A vision-based agricultural robot control system and navigation path extraction method

Abstract

This invention discloses a vision-based agricultural robot control system and navigation path extraction method, comprising: a vision processing module, a controller, a power supply module, a motor drive module, an infrared reflection sensor, an ultrasonic sensor, and a peristaltic pump module. Crop information is collected by a camera and transmitted to the vision processing module. The vision processing module processes the data to obtain the navigation centerline and transmits the required navigation parameters to the controller, thereby controlling the motor speed and achieving motion control of the agricultural robot. Simultaneously, the infrared reflection sensor detects crops, and the detected parameters are transmitted to the controller. The controller then controls the peristaltic pump module to spray pesticides, and the ultrasonic sensor performs obstacle avoidance to ensure the safety of the agricultural robot. This invention exhibits high accuracy, low processing time, good adaptability, and robustness in various crop environments.

Description

Technical Field

[0001] This invention relates to the field of agricultural robot navigation and control, specifically to a vision-based agricultural robot control system and navigation path extraction method. Background Technology

[0002] Since the beginning of the 21st century, intelligent technology has gradually entered the agricultural field. Agricultural robots offer higher precision and efficiency in agricultural production. They can use cameras to collect images of crop growth and weeds, and then plan navigation routes based on different crop conditions to precisely control fertilizer and pesticide application rates. Obtaining the navigation path for agricultural robots is crucial for achieving agricultural automation, and the extraction of crop rows has a significant impact on the navigation accuracy of these robots.

[0003] Agricultural robots primarily operate in vast farmlands, and their control systems and navigation path extraction are complex, nonlinear, and unstable systems. Furthermore, my country has a vast territory with diverse crops, and current agricultural robots are mainly used for single crops; no universal agricultural robots have been designed, hindering their widespread adoption across various regions. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a vision-based agricultural robot control system and navigation path extraction method. The system preprocesses the crop rows of different crops and fits the navigation line, which has high accuracy, low time consumption, good adaptability and robustness.

[0005] To achieve the above objectives, the technical solution adopted by this invention is as follows: a vision-based agricultural robot control system, characterized in that it includes a vision processing module, which is connected to a camera via a USB interface. The camera is installed directly above the center of the agricultural robot, with the camera shooting direction facing the front of the robot body. The vision processing module processes the crop row information collected by the camera to obtain navigation parameters. The vision processing module is connected to a controller via an SPI communication interface, and transmits the navigation parameters to the controller. The controller then transmits the received navigation parameters to a motor drive module. The motor drive module is connected to a DC geared motor, and transmits control signals to the DC geared motor. The DC geared motor is divided into a left-side DC geared motor and a right-side DC geared motor.

[0006] Furthermore, encoders are installed on both the left and right DC geared motors. The encoders are connected to the controller via an SPI communication interface. The encoders are used to detect the operating status information of the DC geared motors and transmit the detected information to the controller. The controller uses a fuzzy PID algorithm to adjust the control signal and transmits the control signal processed by the fuzzy PID algorithm to the motor drive module in the form of a PWM pulse width signal.

[0007] Furthermore, the controller connects to ultrasonic sensors and infrared reflection sensors via the IIC bus. The ultrasonic sensors are used to detect obstacles in the farmland, and the infrared reflection sensors are used to detect information about crops. The ultrasonic sensors transmit obstacle information to the controller, which then controls the DC geared motor to stop operating via the motor drive module. The controller is also connected to a peristaltic pump module, which in turn connects to the peristaltic pump's DC geared motor. The infrared reflection sensors transmit information about detected crops to the controller, which then controls the peristaltic pump's DC geared motor to operate via the peristaltic pump module.

[0008] Furthermore, the vision processing module, controller, motor drive module, and peristaltic pump module are connected to the power supply module. The power supply module has two output power ports: 12V and 5V. The 5V output power port supplies power to the vision processing module and controller, while the 12V output power port supplies power to the motor drive module and peristaltic pump module.

