APPARATUS AND METHODS FOR WEEDING RAILWAY TRACKS, RAILWAY WEEDING VEHICLES

Description

Equipment and methods for weeding tracks railway, railway weed control vehicles The present invention relates to an apparatus of weed control designed to be installed on a 5 railway vehicle for weed control and a method for weeding of railway tracks which avoids the growth of weed species in proximity of the railway tracks. In particular, the invention relates to an apparatus 10 for weed control using software that implement artificial intelligence logic for optimize the weeding procedure, so that identify weed species and their position. 15 Furthermore, the invention relates to a method for carry out weed control of said herb species pest control with a device that can be installed on vehicles railways, which optimizes weed control methods, trying to avoid waste and errors. 20 In the known art the weed control devices are prepared to be installed on railway vehicles can make use of software based on logics of artificial intelligence (AI) which predict the processing of data matrices, which 25 appropriately reworked provide an output probability function of the presence of species of weeds. Based on a probability function, a unit control operates a weed control unit 30 optimizing its operation so that it uses the weed killer in the necessary quantity and in necessary points, defined by the presence of species of weeds. The data matrices processed by the AI ​​software are generally extracted from a digital input provided from sensors included in the device, for example 5 cameras. The types of sensors can vary depending on of the devices, in addition to video cameras that detect the visible spectrum there may also be infrared cameras, tachometers, accelerometers, 10 and so on. Unfortunately, in the prior art the matrices are reworked, preferably through operations of convolution, without using factors different multiplicatives for different matrices, in 15 based on considerations regarding the origin of the data and data quality depending on conditions external factors that could affect said quality. In this way, data fusion may not be optimal, leading to imprecise interventions or 20 redundant with consequent waste of agent herbicide or, in the absence of intervention on some weed species. Therefore, it is clear how much it can be important to rework the data matrices in a way 25 that the latter are recombined with factors different multipliers depending on their quality, which can be determined by several factors external, in order to avoid waste and bad operations. 30 The technical problem underlying this invention is to provide a weed-killing apparatus designed to be installed on a vehicle railway for weeding and a method for weeding of railway tracks that allow to avoid the drawbacks mentioned with reference to the known technique. 5 In its first aspect, this problem is solved from a weed control apparatus set up to be installed on a railway vehicle for weed control as defined in appended claim 1. This weed control apparatus is, specifically, 10 prepared to be installed on a vehicle railway for weed control. It includes a first camera that sets up to detect at least the visible spectrum, and it is capable of acquiring a first digital image of a 15 portion of railway track, and said first image is expressed digitally through a first array of pixel values. Additionally, a second camera is provided which is designed to detect at least the spectrum of the 20 infrared, and is therefore configured to acquire a second digital image of a portion of railway track, and the second image, similarly to the first, it is expressed digitally through a respective second matrix of values 25 pixels. The apparatus then comprises a device for remote sensing, of the LIDAR type, which operates with a laser emitter and is configured to acquire a digital scan of a portion of the track 30 railway, and this digital scan is expressed digitally through a respective third matrix of pixel values. In addition to the above-mentioned image-detecting devices, the apparatus includes a weeding unit, supplied to weed out invasive species, and a module of control that implements a first software that in its 5 times it is able to receive a signal from detection of pest species and to control the spraying of chemical agent from said module feeding through at least one nozzle, acting based on the species detection signal 10 weeds. In addition, an identification unit is provided of the invasive species which in turn includes a communication form prepared to receive, the first digital image, second image 15 digital and digital scanning and sending a pest species detection signal called weed control module. It also includes a memory module, suitable for memorize said first and second images and said 20 scan, and a processing module that features a second software designed to run at least a convolution operation between a matrix of pixels of each digital image and of the digital scanning and a respective kernel 25 default stored in memory module for obtain a representation matrix of the characteristics of each digital image and of digital scanning. The processing module, through the said second 30 software, can thus calculate at least one first, one second and third probability of presence and location of a pest species, in a database of infesting species, being said first, second and third probability of presence and position respectively associated with the said first image digital, to said second digital image and to said scan, so that the said module 5 processing identifies the presence and position of said invasive species, At this point, the processing module generates a pest species detection signal on the basis of a resulting probability of presence and 10th position, calculated as the sum of the first, second and third probability of presence and position, respectively multiplied by a prime, a second and a third multiplying factor defined based on the presence of precipitation and 15 brightness; in addition, the processing module sends said detection signal to the control module which is designed to activate said unit weed control, so that it only acts in areas where where invasive species have been detected. 