Implementing this system requires tools such as the OCR algorithm, the Python language, the openCv library, the road vision system and queueing theory. In addition to the aforementioned tools, others such as the matplotlib library, numpy and pandas are also required, all of which will help us achieve our work objective.

Speeding is detected by infrared sensors operating on the laser principle. This mechanism automatically takes photos of the vehicles. This photo is then encrypted and sent automatically to the server [5] .

The character scanned by the OCR algorithm will be used to identify the owners of the offending vehicle, based on an Application Programming Interface (API) linking the server and the public institution in charge of registration [6] . The system will automatically classify the type of offence, the height of the fine, the time/date and location of the offence. At the end of each day, the ITS will send the vehicle owner a notification message including the offence committed, the height of the fine, the time limit for payment of the fine, as well as an indication of the additional fine if the payment deadline is exceeded.

From a methodological point of view, we will use data collected in institutions such as the Driving and Road Safety Police (DR&SP), the Burundi Revenue Authority (BRA), the Burundi Transport Association (BUTRA) and the Burundi Road Agency (BRA). The data obtained from these state institutions will help us to establish a comparison between the taxes collected over the last three years via the quasi-manual system. To achieve this, we intend to simulate the results of traditionally collected taxes using the new Intelligent System. Under this system, all offending drivers will be billed according to the offence committed, while weekly and monthly reports, together with a trend graph showing the reduction in traffic offences and the increase in revenue for the public purse, will demonstrate the relevance of this system [7] [8] .

Graphic 1 compares the use of manual and intelligent systems for collecting road fees. The results show the relevance of the intelligent transport system in Burundi over the last three fiscal years covered by our research.

Graphic 2 also shows the comparison between the use of the manual and intelligent systems in reducing road accidents. The simulation carried out using the intelligent system compared with the manual system shows a clear improvement in accident reduction over the last three years of the study.

Graphic 1. Comparison of road tax data between the manual system and the simulation of the intelligent system.

Graphic 2. Comparison of the increase in road accidents between the manual system and the intelligent simulation system.

The wireless sensor network is one of the technologies applied in ITS. It enables the exchange of data between several control points, while routing the data analyzed by the intelligent algorithm and image processing algorithm to the remote main server [9] . In turn, the control point facilitates the monitoring of nearby intersections.

The Intelligent Transport System is one of the most widely deployed and developed systems in certain cities in African countries that are relatively advanced in terms of transport technology (Kigali, Kinshasa, Libreville, Yaoundé, etc.) [10] [11] . This system is a modern means of information transmission, communication, recording and control, including via the Internet, which are used to increase the level of security, stability, efficiency and improvement of the system for collecting taxes and other charges. The intelligence mechanism in the transport sector is a solution based on New Information and Communication Technologies (NICT). This technology can be used to solve problems linked to road insecurity [12] .

To achieve this [13] , the ITS subsystem should include modules for data reception, processing and decision-making based on data received in the server.

As a result, ITS can be applied to every mode of transport. The main thrust of this innovation is the integration of existing technologies to increase efficiency, resilience and reliability. Intelligent Transport Systems such as traffic management software and security cameras react in real time, helping to protect transport systems and infrastructure from all types of threat [14] [15] .

Optical Character Recognition (OCR) is a method for converting digital images into electronic text. In our academic work, OCR will help the PR&SR to make a trust of the used for license plate detection. During plate detection, this algorithm forms a rectangle around the plate characters as reference points for extraction [16] .

Speed detection will be carried out using infrared sensors installed on poles at each identified intersection. The sensor takes a photo of the offending vehicles, which is then encrypted by an algorithm (OCR) and automatically sent to the National Center for Traffic Regulation Studies (NCTRS) server for identification. The data obtained, including information on the discernment of the license plate number, the critical speed reached, and the exact time and place at which the violation was committed, will be recorded [17] . Technically, the detector (transmitter) transmits a signal to the control point (receiver), located at identified intersection. Technically, the transmitter’s movement must be divided into two parts: the radial movement towards the receiver, which will be affected by the Doppler effect, and the orthogonal movement, which will remain constant [18] .