[0009] A vision-based method for extracting navigation paths in an agricultural robot control system includes the following steps:

[0010] S1. Normalize the R, G, and B values ​​of the acquired image. The normalization formula is shown in (1):

[0011]

[0012] S2. Grayscale processing of crop row images; After normalization, the image is grayscaled using an improved super-green feature algorithm. The formula for the improved super-green feature algorithm is shown in (2):

[0013] G(x,y)=(1.88g(x,y)-r(x,y)-b(x,y)) (2)

[0014] S3. Perform image segmentation to convert the grayscale image into a binary image;

[0015] Threshold segmentation is performed using the Otsu method, with t set as the threshold for the segmented region. The area ratios θ1 and θ2 of the separated regions 1 and 2 in the entire image are θ1 and θ2, respectively. The relationship between the gray level of the entire image and the average gray level of regions 1 and 2 is expressed by equation (3):

[0016] u=u1θ1+u2θ2(3)

[0017] To achieve optimal image separation, region variance is an effective parameter for describing gray-level differences, and its expression can be represented by equation (4):

[0018]

[0019] This represents the variance between two regions after the image is segmented by a threshold. Therefore, the optimal separation state is considered when the variance between the two segmented regions is maximized. The threshold T is determined accordingly, and its expression can be represented by equation (5):

[0020]

[0021] S4. Image smoothing to eliminate image noise: First, a 5×5 pixel structure is used to perform median filtering on the binary image. Then, a 3×3 pixel structure is used to perform erosion processing on the image. Finally, a 3×3 pixel structure is used to perform dilation processing on the image, thereby removing small-area noise.

[0022] S5. Use the vertical projection method to obtain crop row information and detect crop row feature points;

[0023] S6. Use the least squares method to fit feature points and obtain the navigation line. The least squares method first sets the coordinates of the feature points as (x... i y i (i = 1, 2, 3...), the slope and intercept of the fitted line are a and b respectively, and the equation of the line is:

[0024] f(x) = ax + b (6)

[0025] Let the residual function be:

[0026]

[0027] Determine a set of a and b that minimizes the parameter function E. The derivative of E with respect to a and b is the final fitted line. According to formulas (8) and (9), the values ​​of a and b can be obtained, which are the slope and intercept of the fitted line.

[0028]

[0029] S7. The offset angle d is obtained by comparing the centerline with the center of the image window. Combined with the camera calibration parameters, the position in the world coordinate system is determined, and the navigation parameters are transmitted to the controller.

[0030] Furthermore, step S5, which uses the vertical projection method to obtain crop row feature points, specifically includes the following steps:

[0031] S501. Divide the image into 10 equal parts horizontally;

[0032] S502. In the binary image, sequentially select each horizontally divided image block for vertical projection; the vertical projection accumulates all column pixels in the binary image band, compares the accumulated pixels with a threshold, the area where the accumulated value is greater than the threshold is regarded as the crop area, and the area where the accumulated value is less than the threshold is regarded as the background. The formula is as shown in (10):

[0033]

[0034] Where B(i, j) is the pixel value of each pixel in the image band; the value of Δk is one-tenth of the image, that is, Δk = 0.1H a , H a is the image height;

[0035] S503. Calculate the cumulative mean M of the column pixels. The formula is as shown in (11):

[0036]

[0037] S504. Calculate the standard deviation E of the column pixels. The formula is as shown in (12):

[0038]

[0039] S505. Set the threshold When G(J)>T, it is the crop target area. Traverse the entire crop area. If G(J)>G(J - 1), then C[m][0] = J represents the left edge point of the crop row; if G(J)<G(J - 1), then C[m][1] = J represents the right edge point of the crop row; set the left and right edge thresholds U of the crop row i , used to determine whether the detected edge point is a true edge point; if U i >C[m][1] - C[m][0], then the crop row is retained, U i <C[m][1] - C[m][0], then it is discarded, and then calculate the midpoint of the left and right edges of the crop row; repeat the above steps, J = M a Stop the search; by scanning each image band, obtain the target points on the image band. In order to fit the center line of each target, classify the center points on the image band.

