Light-based detection system for a motor vehicle

Abstract

The invention relates to a detection system (1) for a motor vehicle, the system comprising an emission module (2) capable of emitting a visible light beam contributing to a lighting or signalling function, modulated on the basis of an emission signal (Sseq) comprising the repetition of emission pulse trains at a predetermined emission frequency; a receiving module (3) capable of receiving a reflection of the emitted beam (F1) off an object (O) in the environment, comprising at least one photodetector; and a computing unit (4) configured to discretely compute correlation values (Vcorr) between the received signal (Sel) and the emission signal (Sseq) according to a predetermined correlation time interval, the computing unit (4) being configured to identify specific information, from the correlation values (Vcorr), corresponding to the nature and / or distance of the object (1).

Description

DESCRIPTION TITLE: Light detection system for a motor vehicle

[0001] The present invention relates to a light detection system for a motor vehicle. It also relates to a headlight for a motor vehicle comprising such a light detection system. It also relates to a motor vehicle itself, comprising such a headlight or light system. Finally, it relates to a detection method for a motor vehicle, as well as a driver assistance method for a motor vehicle.

[0002] Various optical detection systems for motor vehicles designed to assist the driver are known from the state of the art. Some of these optical detection systems take the form of optical sensor arrangements. These arrangements typically include one or more optical detection devices, such as scanning lidar sensors. Such an optical detection device can monitor an area surrounding the motor vehicle and, for example, detect objects in the surrounding area and provide a driver assistance system with information about the detected object, such as the distance between the object and the motor vehicle or the object's geometric dimensions.

[0003] In scanning lidar sensors, an infrared laser beam is emitted onto a deflection device, such as a reflecting mirror, and reflected back into the surrounding area. The deflection device is typically movably mounted around a rotational axis and can be driven to rotate about this axis. The laser beam is thus deflected by a specific angle at predefined scan intervals or times, and the surrounding area is thereby scanned in a given direction.

[0004] These devices are, however, complex and bulky. Furthermore, they must be positioned strategically within the vehicle to effectively observe the surroundings. Their integration into a vehicle Automotive surveillance is therefore generally a delicate matter. Furthermore, the digital signals provided by these optical detection systems can, in some cases, lead to a false perception of the environment, which can endanger vehicle occupants. Finally, they require complex calculations to process large amounts of data if detailed information about a vehicle's surroundings is desired, a condition that sometimes proves difficult to achieve due to physical limitations, computation time, or the cost of the devices.

[0005] The aim of the invention is to provide an optical detection system that can be easily integrated into a vehicle, making it possible to improve the reliability of environmental perception, in particular by obtaining detailed information on any object positioned in the environment of a vehicle.

[0006] To this end, a first aspect of the invention relates to a light detection system for a motor vehicle comprising: an emission module including a light module capable of emitting a light beam whose spectrum has at least a portion in the visible spectrum, the light beam contributing to a lighting or signaling function, and a modulation unit arranged to modulate said light beam emitted by the light module from an emission signal (Sseq) comprising a train of emission pulses, so as to emit a modulated emitted light beam (F1) comprising the repetition of elementary pulse trains according to a predetermined emission frequency;a receiving module capable of receiving a received light beam (F2), the received beam originating at least in part from a reflection of the emitted light beam (F1) on an object (O) in the environment, comprising an acquisition module including at least one photodetector capable of converting a received light signal into a received electrical signal (Sel); a computing unit configured to calculate correlation values ​​(Vcorr) discretely between the received electrical signal (Sel) and the emitted signal (Sseq) for several relative time offset values ​​of these two signals, these time offset values ​​being defined according to a predetermined correlation time step, in which the computing unit is configured to implement a processing of the correlation values ​​(Vcorr), so as to identify specific information from the correlation values ​​(Vcorr), such as one or more maxima, which correspond to a time offset corresponding to a time of flight of an emitted light beam (F1) until the reception by at least one photodetector of the part of the received light beam (F2) reflected on the object (O), to deduce the nature and / or distance of said object positioned in the environment of the light detection system.