This is the case with speed cameras. Placed at an angle of 450˚ to the direction of the road, it transmits electromagnetic waves of frequency F in the direction of the vehicle.

$F=f+\Delta f$ (1)

the simple Doppler effect is:

$\Delta f=f_R-f_E$ (2)

While the vehicle is in motion, it will receive waves of a specific frequency f’ and send them back to the radar.

For a double Doppler effect

$\Delta f=2f\cdot V\cdot Cos\left(\theta \right)/C$ (3)

Of which: V: vehicle speed, C: celerity (speed of light or electromagnetic waves), fr: reception frequency, fe: transmission frequency of which:

$\Delta f=f\cdot V/C$ (4)

The infra-red sensor continuously emits two frequencies at a constant speed more or less equal to the celerity; the latter is transmitted at different angles. When a vehicle is hit by the first frequency, the vehicle will reflect this wave at time $t_1$, and the space between this point of reflection and the light source will be calculated as follows:

$E_1=t_1\cdot C$ (5)

The moving vehicle is struck by the second frequency at a new point at time $t_2$. The space between this point and the light source is calculated as

$E_2=t_2\cdot C$ (6)

In addition, the sub-threshold Moving Velocity is calculated as follows:

$V_m=\left(e_1-e_2\right)/\left(t_1-t_2\right)\text{where}e=e_1-e_2\text{and}t=t_1-t_2$ (7) [19]

If the mobile’s speed Vm exceeds the set threshold, an impulse will automatically be sent to a camera installed on the pole near the control point; this enables image detection of the license plate that has exceeded the required speed in the threshold environment, as shown in Figure 1 below:

Note that calculating/measuring speed, distance and time is essential for understanding and quantifying transport movements.

$\text{Speed}\left(S\right)=\frac{D}{T},\text{Distance}\left(D\right)=V\cdot T\text{and Time}\left(T\right)=\frac{D}{V}.$

To find the average speed of a moving object between two given instants of time, we use the following formula:

$V_m=\frac{\text{Change in position}}{\text{Change in time}}$

In this case, the change in position is the difference between the position at time T₂ (x at T₂) and the position at time T₁ (x at T₁). The time change is simply the difference between instants T₂ and T₁. So the formula becomes:

$V_m=\frac{X\text{a}{T}_2-X\text{a}{T}_1}{T_2-T_1}$

In the scientific expression given for velocity $V=\sqrt{\left(5x+36\right)}$, we first need to find the position values at T₁ and T₂ using the given time values (T₁ = 2 seconds and T₂ = 10 seconds). Once we’ve obtained the position values, we’ll be able to use the formula above to calculate the average speed of the mobile between instants T₁ and T₂. The average speed of the mobile between instants T and T + Dt is equal to the ratio (∆X)/∆T, hence the instantaneous speed at instant T, noted v(T), as the limit of this ratio when the real result Dt.

$V\left(T\right)=\text{lim}\frac{x\left(t+\Delta T\right)-x\left(t\right)}{\Delta T}\text{si}\Delta T\to 0$

Hence V(T) by definition, the derivative of the function X and Y gives: F = R → R, i.e. a real-valued function of the variable X. The function F is said to be

derivable in a if the ratio $\frac{f\left(x\right)-f\left(a\right)}{x-a}$ admits a finite limit.

When it comes to transmitting information to the central server, the fiber-optic network facilitates optical data arriving at the server in real time. When it comes to storage, all you need to do is rent a high-speed bandwidth dedicated to your system. The aim here is to avoid response latency [21] .

The data collection technique consisted of a questionnaire containing one or more categories; this served as a guide for interviews conducted in the study area. Questions were put mainly to agents of the PR&SR, ATRABU, OBR, ARB, and to drivers, as well as to users taken individually to get out a given character. These questions provided us with information on: the type of road accidents, the major causes of accidents, the most common location, the number of accidents recorded, the type of equipment used, the types of fines, the fine payment system, the number of roads damaged by vehicles, the road or boulevard permitted for heavy goods vehicles, and the annual rate of fines. All these questions helped us to find out more about how, when and where the fee collection system works, as well as the number of incidents recorded by PR&SR, not forgetting the annual assessment of the road fee collection rate [22] .