[0040] The beneficial effects of the present invention are as follows:

[0041] This navigation and control system can accurately identify crop rows for different crops based on image information captured by a camera positioned directly above the robot. The vision processing module calculates navigation parameters for these rows and transmits them to the controller. The controller uses a fuzzy PID algorithm to adjust the control signal and transmits the processed signal as a PWM pulse width signal to the motor drive module, thus enabling autonomous navigation for the agricultural robot. When an obstacle appears in front of the robot, ultrasonic sensors detect it in advance and stop the robot, ensuring its safety. Infrared reflection sensors installed at the rear of the robot identify crop seedlings, and the identification parameters are transmitted to the controller. The controller then controls the peristaltic pump module for precise spraying. Attached Figure Description

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a schematic block diagram of the overall structure of the present invention;

[0044] Figure 2 Workflow diagram for the vision processing module;

[0045] Figure 3 Ultrasonic sensor control flowchart;

[0046] Figure 4 Infrared reflection sensor control flowchart;

[0047] Figure 5 This is a schematic diagram of the fuzzy PID algorithm in this invention. Detailed Implementation

[0048] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0049] Numerous specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways than those described herein, and therefore the invention is not limited to the specific embodiments disclosed in the following specification.

[0050] Depend on Figure 1-4As shown in the figure, an embodiment of the present invention provides a vision-based agricultural robot control system and navigation path extraction method, including a vision processing module, a camera, a controller, a motor drive module, a motor, a power supply module, an encoder, an ultrasonic sensor, an infrared reflection sensor, and a peristaltic pump module.

[0051] The camera is mounted directly above the center of the agricultural robot, at a vertical height of approximately 1-1.2 meters above the ground. The angle between the camera and the horizontal direction is 45°-50°, and the camera is positioned in front of the robot. Preferably, an OV2710 camera module is used. The camera and vision processing module are connected via a USB interface. Preferably, the vision processing module is a Raspberry Pi 4B. The vision processing module processes the crop row information collected by the camera to obtain navigation parameters. The vision processing module is connected to the controller via an SPI communication interface, and transmits the navigation parameters to the controller. Preferably, the controller is an STM32F103C8T6. The controller transmits the received navigation parameters to the motor drive module, which is connected to a DC geared motor. The motor drive module transmits control signals to the DC geared motor, which consists of a left-side DC geared motor and a right-side DC geared motor.

[0052] Encoders are installed on both the left and right DC geared motors. The encoders are connected to the controller via an SPI communication interface. The encoders transmit the status information of the DC geared motors to the controller. The controller uses a fuzzy PID algorithm to adjust the control signal. That is, the controller uses a fuzzy PID algorithm to adjust the magnitude of the output control signal according to the operating parameters of each DC geared motor. The control signal processed by the fuzzy PID algorithm is transmitted to the motor drive module in the form of a PWM pulse width signal to adjust the current operating parameters of each DC geared motor.

[0053] Fuzzy PID control can quickly adjust the wheel speed error of agricultural robots. The fuzzy PID algorithm combines the high control accuracy of PID with the flexibility and adaptability of fuzzy control. The structure of a fuzzy PID control system is as follows: Figure 5 As shown, the system input is the given wheel speed driven by the motor, and the feedback value is the digital value fed back by the encoder, Δk. p Δk i Δk d This is a correction factor. Δk p Δk i Δk d From formula (1):

[0054]

[0055] The controller outputs a control voltage of 0-5V to the motor driver, and the PWM wave parameter adjustment range is 0-255. This means that the voltage at the microcontroller's output pin can be changed by adjusting the PWM wave parameters. The relationship is as follows:

[0056]

[0057] Where: m is the input parameter of the PWM wave, and u is the pin output voltage of the STM32 lower-level machine.

[0058] The encoder on the left DC geared motor feeds back the motor's speed to the controller in real time, enabling closed-loop control of the motor's speed by the driver. Similarly, the encoder on the right DC geared motor feeds back its speed to the controller in real time, achieving closed-loop control of the motor's speed by the driver. When the control inputs for the left and right DC geared motors are the same, the left and right wheels rotate at the same speed, allowing the robot to travel in a straight line. Increasing the control input for the left DC geared motor and decreasing the control input for the right DC geared motor allows for different speeds for the left and right wheels, achieving electronic differential control and enabling autonomous steering of the agricultural robot.

