Journal of information and
communication convergence engineering
(J. Inf. Commun. Converg. Eng.)

Now and in the future, there are many compact optoelectronics devices with affordable and multifunctional [1]. One of the applications is the target marker and the gun [2,3]. Generally, cameras and Computer Vision (CV) can provide information that depicts an object based on color, pattern, and size. One of the applications of CV in vision marker recognition can be developed into a unique marker recognition system. The system detects unique markers to meet the needs of target shooting automation. The marker is a laser beam with a unique color, shape, size, and blinking [4] so that a programmed computer only recognizes it. Object recognition uses color, dimensions, and shape contour detection methods [5-7]. Meanwhile, the object marked with markers will be detected and tracked as a shot marker. When the object moves, the camera will follow the object movement assisted by a dual-axis servo motor [8].

Furthermore, the camera will become a sensor to determine the distance between the object and the launcher [9,10]. In general, several basic methods exist for estimating the distance between an object and the monocular camera, such as comparing virtual images in the mirrors [4]. When the camera detects the presence of an object, it will determine the size of the object based on the area detected by the computer. Then the data is compared with the object's actual size to find the accurate distance between the camera and the object [11-13]. Moreover, the estimated distance data determines the force to launch a bullet to hit the target precisely [14].

This research aims to build and develop a unique object recognition system using a blinking laser so that the launcher can shoot on objects that have been marked system with precision and accuracy [15].

A. Object Recognition

A marker is an artificial sign that identifies objects easily [16]. Then, the camera with image-processing software functioned to recognize and identify objects [6,16,17]. The camera has a resolution of 640×480 pixels, a frame rate of 30 fps, and a capturing angle of 54°. The image captured by the camera is processed by a computer using the OpenCV library, which contains an image-processing program for recognizing and identifying objects [18-20]. In this research, the features developed by OpenCV for recognizing and identifying objects are the Unique Marker Recognizing [17].

The object is carried out hierarchically based on predefined colors, patterns, and sizes. Then, the coordinates of the target midpoint are detected in the image, and it will be used as feedback on the servo motor [19-21]. Moreover, when the object moves, the camera and its propulsion system will follow the object’s movement, as shown in Fig. 1.

Figure 1. Integration of object detection, unique marker, and bullet launcher mechanism.

The camera with a biconvex lens influences the relationship between the object and its shadow, satisfying the triangle similarity principle. Based on this condition, an estimated calculation regarding the focal length ratio divided by the shadow's width equals the real object distance to the camera divided by the lens width (images) [6]. Here, by increasing the distance between the real object and the camera, the size, and width of the object captured by the camera gets smaller. Moreover, when the size of the object and the focal length of the camera lens are known, the real distance of the object to the camera can be calculated. The distance based on these calculations becomes the input launcher to shoot the target with precision and accuracy.

B. Encoding Laser

Encoding laser is built by laser with the variant of pulsed wave modulation (PWM). The PWM signal consists of a special pattern controlled by time toward a target and measuring the time it takes for pulses to be reflected from the target to the receiver in the form of a camera. Then, such target marking was implemented with a laser lighting quantity [20], the control system against the amount of electricity (dimmer) by using the Arduino Uno-based pulse width modulation (PWM) method [19,22]. In this case, the dimmer circuit uses the principle of voltage control to produce a PWM signal, as shown in Fig. 2.

Figure 2. (a) The PWM of laser blinking consists of the t_on/t_bright, t_off, and t_dim. (b) Observation in real time by the camera, and masking side in computer.

A unique marker is built with a laser coding system, a combination of t_on and t_off, with unique variants, as shown in Fig. 2(a). The marker is a laser beam with a unique size, color, and pattern that the user can only recognize. One of the essential purposes of this research is the application of unique markers, which will be developed using a cue code in the form of a blinking laser.

C. Trajectory System

The launcher controls based on a pneumatic system capable of shooting a unique automatic marker proposed is a close loop control [23]. Furthermore, several calculation steps are needed to adjust the air pressure in the system. When analyzing the air pressure on the launcher, data such as object distance, bullet size, bullet mass, and target size become important [21]. The bullet trajectory analysis follows the parabolic form, and this concept obeys the straight-line motion with constant acceleration. When the camera detects a target, the computer will perform calculations to determine the required pressure. Once the amount of pressure is known, the force needed to eject a bullet can be determined. When the force has been determined, the valve will be controlled to open, and a bullet ejects toward the target.

The block diagram of the bullet launcher with a unique automatic marker is shown in Fig. 3. Here, the block diagram of the bullet launcher consists of several blocks, such as process, actuator, and feedback. The primary variable designed for this system is the air pressure inside the compressor.

Figure 3. Block diagram of bullet launcher with a unique automatic marker.