20 The main advantage of the weeding apparatus lies in multiplying each matrix by feature representation corresponding to a different data source, in particular to the first, to the second image 25 digital and digital scanning, for a factor multiplicative which is customized based on the weather conditions and time of day (day / night). In this way, more weight is given to more reliable data; for example, during the day 30 and in the absence of rain the first camera is from consider yourself very reliable, possibly more reliable of the second camera and the remote sensing device; otherwise, of at night and in the presence of atmospheric precipitation or heavy fog, the second camera is more reliable first camera and device remote sensing; or at night and with conditions 5 serene, the remote sensing device is considered the most reliable. In a second aspect, this problem is solved by a method for weeding tracks railways as defined in the attached claim 10 9. The method for weeding railway tracks involves to use a weed control device designed for be installed on a railway vehicle as is defined in the said first aspect. 15 Furthermore, specifically, the weeding method It involves the stages of acquiring a first image digital in the visible of a portion of the track railway, a second digital image in the infrared of a portion of the track 20 railway and a laser scan of a portion of railway track, each of the said images and said scansion being expressed with a respective a matrix of pixels. Then, through the said communication form, 25 images acquired and the scan is sent to a processing module and store and, through said second software, a convolution between a matrix of pixels of each digital image and digital scanning and a 30 respective kernel stored in the memory module to obtain a representation matrix of the characteristics of each digital image and of digital scanning. Subsequently, through the said form of processing that implements said second software, at least one probability of presence is calculated and 5 location of at least one pest species in a database of pest species, from a matrix of feature representation of the image, identifying the presence and the position of said invasive species. They are then 10 calculated, through the said processing module, a first, a second and a third multiplying factor defined based on the presence of precipitation and of brightness. The first multiplying factor is then 15 attributed to a first probability of presence and position of the first digital image, called second multiplying factor to a second probability of presence and position of the second digital image, and the so-called third factor 20 multiplicative to a third probability of presence and digital scan position. A resulting probability of is then calculated. presence and position, through the said software, as the sum of the first, second and third probabilities of 25 presence and position, respectively multiplied for the first, second and third factor multiplicative, and then it is generated, through said processing module, a signal of detection of pest species, based on the 30 the aforementioned resulting probability, which is sent to the weed control module for the implementation of weed control operations through called weed control unit. The method as well as the apparatus has the advantage of weight the received data with multiplication factors which may be different from each other The present invention will be hereinafter 5 described according to an example of its implementation preferred, provided for illustrative purposes only and not limiting with reference to the attached drawings in which: ∗ Figure 1 shows a block diagram of a 10 weed-control apparatus according to the invention; ∗ Figure 2 shows the identification of an area of intervention; ∗ Figure 3 shows the identification of further intervention areas in addition to the intervention area of 15 figure 2; ∗ Figure 4a shows a representation Schematic of input data processing to the weed control apparatus; ∗ Figure 4b shows a representation 20 schematic of the data convolution process; ∗ Figure 5 shows a schematic representation front of a railway vehicle on which it is mounted installed the weed control device according to the figure 1; and 25 ∗ Figure 6 shows a schematic representation from the top of a railway vehicle on which it is installed the weed control device according to the figure 1. With reference to the figures, a weeding apparatus, 30 designed to be installed on a vehicle railway for weed control, is indicated in its complex with 1. The apparatus 1 comprises a first camera 3 designed to detect at least the spectrum of the visible and being able to acquire a first 5 digital image of a portion of the track railway 4, said first image including a first array of pixel values. The apparatus 1 then includes a second camera 5 designed to detect at least the spectrum of the 10 infrared and being able to acquire a second digital image of a portion of track railway 4, said second image including a second array of pixel values. The apparatus 1 further comprises a device 15 remote sensing 6, of the LIDAR type, including a laser emitter, the remote sensing device being capable of