[0059] The controller connects to the ultrasonic and infrared reflection sensors via the IIC bus to receive signals from them. When the agricultural robot encounters an obstacle, the ultrasonic sensor transmits the obstacle information to the controller. The controller prioritizes processing the obstacle detection signal from the ultrasonic sensor and then controls the DC geared motor to stop operating via the motor drive module, thereby stopping the agricultural robot. The controller also connects to the peristaltic pump module, which in turn connects to the peristaltic pump's DC geared motor. When the infrared reflection sensor detects crops, it transmits the information to the controller. Upon receiving the crop detection signal from the infrared reflection sensor, the controller controls the peristaltic pump's DC geared motor to operate, enabling the agricultural robot to spray pesticides.

[0060] The vision processing module, controller, motor drive module, and peristaltic pump module are each connected to the power supply module, which has both 12V and 5V output ports. The 5V output port powers the vision processing module and controller, while the 12V output port powers the motor drive module and peristaltic pump module.

[0061] The visual processing module operates as follows:

[0062] S1. Normalize the R, G, and B values ​​of the acquired image. To reduce the influence of irrelevant variables such as sunlight, the RGB color model is used to normalize the R, G, and B values ​​of the currently acquired image. The normalization formula is shown in (3):

[0063]

[0064] S2. Grayscale processing of crop row images. The normalized image is then grayscaled using an improved super-green feature algorithm. The formula for the improved super-green feature algorithm is shown in (4):

[0065] G(x,y)=(1.88g(x,y)-r(x,y)-b(x,y)) (4)

[0066] S3. Perform image segmentation to convert the grayscale image into a binary image.

[0067] Threshold segmentation is performed using the Otsu method, with t set as the threshold for the segmented region. The area ratios θ1 and θ2 of the separated regions 1 and 2 in the entire image are θ1 and θ2, respectively. The relationship between the gray level of the entire image and the average gray level of regions 1 and 2 is expressed by equation (5):

[0068] u=u1θ1+u2θ2(5)

[0069] To achieve optimal image separation, region variance is an effective parameter for describing gray-level differences, and its expression can be represented by equation (6):

[0070]

[0071] This represents the variance between the two regions after the image has been segmented by a threshold. Therefore, the optimal separation state is considered when the variance between the two segmented regions is maximized, and the threshold T is determined accordingly. The expression for T can be represented by equation (7):

[0072]

[0073] S4. Image smoothing to eliminate image noise. First, a 5×5 pixel structure is used to perform median filtering on the binary image. Then, a 3×3 pixel structure is used to perform erosion processing on the image. Finally, a 3×3 pixel structure is used to perform dilation processing on the image, thereby removing small-area noise.

[0074] S5. Obtain crop row information using the vertical projection method and detect crop row feature points. The detection steps are as follows:

[0075] S501. In the image acquisition work, it is found that the camera has a certain angle relative to the ground. Due to the influence of the downward shooting, the crop rows are not completely parallel in the image, and the crop rows finally converge at a central point. The vertical projection method cannot be directly applied to image acquisition. Therefore, it is considered to equally divide the image into 10 blocks in the horizontal direction.

[0076] S502. In the binary image, each block of the horizontally divided image is sequentially selected for vertical projection. The vertical projection accumulates all the column pixels in the binary image band, compares the accumulated pixels with the threshold, and the area where the accumulated value is greater than the threshold is regarded as the crop area, and the area where the accumulated value is less than the threshold is regarded as the background. The formula is as shown in (8):

[0077]

[0078] Where B(i,j) is the pixel value of each pixel in the image band; the value of Δk is one-tenth of the image, that is, Δk = 0.1H a , H a is the image height.

[0079] S503. Calculate the accumulated mean M of the column pixels. The formula is as shown in (9):

[0080]

[0081] S504. Calculate the standard deviation E of the column pixels. The formula is as shown in (10):

[0082]

[0083] S505. Set the threshold When G(J)>T, it is the crop target area. Traverse the entire crop area. If G(J)>G(J - 1), then C[m][0] = J represents the left edge point of the crop row; if G(J)<G(J - 1), then C[m][1] = J represents the right edge point of the crop row. Set the left and right edge thresholds U i , used to determine whether the detected edge point is a true edge point. If U i >C[m][1] - C[m][0], the crop row is retained, and if U i <C[m][1] - C[m][0], it is discarded. Then calculate the midpoint of the left and right edges of the crop row. Repeat the above steps until J = M a Stop the search. By scanning each image band, the target points on the image band are obtained. In order to fit the center line of each target, the center points on the image band are classified.

