the image recognition information, the main controller and the image processor adopt UART for serial data asynchronous transmission; the main controller and the voice broadcast module adopt GPIO mode for data transmission, and the transmission form adopts IO coding.

8. A tracking and identification system for a specific job, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the method in any one of claims 1 to 5.

Images