acquiring a digital scan of a portion of railway track 4, called digital scanning comprising a matrix of 20 pixel values. Apparatus 1 also includes a weeding unit designed to weed out invasive species. Finally, the apparatus 1 comprises a module of control 12, which implements a first software 25 set to receive a detection signal of weed species and to command weeding through said weeding unit on the basis of said pest species detection signal and a 13 species identification unit 30 weeds. This identification unit 13 includes, in its time, a communication module 14 that is prepared to receive, the first digital image, the second digital image and digital scanning and send a species detection signal pests to said control module 12 5 of the weeding apparatus 1. It also includes a 15 act memory module to memorize the said first and second images and the said scanning, and a 16 processing module including a second software that is prepared 10 to perform at least one convolution operation between a matrix of pixels of each image digital and digital scanning and a respective default kernel stored in memory module 15, in order to obtain a matrix of 15 representation of the characteristics of each digital image and digital scanning, The processing module 16, through said second software, calculates at least a first, a second and a third probability of presence and position of a 20 pest species, in a species database weeds. The first, second and third probability of presence and position are respectively associated with said first digital image, said second image 25 digital and said scan, so that said processing module 16 identifies the presence and the location of said pest species The processing module 16 generates a signal of detection of pest species based on 30 a resulting probability of presence and position, calculated as the sum of the first, second and third probability of presence and position, respectively multiplied by a first multiplying factor F, a multiplying factor according to F and a third FI multiplication factor multiplication factors 3. are defined on the basis of the presence of 5 precipitation and light conditions. The processing module 16 sends the signal of detection to control module 12 which, in its time, it is set to activate the unit of weed control, so that it only acts in areas where 10 where invasive species were detected. Alternatively, according to another form of realization of the invention that integrates what described above, the processing module 16 may comprise a plurality of nuclei of 15 parallel processing, and each core of parallel processing is designed to perform, through said second software, at least one operation of convolution between a submatrix constructed by neighboring pixels of each digital image and the 20 digital scan and a respective kernel default, which is stored in the memory module 15, to obtain a submatrix of representation of the characteristics of each digital image and digital scanning 25 assigning a value to each pixel in the image. Preferably, the weeding apparatus comprises a 100 external storage space, optionally even in the cloud, on which the crops when said weed control apparatus 1 is in use are saved and 30 stored for future use. Furthermore, the processing module 16, through the according to software, it can calculate at least one probability of presence and location of a species pest in a database of pest species from each representation matrix of the characteristics of the image constructed by the 5 submatrices of representation of the characteristics, in turn built by the nuclei of parallel processing, identifying the presence and the location of said weed species, and can generate a species detection signal 10 pests based on a resulting probability of presence and position, calculated as the sum of the first, second and third probability of presence and position, respectively associated with the said first digital image, said second digital image 15 and said scansion, respectively multiplied for a first, a second and a third factor multiplicative defined based on the presence of precipitation and brightness. This detection signal is then transmitted 20 to the control module which is designed to activate the spray unit, so that it acts only in areas where species have been detected weeds. In some embodiments, the apparatus 25 weeding 1 includes a first device of brightness detection set for identify a brightness parameter λ, where, if λ > λ the presence of light is considered, that is min enough brightness to take pictures, and 30 on the other hand it's dark. Additionally, a second device can be provided of precipitation detection in the form of rain and / or hail and / or snow, which is set up for identify an atmospheric parameter μ, where if μ > μ it is established that there is precipitation min such as to influence the weeding process, or on the contrary, he is serene. 5 If the first and second detection device are present, the communication module can be set to receive the brightness parameter λ and the atmospheric parameter μ and where the said module of processing 16, through said software, 10 establishes the said first, second and third factors multiplicative based on the parameter of luminosity λ and the atmospheric parameter μ, in which -if λ > λ and μ < μ then F > F ≥ F, or min min 1 2 3 15 -if λ > λ and μ > μ then F > F > F, or min min 2 1 3 -if λ < λ and μ < μ then F ≥ F > F, or min min 3 2 1 -if λ < λ and μ > μ then F > F > F min min 2 3 1. In this way, depending on the weather conditions and to the brightness, it is possible to define three factors 20 multiplicatives F , F , F , which will weigh the respective 1 2 3 matrices depending on the greater reliability of each of them. In fact, in the presence of rain, heavy fog, snow or hail the scan obtained with the device 25 of remote sensing 6 is to be considered less reliable first and second digital image, which in turn could be considered with different multiplication factors depending on whether detected day or night. 