[0084] S6. Use the least squares method to fit the feature points to obtain the navigation line. The least squares method first sets the feature point coordinates as (x iy i (i = 1, 2, 3...), the slope and intercept of the fitted line are a and b respectively, and the equation of the line is:

[0085] f(x) = ax + b (11)

[0086] Let the residual function be:

[0087]

[0088] Determine a set of a and b that minimizes the parameter function E. The derivative of E with respect to a and b is the final fitted line. According to formulas (13) and (14), the values ​​of a and b can be obtained, which are the slope and intercept of the fitted line.

[0089]

[0090] S7. The offset angle d is obtained by comparing the center line with the center of the image window. Combined with the camera calibration parameters, the position in the world coordinate system is determined, and the navigation parameters are transmitted to the controller.

[0091] In summary, this invention provides a navigation control system and method for agricultural robots. A camera mounted directly above the agricultural robot collects information about crop rows. After visual processing, the collected crop row images undergo row normalization. The resulting grayscale image is then converted to grayscale using a super-green feature algorithm. Next, threshold segmentation and morphological processing are performed using OTSU to remove noise. Finally, feature points are extracted using the vertical projection method, and the crop row centerline is fitted using the least squares method. The centerline is compared with the center of the image window to obtain the offset angle 'd'. Combined with the camera's calibration parameters, the world coordinate system position is determined. The visual processing module transmits navigation parameters to the controller, which uses a fuzzy PID algorithm to adjust the control signal. The control signal processed by the fuzzy PID algorithm is transmitted to the motor drive module as a PWM pulse width signal, thereby achieving autonomous navigation for the agricultural robot. When the agricultural robot encounters an obstacle, the ultrasonic sensor transmits the obstacle detection information to the controller, causing the agricultural robot to stop. When the infrared reflection sensor detects crops, it transmits the information to the controller, which then controls the peristaltic pump module to spray pesticides.

[0092] All electrical components in this case are connected by wires. The specific connection and control sequence should refer to the working principle described above. The detailed connection methods are well-known technologies in this field. The above mainly introduces the working principle and process, and will not provide a detailed description of the electrical control.

[0093] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications 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 protection scope of the present invention.

Claims

1. A vision-based agricultural robot control system, characterized in that, The system includes a vision processing module connected to a camera via a USB interface. The camera is mounted directly above the center of the agricultural robot, with its view angled towards the front of the robot. The vision processing module processes the crop row information captured by the camera to obtain navigation parameters. The vision processing module is also connected to a controller via an SPI communication interface, transmitting the navigation parameters to the controller. The controller then sends the received navigation parameters to a motor drive module connected to a DC geared motor. The motor drive module transmits control signals to the DC geared motor, which consists of a left-side DC geared motor and a right-side DC geared motor. The vision processing module performs the following operations: S1. Normalize the R, G, and B values ​​of the acquired image. The normalization formula is shown in (1): S2. Grayscale processing of crop row images; After normalization, the image is grayscaled using an improved super-green feature algorithm. The formula for the improved super-green feature algorithm is shown in (2): G(x,y)=(1.88g(x,y)-r(x,y)-b(x,y)) (2) S3. Perform image segmentation to convert the grayscale image into a binary image; Threshold segmentation is performed using the Otsu method, with t set as the threshold for the segmented region. The area ratios θ1 and θ2 of the separated regions 1 and 2 in the entire image are θ1 and θ2, respectively. The relationship between the gray level of the entire image and the average gray level of regions 1 and 2 is expressed by equation (3): u=u1θ1+u2θ2 (3) To achieve optimal image separation, region variance is an effective parameter for describing gray-level differences, and its expression can be represented by equation (4): This represents the variance between two regions after the image is segmented by a threshold. Therefore, the optimal separation state is considered when the variance between the two segmented regions is maximized. The threshold T is determined accordingly, and its expression can be represented by equation (5): S4. Image smoothing to eliminate image noise: First, a 5×5 pixel structure is used to perform median filtering on the binary image. Then, a 3×3 pixel structure is used to perform erosion processing on the image. Finally, a 3×3 pixel structure is used to perform dilation processing on the image, thereby removing small-area noise. S5. Use the vertical projection method to obtain crop row information and detect crop row feature points; S6. Use the least squares method to fit feature points and obtain the navigation line. The least squares method first sets the coordinates of the feature points as (x... i y i (i = 1, 2, 3...), the slope and intercept of the fitted line are a and b respectively, and the equation of the line is: f(x) = ax + b (6) Let the residual function be: Determine a set of a and b that minimizes the parameter function E. The derivative of E with respect to a and b is the final fitted line. According to formulas (8) and (9), the values ​​of a and b can be obtained, which are the slope and intercept of the fitted line. S7. The offset angle d is obtained by comparing the centerline with the center of the image window. Combined with the camera calibration parameters, the position in the world coordinate system is determined, and the navigation parameters are transmitted to the controller.