The feedback system is designed using a pressure sensor as a response. The microcontroller will process the amount of pressure detected as feedback, and this value will be compared with the set point and affect the process output. Here, an actuator used in an automatic unique marker launcher is a solenoid valve consisting of a drain valve and a firing valve.

The process of the pneumatic system on this bullet launcher is that the compressor provides the air pressure when the solenoid valve is open. The regulator measured the pressure. Then the storage filling process is completed according to the set point. Here, the set point value is the result of converting the object’s distance to the launcher. Another process is the solenoid valve connected to the drain will open and connect to the pressure sensor. When the measured pressure follows the set point, the output solenoid valve connected to the plant will open. This entire process is finalized at the user's command to launch the bullets. If there is still air in the storage tank after processing, the solenoid valve connected to the drain will open to remove the remaining pressure.

D. Dual Axis of Bullet Launcher

The turret and gun system are integrated systems for the bullet launcher system. The turret and gun are 2-axis; a turret is a horizontal movement while the gun is the up and down movement. Here, the servo motor movement is controlled by an Arduino microcontroller at the coordinates. The coordinates obtained by the computer are sent serially to Arduino, and it is used as input to drive the servo motor. Moreover, the system will signal to stop the motor when the coordinates match the input data. The output generated from this control is the angle of the motor movement [8].

The rotation of the servo motor uses the PWM (pulse width modulation) method with a closed-loop system. When the servo motor rotates, the camera reads each rotation angle so that the difference in coordinates between the camera and the object will be known. Here, the Arduino microcontroller processes the information and adjusts the angle according to the reference [24-26]. The reference value is the output from the potentiometer with a length of 10 bits (0-1023). The output generated from the servo motor control system is the angle corresponding to the position input in the software [27].

A. Color Detections

Color detection of indoor conditions is carried out with several color variations like red, green, and blue. Here, the purpose of color detection is to discover that Computer Vision can recognize object colors in various conditions. Color detection is carried out by taking the Hue Saturation Value (HSV) parameter. Moreover, the filtering mechanism based on object color is shown in Fig. 4.

Figure 4. Color filter mechanism as a detector in indoor conditions.

Indoor conditions with stable lighting give the stable threshold parameter as well on the CV when detecting the color of an object. Meanwhile, in outdoor conditions, the threshold parameters must be adjusted. Due to outdoor lighting conditions depending heavily on sunlight, this light is affected by weather, clouds, and measurement time.

B. Pattern and Size Detection

Pattern detection is carried out with several variations of shapes, such as circles, squares, and triangles. One of the results of object pattern detection at a certain color is shown in Fig. 5.

Figure 5. Object with blue color in pattern detection

Meanwhile, object size detection is performed to determine the real distance between the camera and the object. This detection is carried out with several variations in sizes and shapes, such as circles, triangles, and squares, with dimensions of 5, 10, and 15 cm, respectively. Table 1 shows the results of the detection of pattern and size.

Table 1 . Detection result of pattern, size, and approximate distance.

Real Object Side Size (cm)	Detection Distance		Pattern	Detected Object Size
	Min (cm)	Max (cm)		Min (pixel)	Max (pixel)
5	120	240	Circle	26x23	11x10
10	120	400		52x49	12x11
15	120	680		78x75	12x11
5	120	240	Rectangle	26x25	12x11
10	120	400		53x53	14x13
15	120	640		80x83	13x13
5	120	120	Triangle	22.4x18.6	22.4x18.6
10	120	200		45x45	25x25
15	120	360		75x73.1	22x23

The measurement results show that the camera can detect distances well in several object patterns, such as a circle with a diameter of 15 cm detected at a distance of 680 cm. Meanwhile, for other objects, it is measured by the edge of the pattern.

Meanwhile, the camera’s ability to distinguish the size of objects in the same shape and color is illustrated in Fig. 6.

Figure 6. Determination of the size of the object at the same shape and color

C. Detection of Approximate Distance

The approximate distance measurement by the camera is carried out on an object with a diameter of 15 cm. here, the real distance is 20 to 240 cm, with measures every 20 cm.

The results of measuring the circle (see Fig.7) show that the standard deviation value for the slope is 0.002715 and 0.002548 for up and down measurements, respectively.

Figure 7. Distance measurement by the camera on a circle

Meanwhile, Fig. 8 shows that the results of the distance approximation of the square have the highest value for 100% accuracy and precision obtained at the distance of 240 cm.

Figure 8. Distance measurement by the camera on a square

Then, the results of measuring the square show that the standard deviation value for the slope is 0.002536 and 0.000324 for up and down measurements, respectively.

Then the calculation of the average value of the camera measurement error as a distance sensor is ±5% for square. Moreover, measuring the triangle results show that the slope’s standard deviation value is 0.003330 and 0.004332 for up and down measurements, respectively. And Fig. 9 shows the calculation of the average value of the camera measurement error as a distance sensor is ±6%.