30 According to the realizations shown in the figure, the weeding unit includes a spraying unit 7 which features a power module 8 with a 9 tank of chemical agent and a valve delivery 10, and at least one nozzle 11, for spraying a chemical agent from said power module 5 8. Control module 10 is designed to receive a pest detection signal and for command the spraying of chemical agent from feed module 8 through nozzle 11 10 based on species detection signal weeds. Furthermore, the processing module 16 sends the detection signal to control module 12 which is set to activate the spray unit 4, 15 so that it acts only in areas where they are No pest species were detected. According to other preferred embodiments of the invention, the processing module 16, using the second software, calculate at least one 20 resulting probability that, in addition to the presence and the position, consider the typology of at least one invasive species, identifying, precisely, the presence, position and type of the species infesting. 25 The processing module 16 then generates a signal of pest species detection on the basis of of the resulting probability of presence, position and type, and sends said detection signal to said control module 12 arranged for 30 activate, via the second software, the unit of weed control, so that it only acts in areas where where invasive species have been detected and preferably with a dosage of chemical agent suitable for the type of infesting species. In addition to a personalized dosage, for weeding specific types of pest species, the module 5 power supply 8 can include at least two tanks, each with a different chemical agent from others, and the processing module, through said second software, can generate a signal of detection of different weed species on the 10 basis of said resulting probability of presence, position and type and send said signal to detection to said control module 12 prepared to activate the spray unit 7, so that it sprays the different chemical agents according to 15 to the type of infesting species. According to further preferred embodiments of the invention, the weeding unit comprises a weed-control device designed to emit a electric discharge and / or emit a laser beam. 20 When the weed control device is present, the control module 10 is designed to receive a pest detection signal and for command the activation of said device weeding, based on said detection signal 25 of infesting species and in which the said module of processing 16 sends said detection signal to the said control module 12 which is set up for activate the said weed-killing device, so that operate only in areas where they have been detected 30 invasive species. When both spray unit 7 are present both the weeding device, the module of processing 16, through said second software, calculate at least one resulting probability of presence, position, typology of at least one species infesting, identifying the presence, the position 5 and the type of said at least one pest species and generates a detection signal of the different invasive species based on said probability resulting from presence, position and typology and sends said detection signal to said module 10 control 12 ready to activate selectively, through said software, the unit of spraying 7 and / or the weed control device according to which is considered most suitable for weed control a type of pest species, in a 15 certain position, and so that they act only in areas where the invasive species. In fact, in addition to the type of infesting species, in this case the position is also relevant, 20 since, preferably, the weed control device It is positioned underneath the railway vehicle 2. This preferential positioning could make the weed control device designed to act on underneath railway vehicle 2, then on the tracks 25 4 and in the vicinity of these, without reaching areas very far from the railway vehicle 2. In In particular, this limitation could concern mainly the possible realization of the weed-killing device that acts by discharge 30 electric. According to more advanced embodiments, the weeding apparatus includes an anemometer designed to detect an intensity parameter of the W wind, and / or a speedometer set up for detect a vehicle speed parameter V on which the device travels. In the presence of an anemometer and / or tachometer, the module 5 of communication 14 can receive said parameter of wind intensity W and / or the so-called speed parameter V e, said memory module can store them and the processing module 16, through said software, verifies that the conditions W ≤ W ≤ W and that V ≥ min max 10 V are satisfied being W and W values ​​of min, min max wind intensity parameter threshold stored in said memory module 15, and being V is a threshold value of the speed parameter min stored in said memory module 15. 15 If the conditions are met, the form of control receives said detection signal and controls the weeding of weed species through said weeding unit, based on said signal for the detection of pest species. 