[0007] Advantageously, the processing of correlation values ​​(Vcorr) implemented by the computing unit consists of a mathematical interpolation of the correlation values ​​to generate a continuous correlation function (Fcorr) and / or a continuous derivative function of the correlation function (Fcorr), notably by the least squares method, so that the distance of an object positioned in the environment is calculated by finding the maxima of the continuous correlation function (Fcorr).

[0008] Advantageously, said mathematical interpolation is based on the least squares method, considers a predefined mathematical function, such as a cardinal sine, approximated by harmonics based on the predetermined emission frequency, associated with coefficients in the form of polynomials, in particular third-degree polynomials.

[0009] In one embodiment of the invention, the processing of correlation values ​​(Vcorr) implemented by the computing unit consists of estimating the distance and / or automatically identifying the nature of an object positioned in the environment associated with correlation values ​​by an embedded artificial intelligence tool that has undergone learning.

[0010] In one embodiment of the invention, at least one photodetector of the acquisition module is sized to receive a received light beam (F2) from a surface of the environment of dimension less than a predefined threshold, so that said surface is considered in a simplified way as a point and in that the processing of the correlation values ​​(Vcorr) implemented by the computing unit consists of a mathematical interpolation of the correlation values ​​to generate a continuous correlation function and / or a continuous derivative function of the correlation function.

[0011] In one embodiment of the invention, at least one photodetector of the acquisition module is sized to receive a received light beam (F2) from a surface of the environment with a dimension greater than a predefined threshold, so that said surface is considered three-dimensional, and in that the processing of the correlation values ​​(Vcorr) implemented by the computing unit consists of an estimation of the distance and / or an automatic identification of the nature of an object positioned in the environment associated with the correlation values ​​by an embedded artificial intelligence tool that has undergone learning.

[0012] Preferably, the predetermined transmission frequency is between 10 MHz and 100 MHz.

[0013] Advantageously, the predetermined correlation time step is greater than or equal to 500ps, or greater than or equal to 1 ns, and optionally less than or equal to 4ns.

[0014] Preferably, the acquisition module of the receiving module includes an array of at least 32 photodetectors, and optionally of less than 4000 photodetectors.

[0015] Advantageously, the emission module emits a beam in a visible spectrum comprising a first part, notably in the blue spectrum, comprising said modulated emitted light beam intended for detection of at least one object in the environment, and comprising a second part intended to complement the first part so that the beam contributes to the function of lighting or signaling, such as a daytime running light, a high beam or a low beam.

[0016] In a particular embodiment, the light detection system includes a neural classifier configured to process separately the correlation histograms from each photodetector in a group of photodetectors selected from the photodetector array. The classifier receives as input the individual correlation histograms of each photodetector in the group, each histogram representing the correlation values ​​(Vcorr) calculated for different time shifts. Preferably, the classifier also receives a vector of relative coordinates indicating the position of each photodetector with respect to the center of the group. Even more preferably, the classifier receives a vector of absolute coordinates indicating the position of each photodetector in the complete photodetector array.This architecture allows the classifier to exploit not only the temporal information contained in the correlation histograms, but also the spatial relationships between the different photodetectors, which is particularly advantageous for recognizing complex patterns. Preferably, the classifier is implemented as a neural network comprising a photodetector processing branch, composed of one-dimensional convolutional layers to extract the temporal features from the histograms, a fusion layer that integrates relative and absolute position information, a graph network that models the spatial relationships between the photodetectors, and fully connected layers that produce the final classification. This embodiment is particularly effective for accurately identifying the nature of complex objects positioned in the environment of a motor vehicle.

[0017] In another embodiment, the light detection system includes a classifier configured to process an aggregated representation of the correlation histograms of a group of photodetectors. For each identified group of photodetectors, the processing unit sums the individual correlation histograms, producing an aggregated histogram that represents the collective response of the group. This approach has the advantage of significantly reducing the dimensionality of the classifier's input data while preserving the main features of the temporal signature of the detected object.For example, the classifier is implemented as a neural network comprising one-dimensional convolutional layers to extract features from the aggregated histogram, attention layers that focus on the most informative portions of the histogram, and fully connected layers that produce the final classification. This implementation is particularly well-suited for objects with a distinctive temporal signature but a less discriminating spatial distribution, such as highway guardrails with regularly spaced reflectors. It also offers the advantage of high computational efficiency, enabling real-time processing even with limited resources.

