Design of intelligent camp anti-shock system based on FPGA

Abstract: In order to improve security measures, delay the actions of criminals, and ensure the safety of the camp, a solution for an intelligent anti-impact system for the camp is proposed. This solution uses the shape of moving objects, license plate information, and vehicle speed as input features, and adopts virtual coil sensing, license plate recognition, vehicle speed measurement, system control and other methods to achieve early warning and automatic pop-up of vehicle-blocking nails to prevent the passage of illegal vehicles. The system structure of the solution and the working principle of each module are emphasized, and a hardware platform based on the Virtex 5 series XC5VSX50T chip is designed and implemented.
Keywords: Virtex 5; virtual coil; neural network; license plate recognition; vehicle speed measurement; anti-impact system

0 Introduction
In recent years, the international and domestic security situation is not optimistic, and terrorist activities are on the rise. The number of incidents that impact the government, the military, and enterprises due to various problems has increased year by year. Among them, there are cases where terrorists use cars as tools to carry explosive devices and rush into target units to detonate, causing casualties. How to quickly and effectively intercept suspected criminal vehicles has raised new issues for the public security and armed police, and improving intelligent interception has become a top priority. Through research, it was found that the security measures for the camp gates of most camps in China are relatively simple, and the defense vehicle impact equipment is basically speed bumps, car stoppers, etc., and the car stoppers are all wired or wireless remote control triggers. Due to the sudden terrorist activities, cars quickly broke into the camp, and the on-duty personnel had no time to react, and the car stoppers could not pop out at the right time. Therefore, in order to ensure the safety of target units such as the government, the military, and enterprises, a camp anti-impact system is designed. Through the simulation of coil induction, license plate recognition, vehicle speed measurement, system control, early warning and other technologies, the car stoppers are automatically popped out, which can effectively intercept terrorists driving into the camp.

1 System implementation block diagram and working principles of each part
1.1 Intelligent camp anti-impact management system solution
Government, military, enterprises and other units generally have special lanes for vehicles to enter and exit, and they all have security personnel, as shown in Figure 1.

When the vehicle reaches 10 m from the camp entrance, i.e. the virtual coil sensing area in Figure 1, the camera automatically detects the vehicle's license plate information. The basic process is: whether the vehicle is a unit vehicle → whether the vehicle blocking nail pops out. After judgment, the vehicles entering and leaving the camp are handled according to the following three situations:
(1) Unit vehicles
When the vehicle reaches the virtual coil sensing area, the system automatically identifies the vehicle's license plate and determines whether it is a unit vehicle. If it is a unit vehicle, the system is reset, the camera is reset, the speed measuring device does not work, and the system is idle.
(2) Temporary vehicles
When the vehicle reaches the virtual coil sensing area, the system determines that it is not a unit vehicle, automatically captures it, and the camp duty officer stops the vehicle. After registering the information, the driver is issued a slow-speed entry command. If the vehicle does not accelerate upon command, the system has two speed measurement lines. When the vehicle passes the first speed measurement line, if it exceeds the speed limit, the system will quickly pop up a warning sign to slow down immediately and sound an alarm to remind the driver to slow down. When the vehicle passes the second speed measurement line, if it does not slow down but still exceeds the speed limit, the vehicle stopper will pop up to indicate that it is a dangerous vehicle.
(3) Special vehicles
Special vehicles include military vehicles, police cars, ambulances, etc. The gate duty personnel can make corresponding handling according to the actual situation.
1.2 Basic block diagram of the intelligent camp anti-impact system
The camp anti-impact system discussed in this paper is theoretically mainly designed based on FPGA and DSP systems. FPGA is used to realize the acquisition and transmission of image and speed signals, the overall control of the system and other operations. The fast computing power of DSP is used to realize license plate recognition. The basic block diagram of the system is shown in Figure 2.

The camera continuously transmits the surveillance video stream to the video decoding chip A/D. When a vehicle passes the trigger line, the vehicle detector is triggered, and the detector gives the FPGA a detection pulse, triggering the FPGA to capture a frame of image from the video stream. The FPGA caches a row of image data, and uses an interrupt to notify the DSP to obtain the image when the row is full. The DSP uses EDMA to obtain image data from the FPGA. After receiving a frame of image data, it starts the algorithm to process the image in units of frames. The obtained results are written into the FPGA and compared with the stored data. The speed measurement signal is sent from the speed measurement sensor to the FPGA, and is directly processed inside the FPGA. After that, according to the judgment result, a control signal is issued to determine whether the alarm alarms or whether the vehicle blocking nails pop up.
1.3 Hardware platform design of intelligent camp anti-impact system
At present, most of the design schemes used for pattern recognition, system speed measurement, etc. use a master-slave structure of two or more microprocessors, one for high-speed data acquisition tasks, and one for complex signal processing, such as FPGA+DSP, ARM+DSP, etc. The key technologies involved in this paper include analog coil induction design, license plate automatic recognition design, and automobile infrared light speed measurement design. Considering the small traffic volume in Yingmen District, the data processing flow is not large, but the system has a high degree of modularity. In combination with the characteristics of FPGA+DSP structure,
the XC5VSX50T chip of Xilinx's Virtex 5 series is used. As shown in Figure 2. The Virtex 5 series is the first FPGA series product that fully utilizes the performance, density and cost advantages of the 65 nm process. It provides 550 MHz DSP48ESl-ice logic chip support and has built-in 25×18MAC, which can achieve single-precision floating-point operations with a resource utilization rate of less than 50%. Its characteristics are: higher speed level, lower power consumption, stronger embedded processor capability, greater DSP computing power, simplified external interface design, and further cost reduction. Therefore, according to the requirements of completing both system control and data acquisition tasks and complex signal processing, the chip meets the system requirements and meets the requirements of low cost, light weight, and low power consumption.
1.4 Virtual coil induction design
In order to extract vehicle information and prevent the mobile target from confusing the system and causing wrong judgment, a method of identifying vehicle models with a virtual coil is adopted. This method combines the detection line with the virtual coil, pre-processes the extracted mobile information, detects the mobile target arriving at the preset position with the trigger line, and then triggers the virtual coil to extract the characteristics of the mobile target. The BP neural network is used to identify whether it is a car. If it is a car, the judgment is "1", and the system enters the next step. If it is not a car, the judgment is "0", and the system is reset, which reduces the system error rate and achieves the purpose of normal operation of the system.
The trigger line is a parallel strip line set at an appropriate position in the image sequence. When a vehicle enters the trigger line, the pixel at the trigger line position at that time is compared with the pixel at the trigger line position of the corresponding background image. If the gray value change of the pixel at the trigger line position is greater than a certain threshold, it is considered that the vehicle has been detected to reach the position of the detection line. As shown in Figure 3.