20 An example of convolution operations is schematically reported in figure 4, in which are obtained: a parameter ∈ (0,1) which indicates the activation or not of the weed control apparatus 1, a parameter P1 and one P2 which indicate respectively 25 activation of the spray unit 7 and the weed control device and a parameter that indicates the intensity to apply to different nozzles. To carry out weeding of the railway tracks 4 the the first and second software can take advantage of 30 Artificial Intelligence Logics. A favorite method for weeding railway tracks 4 plans to activate the device of weeding, acquire a first digital image in the visible part of a railway track 4, a second digital infrared image of a portion of railway track 4 and a scan 5 lasers of a portion of railway track 4, provided that each of the said images and said scan is expressed as a respective matrix of pixels. In order to process pixel matrices, the module 10 of processing 16 receive via the said form communication 14 the acquired images and the scan in turn stored in the module memory 14. Preferably, the processing module 16 performs 15 a convolution, through the second software, between a matrix of pixels of each digital image and digital scanning and a respective kernel stored in memory module 15, to obtain a representation matrix of the 20 characteristics of each digital image and of digital scanning. The method then provides for calculating, through the processing module 16 which implements the second software, at least a probability of presence and 25 location of at least one pest species in a database of pest species, from a matrix of feature representation of the image, identifying the presence and the location of said pest species. 30 Furthermore, the processing module 16 can calculate a first, a second and a third factor multiplicative defined based on the presence of precipitation and brightness and can multiply them, respectively, with a first probability of presence and position of the first digital image, with a second probability of presence and position 5 of the second digital image and with a third probability of presence and location of the scan digital to calculate a resulting probability of presence and position. Finally, the processing module 16 generates a 10 signal of detection of pest species on the basis of said resulting probability of presence and location of a pest species and sends, via said communication module 14, said signal detection of pest species on the module 15 control 12 which can control the weeding of the weed species, through the said weed control unit, based on the species detection signal weeds. The weeding unit may include the unit of 20 spraying 7 and / or the weed control device already previously described. The detection signal may also include information on the quantity of chemical agent to be apply based on the type of pest species. 25 If the spray unit 7 includes two or more tanks, each containing a chemical agent different from the others, the detection signal is based on a probability of presence position and typology and may include information on the different 30 herbicides to use based on the type of species infesting and active said spraying unit 7 in so that it sprays the different chemical agents according to to the type of infesting species. If the weeding apparatus 1 includes both the unit of spraying 7 is the weed control device, the processing module can perform the choice between 5 which is better to use or whether to use both and send the detection signal, including the choice, to the said control module which can activate selectively, through the said first software, said spray unit 7 and / or said spraying device 10 weed control. According to some preferred embodiments, in where the weeding apparatus includes an anemometer and a speedometer, the activation of the apparatus weed control 1 involves measuring the wind intensity 15 at a certain instant of time t with the anemometer and store a wind intensity parameter W(t), measure the speed of the railway vehicle at a certain moment in time you memorize a speed parameter V(t) in memory module 15 20 and communicate them, via the aforementioned communication form 14, to the said processing module 16. Finally, the activation of the weed control system plans to verify, through the said second software implemented on said module 25 processing 16 that the conditions W ≤ W(t) ≤ W and min max that V(t) ≥ V are satisfied if these conditions min , they are satisfied, the weed control apparatus 1 is proceeding with subsequent actions, alternatively not proceeds. 30 When the weeding apparatus 1 comprises a speedometer, the weeding method may include measure the speed of the railway vehicle 2 and store a speed parameter V in the module memory 15 and communicate it via the said form communication 14 to said processing module 16. Memory module 15 can also communicate a time 5 t of digital image processing and And scanning and an activation time t of the unit to weeding, both being saved in the module memory, to said processing module 16. The processing module 16 can calculate, via 10 said second software, a minimum distance of one relief area 200 X = V(t +t ) to be taken into account Dmin and account in the processing of the detection signal of invasive species. Identifying the relief area 200 allows you to 15 process images and scans of reality surrounding present at a distance from the point of installing the weed control unit on the vehicle railway 2, and this allows the apparatus to weed control to act effectively for weed control 20 invasive species. Preferably, the weeding apparatus 1 allows for identify a relief area 200 and the areas neighboring 201, which correspond to different distances from the point of installation of the weed control unit on the 25 railway vehicle 2. To the above described weed control system and related method of a technician in the field, with the aim of satisfying further and contingent needs, may bring numerous further modifications and variations, all 30 also included in the scope of protection of the present invention, as defined by claims attached.