[0018] In an advantageous embodiment, the light detection system includes an adaptive analysis module configured to dynamically define groups of contiguous photodetectors based on preliminary correlation results. The process takes place in two phases. In a first preliminary analysis phase, the computing unit performs an initial calculation of the correlation values ​​(Vcorr) for all photodetectors in the array. Preferably, mathematical interpolation is applied to generate a continuous correlation function for each photodetector. Even more preferably, a segmentation algorithm identifies regions of interest where the correlation values ​​exceed an adaptive threshold. In a second phase of In dynamic clustering, for each region of interest, a region growth algorithm defines a group of contiguous photodetectors. Preferably, the growth criteria include spatial proximity and similarity of correlation functions. Even more preferably, the shape and size of the group are optimized to encompass the entire signature of the detected object. The groups thus formed are then processed by one of the classifiers described in the preceding embodiments, enabling targeted analysis of the detected objects. This dynamic approach has the advantage of automatically adapting to the size and shape of the detected objects, thereby optimizing the use of computing resources while maximizing the accuracy of detection and classification.

[0019] In a particularly advantageous embodiment, the light detection system includes a memory classifier configured to integrate temporal information from previous detections. The classifier maintains an internal state that evolves with successive detections, thus enabling it to track objects over time and progressively improve classification accuracy. Preferably, this architecture is based on a short-term memory module that stores features extracted from the last N detections, typically N being between 5 and 10, and a long-term memory module that stores consolidated representations of objects tracked over longer periods. Even more preferably, the classifier includes a recurrent network that integrates this temporal information to produce a robust classification.Preferably, the system also maintains a consolidated detection database, including the trajectories of detected objects, the evolution of their characteristics over time, and successive classifications with their confidence levels. This temporal approach effectively resolves classification ambiguities that can arise from isolated detections, particularly under challenging conditions such as partial occlusions or adverse weather.

[0020] In an advanced embodiment, the light detection system simultaneously implements several of the approaches described above, operating in parallel to maximize detection robustness and accuracy. For example, the parallel architecture includes a processing path for individual photodetectors, a processing path for static groups, a processing path for dynamic groups, and a time-domain processing path. In this example, for the individual photodetector processing path, the computing unit performs correlation calculations and interpolation for each photodetector, detects local maxima to identify point objects, and accurately estimates the distance through fine-grained analysis of the correlation peaks.In the static group processing path, the computing unit analyzes predefined groups of photodetectors corresponding to different regions of the field of view and performs classification based on the aggregated histograms of these groups. In the dynamic group processing path, the computing unit adaptively defines groups of contiguous photodetectors and performs classification based on individual histograms and spatial relationships. In the temporal processing path, the computing unit tracks detected objects over time and performs classification enriched with historical data. Preferably, each processing path produces its own detections and classifications, with associated confidence levels, which are then combined by a fusion module.

[0021] The invention also relates, according to a second aspect, to a front headlight for a motor vehicle comprising a light detection system according to the first aspect.

[0022] The invention also relates, according to a third aspect, to a motor vehicle comprising a light detection system according to the first aspect or at least a front projector according to the second aspect.

[0023] The invention also relates to a detection method for a motor vehicle, comprising the following steps: emission of a light beam whose spectrum includes at least a portion in the visible spectrum, from an emission signal comprising a train of emission pulses, such that the emitted light beam is modulated on the basis of said emission signal and comprises the repetition of elementary pulse trains according to a predetermined emission frequency; reception of a received light beam by at least one photodetector and transformation into a received electrical signal; calculation by a computing unit of correlation values ​​discretely between the received electrical signal and the emission signal for several relative time offset values ​​of these two signals, these time offset values ​​being defined according to a predetermined correlation time step, and comprising a step of processing the correlation values ​​by a computing unit, over at least a range of values,so as to identify specific information from the correlation values, such as one or more maxima, which correspond to a time lag corresponding to the time of flight of an emitted light beam until the reception by at least one photodetector of a received light beam originating at least partially from the reflection of this emitted light beam on an object in the environment, and implementing a calculation of the nature and / or distance of said object positioned in the environment of the light detection system from said processing of the correlation values.