Figure 9. Distance measurement by the camera on a triangle.

D. Tracking Objects Measurement

Object tracking is carried out with various distances from 50 to 300 cm. Here, synchronization between object recognition by the camera and the drive system on the servo motor by the microcontroller, as shown in Fig. 10.

Figure 10. Tracking objects mechanism (a) instrument position and objects, and (b) target on the computer view.

This measurement was carried out with seven times repetitions to see the consistency of the results. Then, the ability of the camera to track a moving object is shown in the accuracy.

E. Characteristics of Encoding Laser

Laser blinking is made by controlling the PWM signal.

Fig. 11 shows the results of measuring variations in the duty cycle on blue, red, green, and blue diode lasers. Here, the combination of ton and toff represents the duty cycle.

Figure 11. The effect of increasing the duty cycle on the output voltage

Fig. 11 shows that the increasing the percentage of duty cycle that is set, the voltage that is identified also increases.

Furthermore, the results of outdoor measurements are presented in Table 2. Here, the intensity measured by the lux meter is 91.6 lux in the middle of the day and 0 lux at night. Measurements in outdoor conditions were carried out at 50 cm and 300 cm with variations in duty cycle starting from 20 to 60%. When the intensity of sunlight gets brighter, it makes it difficult for the camera to recognize the laser’s color, so the laser’s intensity also increases, and the camera can detect it properly. In addition, the greater the intensity of the laser light, the farther the range in recognizing objects, and vice versa. So, it can be concluded that the HSV value set is the same for the intensity and distance of different objects. This indicates that the trackbar configuration on the system has been properly calibrated.

Table 2 . Condition of laser measurement as a target marker, this measurement is carried out outdoors.

Measurements were made to determine the effect of ambient light on the camera in recognizing the laser as a target marker. The effect of ambient light is differentiated in indoor and outdoor measurements. Figs. 12 and 13 show the results of the effect of the duty cycle on the light intensity of the blue laser.

Figure 12. The effect of increasing duty cycle on intensity. These measurements were taken condition indoors.

Figure 13. The effect of increasing duty cycle on intensity. These measurements were taken condition outdoors.

From Figs. 12 and 13 regarding the measurement of laser intensity in indoor and outdoor conditions, there are significant differences in results between indoor and outdoor conditions.

Moreover, the ambient light intensity affects the detection system. Moreover, the darker ambient light conditions will make the camera work better.

F. Measurement of Trajectory

Measurement of accuracy and precision of the bullet launcher is divided into three categories, such as response time, the accuracy of the bullet launcher against the target, and measurement of shot precision. The delay time measurement is needed to determine how fast the compressor response can fill the air storage, the servo motor response to executing the commands, and the camera response to the movement of the target marker. In this study, measurement of the delay time was carried out using variations in air pressure from 100 to 500 kPa with a stepping of 50 kPa. Table 3 shows the measurement data for air-filling delay time in storage.

Table 3 . Measurement of the time to produce air pressure and store in the storage.

Pressure (kPa)	Compressor Charging Respond Time (s)
100	0,6
150	0,8
200	1,01
250	2,05
300	2,56
350	2,83
400	3,86
450	4,39
500	5,24

Table 3 shows that the delay time increases linearly with increasing air pressure requirements. To produce a pressure of 100 kPa, the compressor takes 0.6 seconds to fill the air storage, while for a pressure of 500 kPa, the compressor takes 5.24 seconds to fill the air storage.

Furthermore, launcher accuracy measurements were carried out indoors and outdoors with several variations of distance. This measurement is carried out after the microcontroller synchronizes the object recognition system, servo motor movement system, and launcher system. The purpose of this measurement is to find out if the synchronization process is working well. The camera can recognize and send commands to the servo motor so that the prediction of the launcher can be at the target’s midpoint. Measurements were made at 50 cm to 300 cm with a measurement range of 50 cm. The measurement results are shown in Tables 4 and 5.

Table 4 . The measurement of the accuracy of the bullet launcher with several repetitions. This measurement was taken in indoor conditions.

No	Object Distance from System	Repeat Number (1=success, 0=fail)							The accuracy of the automatic weapon aiming system (%)
		1	2	3	4	5	6	7
1	50 cm	1	0	1	1	1	1	1	85.71%
2	100 cm	1	1	1	1	1	1	1	100.00%
3	150 cm	1	1	1	1	1	1	1	100.00%
4	200 cm	1	1	1	1	0	1	1	85.71%
5	250 cm	1	1	1	1	1	1	1	100.00%
6	300 cm	1	1	1	1	1	1	1	100.00%
Average Accuracy Test									95.24%

Table 5 . The measurement of the accuracy of the bullet launcher with several repetitions. This measurement was taken in an outdoor condition.