Claims

1. A weed control apparatus (1) designed to be installed on a railway vehicle (2) for weed control, comprising: • a first camera (3) designed to detect at least the visible spectrum and being capable of acquiring a first digital image of a portion of railway track (4), said first image comprising a first matrix of pixel values; • a second camera (5) designed to detect at least the infrared spectrum and being capable of acquiring a second digital image of a portion of railway track (4), said second image comprising a second matrix of pixel values; • a remote sensing device (6), of the LIDAR type, comprising a laser emitter, the remote sensing device being capable of acquiring a digital scan of a portion of railway track (4),said digital scan comprising a matrix of pixel values; • a weeding unit (77) arranged to weed out weed species; • a control module (12) implementing a first software arranged to receive a weed species detection signal and to control weeding via said weeding unit on the basis of said weed species detection signal; • a weed species identification unit (13) comprising: * a communication module (14) arranged to receive the first digital image, the second digital image and the digital scan and to send a weed species detection signal to said control module (12) of the weeding apparatus (11), * a memory module (15) arranged to store said first and second image and said scan,and * a processing module (16) comprising a second software arranged to perform at least one convolution operation between a pixel matrix of each digital image and of the digital scan and a respective predefined kernel stored in the memory module (15) to obtain a matrix representing the characteristics of each digital image and of the digital scan, wherein the processing module (16), through said second software, calculates at least a first, a second and a third probability of presence and position of a pest species, in a database of pest species, said first, second and third probabilities of presence and position being respectively associated with said first digital image, said second digital image and said scan, in such a way that said processing module (16) identifies the presence and position of said pest species,and wherein the processing module generates a pest species detection signal based on a resulting probability of presence and location, -23 calculated as the sum of the first, second and third probabilities of presence and location, respectively multiplied by a first (Fi), a second (F2) and a third multiplicative factor (F3) defined on the basis of the presence of precipitation and brightness, and wherein said processing module sends said detection signal to said control module set up to activate said weed control unit so that it acts only in areas where pest species have been detected.

2. Weed control apparatus (1) according to claim 1, comprising a first brightness detection device and / or a second precipitation detection device in the form of rain and / or hail and / or snow, wherein said first detection device is arranged to identify a brightness parameter λ, wherein if λ > Xmin there is light, vice versa it is dark and, wherein said second detection device is arranged to identify an atmospheric parameter μ, wherein if μ > μmin there is precipitation, vice versa it is clear.

3. Weed control apparatus according to claim 2, wherein said communication module (14) is arranged to receive the brightness parameter λ and the atmospheric parameter μ and wherein said processing module (16), via said software, establishes said first, second and third multiplication factors on the basis of the brightness parameter λ and the atmospheric parameter μ, wherein - if λ > λ!^ and μ < μmin then Fi > F2 h F3, or - if λ > λϋίπ and μ > μmin then F2 > Fi > F3, or - if λ < Àmin and μ < μmin then F3 h F2 > Fi, or - if λ < Àmin and μ > μmin then F2 > F3 > Fi.

4. A weed control apparatus according to any of the preceding claims, wherein said weed control unit has a spraying unit (4) comprising: * a supply module (8), including a chemical agent tank (11) and a dispensing valve (12); and * at least one nozzle (9) for spraying a chemical agent from said supply module (8); wherein said control module (10) is arranged to receive a pest species detection signal and to control the spraying of chemical agent from said supply module (8) through said at least one nozzle (11) based on said pest species detection signal and wherein said processing module (16) sends said detection signal to said control module (12) which is arranged to activate said spraying unit, so that it acts only in areas where pest species have been detected.

5. Weed control apparatus according to claim 4, wherein said supply module (8) comprises at least two tanks each with a chemical agent different from the others and wherein said processing module (16), via said second software, generates a detection signal of the different weed species on the basis of a resulting probability of presence, position and typology and sends said detection signal to said control module (12) arranged to activate said spraying unit (7), so that it sprays the different chemical agents based on the type of weed species.

6. Weed control apparatus (1) according to any of claims 1 to 3, wherein said weed control unit comprises a weed control device arranged to emit an electric discharge and / or emit a laser beam, and wherein said control module (10) is arranged to receive a weed detection signal and to command the activation of said weed control device, on the basis of said weed detection signal and wherein said processing module (16) sends said detection signal to said control module (12) which is arranged to activate said weed control device, so that it acts only in areas where weed species have been detected.

7. Weed control apparatus (1) according to claims 4 and 6, wherein the processing module (16), through said second software, calculates at least one resulting probability of presence, position, typology of at least one infesting species, identifying the presence, position and typology of said at least one infesting species and, wherein the processing module (16) generates a detection signal of the different infesting species on the basis of said resulting probability of presence, position and typology and sends said detection signal to said control module (12) arranged to selectively activate, through said software, said spraying unit (7) and / or said weed control device -26 based on which is considered most suitable for weeding a type of infesting species, in a given position, and, so that they act only in areas where the infesting species have been detected.

8. Weed control apparatus according to any of the preceding claims, comprising an anemometer arranged to detect a wind intensity parameter W, and / or comprising a tachometer arranged to detect a speed parameter V.