2. The vision-based agricultural robot control system according to claim 1, characterized in that, The left and right DC geared motors are each equipped with an encoder. The encoder is connected to the controller via an SPI communication interface. The encoder is used to detect the operating status information of the DC geared motor. The encoder transmits the detected information to the controller. The controller uses a fuzzy PID algorithm to adjust the control signal. The controller transmits the control signal processed by the fuzzy PID algorithm to the motor drive module in the form of a PWM pulse width signal.

3. The vision-based agricultural robot control system according to claim 1, characterized in that, The controller is connected to an ultrasonic sensor and an infrared reflection sensor via an IIC bus. The ultrasonic sensor is used to detect obstacles in the farmland, and the infrared reflection sensor is used to detect information about crops. The ultrasonic sensor transmits obstacle information to the controller, and the controller controls the DC geared motor to stop operating through the motor drive module. The controller is also connected to a peristaltic pump module, which is connected to a peristaltic pump DC geared motor. The infrared reflection sensor transmits the detected crop information to the controller, and the controller controls the peristaltic pump DC geared motor to operate through the peristaltic pump module.

4. The vision-based agricultural robot control system according to claim 1, characterized in that, The vision processing module, controller, motor drive module, and peristaltic pump module are respectively connected to the power supply module. The power supply module has two output power ports: 12V and 5V. The 5V output power port supplies power to the vision processing module and controller, and the 12V output power port supplies power to the motor drive module and peristaltic pump module.

5. A vision-based agricultural robot control system according to claim 1, characterized in that, S5 specifically includes the following steps: S501. Divide the image into 10 blocks horizontally; S502. In the binary image, each horizontally divided image block is selected sequentially for vertical projection; the vertical projection accumulates all column pixels in the binary image band, and compares the accumulated pixels with a threshold. The area with the accumulated value greater than the threshold is regarded as the crop area, and the area with the accumulated value less than the threshold is regarded as the background. The formula is as shown in (10): Where B(i,j) represents the pixel value in the image band; the value of Δk is one-tenth of the image, i.e., Δk = 0.1H, H a Image height; S503. Calculate the cumulative mean value M of the column pixels, as shown in formula (11): S504. Calculate the standard deviation E of the column pixels, as shown in formula (12): S505. Set Threshold When G(J) > T, it is the crop target area. Traverse the entire crop area. If G(J) > G(J - 1), then C[m][0] = J represents the left edge point of the crop row; if G(J) < G(J - 1), then C[m][1] = J represents the right edge point of the crop row; set the left and right edge thresholds U of the crop row i , used to determine whether the detected edge point is a true edge point; if U i > C[m][1] - C[m][0], then the crop row is retained, U i < C[m][1] - C[m][0], then it is discarded, and then calculate the midpoint of the left and right edges of the crop row; repeat the above steps, J = M a Stop the search; by scanning each image band, obtain the target points on the image band. In order to fit the center line of each target, classify the center points on the image band.

Images