No	Object Distance from System	Repeat Number (1=success, 0=fail)							The accuracy of the automatic weapon aiming system (%)
		1	2	3	4	5	6	7
1	50 cm	1	1	1	1	1	1	1	100,00%
2	100 cm	1	1	1	1	1	1	1	100.00%
3	150 cm	1	1	1	1	0	1	1	85,71%
4	200 cm	1	1	1	1	0	1	1	85,71%
5	250 cm	1	1	1	1	0	0	1	71,42%
6	300 cm	1	1	1	0	1	0	1	71,42%
Average Accuracy Test									85,71%

Based on measurements with seven repetitions for distance variations, as shown in Tables 3 and 4, the accuracy of a launcher throwing bullets at the object area illuminated by a unique marker with the help of a drive servo motor in directing an automatic system has an average accuracy of 95.24 and 85.71% for indoor and outdoor conditions respectively. Then the precision value that hits the target can be calculated. The accuracy and precision of the shot were measured on a circle with a diameter of 15 cm at an angle of 90.85 of the bullet launchers. This measurement, to obtain the results of the accuracy and precision of the shot, is carried out with seven repetitions. Then the average data is taken for each target position. The measurement results are shown in Table 6.

Table 6 . Calculation results of approximated error, accuracy, and precision of the shot.

Target (90,85)		Object Distance
Distance of the Bullet Against the Center of the Target (cm)	Test Number	50 cm	100 cm	150 cm	200 cm
	1	0.5	1.5	2.8	5
	2	0.8	1.5	2.8	5.2
	3	1	1.6	3	5.3
	4	1.4	2	3.2	5.4
	5	1.5	2.2	3.4	5.4
	6	1.5	2.2	4	5.8
	7	2	2.6	4.2	6.6
RMSE (cm)	0.84	2.06	3.38	5.55
Accuracy	98.32%	97.94%	97.75%	97.23%
Precision	74.06%	81.93%	86.57%	93.06%

Using the Root Mean Square Error (RMSE) method, an estimate of the average measurement error, accuracy, and precision is obtained. Table 6 shows that the estimated error shooting data at 50 cm is 0.84 cm. At 100 cm, it is 2.06 cm. At 150 cm, it is 3.38 cm, and at 200 cm, it is 5.55 cm. With a simple calculation, it can be obtained that the accuracy of the shot at 50 cm is 98.32%, at 100 cm is 97.94%, at 150 cm is 97.75%, and at 200 cm is 97.23%. While the precision value obtained from the calculation results at 50 cm is 74.06%, at 100 cm is 81.93%, at 150 cm is 86.57%, and at 200 cm is 93.06%.

From the measurement of the accuracy and precision of shots with variations in object distance, the highest accuracy rate was obtained for shots at 98.32% at 50 cm, while the highest precision for shots was obtained for shots with 93.06% at 200 cm. So, from the calculation results, it can be concluded that the closer the target distance is, the higher the accuracy will be. Still, the lower the precision of the shot, it's because the bullet’s point spreads around the target’s midpoint and vice versa.

G. Target Position and Response Time tracking measurement

This measurement is carried out to see the response of tracking time to the target and the position of the system moving from right to left, from top to bottom, and vice versa. This movement is done when aiming the gun. Here, the bullet launcher may respond when the target is within range of the camera. The result of this measurement is shown in Table 7.

Table 7 . Measurement of target position and response time tracking to the target.

Position	Servo Degree		Response Time (s)	Accuracy of the system aiming the weapon when the laser on
	Horizontal	Vertical
Right (Axis x)	30	50	1.1	Accurate
Left (Axis x)	15	50	0.65	Accurate
Up (Axis y)	90	60	0.87	Accurate
Down (Axis y)	90	45	0.43	Accurate
Right (Axis x) + Up (Axis y)	115	60	1.57	Accurate
Right (Axis x) + Down (Axis y)	115	45	1.36	Accurate
Left (Axis x) + Up (Axis y)	75	60	1.56	Accurate
Right (Axis x) + Down (Axis y)	75	45	1.34	Accurate
Average Response Time Tracking Target			1.11	Accurate

In this system, the measurement of a bullet launcher system that can shoot accurately based on the target position has an average response time of 1.11 s. The bullet launcher system also directs the gun as soon as the target marker lights up at the set position.

The automatic unique marker shooter system is designed to be able to track, detect, lock, mark, and shoot the objects that a unique marker has marked. This system consists of object recognition, a drive, and a bullet launcher. Here, the automatic unique marker shooter is designed with a closedloop control system. Furthermore, pneumatics is used as the source of bullet power to be launched. Then, the feedback signal on the control system is a pressure sensor. The pressure sensor used in the system has a measurement error value of ±3%. Meanwhile, this system has been able to shoot precisely with an accuracy of 98.32% for a 50 cm target distance and 93.06% for 200 cm.