9. A weed control apparatus according to claim 8, wherein said communication module (14) is arranged to receive said wind intensity parameter W and said speed parameter V e, wherein said memory module (15) is suitable for storing said wind intensity parameter and said speed parameter e, wherein the processing module (16), via said software, verifies that the conditions Wmin < W < Wmax and that V > Vmin are satisfied, Wmin and Wmax being threshold values ​​of the wind intensity parameter stored in said memory module (15) and Vmin being a threshold value of the speed parameter stored in said memory module; if the conditions are satisfied, said control module (12) receives said detection signal and controls the weed control of the weed species via said weed control unit, based on said weed species detection signal.

10. Method for weeding railway tracks (4) carried out using a weeding apparatus (1) designed to be installed on a railway vehicle (2), the apparatus (1) comprising a first camera (3) designed to detect at least the visible spectrum, a second camera (5) designed to detect at least the infrared spectrum, a remote sensing device (6), of the LIDAR type, comprising a laser emitter, a weeding unit, a control module (12) implementing a first software, a weed species identification unit (13) including a communication module (14), a memory unit (15) and a processing module (16) implementing a second software based on artificial intelligence logic, the method comprising the steps of: * activating said weeding apparatus (1);* acquiring a first digital image in the visible range of a portion of railway track (4), a second digital image in the infrared range of a portion of railway track (4) and a laser scan of a portion of railway track (4), each of said images and said scan comprising a pixel matrix, * receiving via said communication module (14) the acquired images and the scan in the processing module (16) and storing said images and said scan in the memory module (15); * performing, via said second software, at least one convolution between a pixel matrix of each digital image and of the digital scan and a respective kernel stored in the memory module (15) to obtain a 28-characteristic matrix representing each digital image and of the digital scan;* calculate, through said processing module (16) which implements said second software, at least one probability of presence and position of at least one pest species in a database of pest species, from a matrix representing the characteristics of the image, identifying the presence and position of said pest species; * calculate, through said processing module (16), a first (Fi), a second (F2) and a third (F3) multiplying factor defined on the basis of the presence of precipitation and brightness; * attribute, through said processing module, said first multiplying factor (Fi) to a first probability of presence and position of the first digital image, said second multiplying factor (F2) to a second probability of presence and position of the second digital image, said third multiplying factor (F3) to a third probability of presence and position of the digital scan;* calculate a resulting probability of presence and position, via said software, as the sum of the first, second and third probabilities of presence and position, respectively multiplied by the first (Fi), second (F2) and third (F3) multiplying factor; * generate, via said processing module (16), a pest species detection signal based on said resulting probability of presence and position of a pest species -29 and send, via said communication module (14), said pest species detection signal to the control module (12); and * command, via said control module (12) the weeding of the pest species by said weeding unit based on the pest species detection signal.; 11. A method according to claim 10, wherein said weeding unit comprising: * a spraying unit (7) including: * a supply module (8), including a chemical agent tank (11) and a dispensing valve (12); * at least one nozzle (9) for spraying a chemical agent from said supply module (8); and / or * a weeding device arranged to emit an electric discharge and / or emit a laser beam, wherein the method comprises the steps of: * calculating, through said processing module, a resulting probability of presence, location and type of at least one pest species, by identifying the presence, location of said at least one pest species; * generating, through said processing module, a pest species detection signal based on said resulting probability of presence, location and type;* select, via said second software, which of said spraying unit and / or said weeding device is considered most suitable -30 for weeding a type of weed species and include the information of the choice in said detection signal; and * send said detection signal, via said communication module, to said control module set up to selectively activate, via said software, said spraying unit and / or said weeding device based on the type and location of the weed species, so that they act only in areas where the weed species have been detected.; 12. Method according to claim 10 or 110, wherein the weeding apparatus comprises an anemometer and a tachometer, and wherein activating said weeding apparatus comprises the steps of: * measuring the wind intensity at a certain instant of time t with said anemometer and storing a wind intensity parameter W(t); * measuring the speed of the railway vehicle at a certain instant of time t and storing a speed parameter V(t) in the memory module and communicating them via said communication module to said processing module; * verifying, via said second software implemented on said processing module, that the conditions Wmin ù W(t) ù Wmax and that V(t) à Vmin are satisfied. If these conditions are satisfied, the weeding apparatus proceeds with the subsequent actions, alternatively it does not proceed.