In Figure 3, the red square represents the virtual coil. The virtual coil is a series of squares consisting of N×N pixels set on the image. Since the vehicle reaches the trigger line, the trigger line detects the pixel change between the two black parallel lines, so a trigger signal is issued. When the trigger signal of the detection line is received, the virtual coil position of the current frame and the background image can be compared to determine whether each small coil is occupied by a vehicle. If it is greater than a certain threshold, it can be determined that a vehicle is occupied, and the specific model can be determined by length, width and area.
1.5 Automatic license plate recognition design
Automatic license plate recognition is one of the main functional modules of the system. Its function is to use the concept that the license plate is the only identifier of the vehicle's identity to intelligently identify and count vehicles. It mainly uses advanced image processing, pattern recognition and artificial intelligence technology to obtain digital information of the vehicle through the acquisition and processing of vehicle images to determine whether the vehicle is a vehicle of this unit.
The automatic license plate recognition module consists of license plate image capture, license plate recognition, and recognition result processing. License plate image capture mainly uses the image acquisition device to effectively transmit the image information collected by the camera to the license plate recognition system for further recognition and processing.
License plate recognition refers to the steps of accurate license plate positioning, filtering and denoising, character segmentation, character recognition, etc. after the collected license plate image is transmitted to the DSP48e processing module of the processor XC5VSX50T chip, and the core recognition algorithm of the license plate recognition module is used to identify the vehicle license plate information. As shown in Figure 4.

Identification result processing After the identification result is sent by DSP, it is compared with the unit's vehicle database inside the FPGA system for verification. If it is a unit's vehicle, it is released and the system is reset; if it is not a unit's vehicle, it goes to the next step.
1.6 Design of automobile infrared light speed measurement
Because the cost of radar speed measurement is too high and the coil speed measurement is prone to aging and damage, the solution adopts the infrared light speed measurement mode. The speed measurement module consists of 4 infrared light sensors and alarm LED lights, which are A, B, C, and D groups. AB is the first group of speed measurement, followed by an alarm and LED speed limit light. CD is the second group of speed measurement. Take group AB as an example, as shown in Figure 5.

In Figure 5, points A and B are two points on one side of the road, each of which is equipped with an infrared light-emitting tube that can emit 38 kHz. The corresponding photoelectric receivers placed at the two opposite points A' and B' receive the light, forming a light-controlled speed measurement area. When a vehicle passes through the light-controlled speed measurement area, the light emitted by point A is first blocked, and the first photoelectric detector A-A' converts the optical signal into an electrical signal and sends it to the signal modulation circuit. After amplification and shaping, it is sent to the FPGA processor. The FPGA receives this trigger signal, turns on the internal counter to start counting, and the value of time t increases accordingly. When the car reaches point B, the second photoelectric detector B-B' converts the optical signal into an electrical signal and sends it to the FPGA processor through the signal modulation circuit to stop counting the counter, and the value of t is determined. The speed v is obtained by dividing the distance S between points A and B by t. If the calculated v value is within the set safe speed range, the LED speed limit light and the alarm circuit will not work. If the vehicle drives from B to A, that is, in the opposite direction, the system will not work. If the v value exceeds the set range, the speed offset △v=Vmax-v is calculated. If △v<"O", the LED speed limit light will flash and the alarm circuit will start working.

2 Software Flow
Xilinx provides the AccelDSP synthesis tool, which is based on the Matlab high-level language to develop Xilinx's DSP module. It can automatically complete the conversion from floating point to fixed point, generate synthesizable VH-DL or Verilog code, and create test benchmark vectors for the verification process. The main functions of the software part are to complete data acquisition, system control and image processing operations. The execution flow of the system software is shown in Figure 6.

3 Conclusions
This paper proposes the design of an intelligent camp gate anti-impact system based on FPGA, and explains the design methods of the main modules. Theoretically verified, the system has a certain effect on preventing terrorists from attacking the camp, can overcome the slow reaction of the camp gate guards in emergency situations, and greatly improve the security of government, military, and enterprise camps.