[0024] In said process, the step of processing the correlation values ​​by a computing unit advantageously includes: a mathematical interpolation of the correlation values ​​to generate a continuous correlation function and / or a continuous derivative function of the correlation function, in particular by the least squares method, such that the distance of an object positioned in the environment is calculated by searching for maxima of the continuous correlation function, and / or estimating the distance and / or automatically identifying the nature of an object positioned in the environment of the motor vehicle associated with the correlation values ​​by an embedded artificial intelligence tool that has undergone learning.

[0025] The invention is more particularly defined by the claims.

[0026] These objects, features and advantages of the present invention will be described in detail in the following description of a particular embodiment, given by way of non-limiting example, with reference to the accompanying figures, among which:

[0027] Figure 1 schematically illustrates a light detection system according to one embodiment.

[0028] Figures 2a to 2c schematically represent the obtaining of correlation values ​​by the light detection system according to the embodiment.

[0029] Figure 3 schematically illustrates a processing carried out on correlation values ​​by the light detection system according to the embodiment.

[0030] In the automotive field, it is common practice to use a pulsed light beam emitted by a light module within a vehicle's headlight system to perform a specific photometric function, such as lighting or signaling. Typically, the light source emitting this beam is controlled by a pulse-width modulated (PWM) electrical signal. The light source is periodically switched on and off by this PWM signal, resulting in a beam composed of successive light pulses with a predetermined emission frequency high enough that they are indistinguishable to the human eye. The intensity of the emitted beam is a function of the duty cycle of this PWM signal, allowing it to be controlled by adjusting... This cyclic ratio allows for the realization of a photometric function of lighting or signaling, such as a daytime running light or a crossing type lighting; various functions can thus be implemented by this type of light module.

[0031] According to the embodiment of the invention, such a lighting system is also used to perform a second telemetry function, thus forming a detection system. Indeed, the light source of the lighting module can be controlled so that the pulses of the emitted light beam carry a data sequence. The lighting system is further equipped with a receiver module to receive the emitted light beam after reflection from an object in the vicinity of the vehicle. A computer unit in the motor vehicle can then, after detecting the data sequence in the received light beam, determine the time of flight of the emitted light beam and thus at least estimate the distance between the vehicle and the object. The invention proposes an optimized lighting system using this detection principle.

[0032] Figure 1 shows in detail a light detection system 1 for a motor vehicle according to an embodiment of the invention.

[0033] The light detection system 1 includes an emission module 2 arranged to emit a light beam F1 and a reception module 3 intended to receive a light beam F2, in particular from the reflection of the light beam F1 by an object O, so as to detect such an object O positioned for example at the front of a motor vehicle.

[0034] In this embodiment, the transmission module 2 and the reception module 3 are arranged in the same front headlight of a motor vehicle, the detection lighting system thus being combined with a front headlight intended for the function of lighting and / or signaling at the front of a motor vehicle, which is advantageous in terms of size. The headlight thus combines the two functions of traditional lighting and / or signaling on the one hand, and object detection on the other. Alternatively, the light detection system according to the embodiment of the invention could naturally be separate from the front headlight of a motor vehicle. Furthermore, the transmission module 2 and the reception module 3 could also be arranged in different locations within the motor vehicle.

[0035] According to the embodiment, the emission module 2 comprises a light module 21 and a modulation unit 22. The light module 2 is arranged so that the emitted light beam F1 it emits has an electromagnetic spectrum, at least a portion of which lies within the visible spectrum. For the purpose of fulfilling the rangefinding function of the invention, this visible spectrum may include an intensity peak P1, or line, in the blue at 450 nm. This is advantageous because natural light generally exhibits a minimum of light in this blue spectrum, thus minimizing its impact, which represents noise in the detection system according to the invention. Alternatively, the spectrum may include other intensity peaks in the visible and / or infrared regions.

[0036] In order to emit this light beam F1, the light module 21 comprises a light source 23, capable of emitting light rays, and an optical unit 24 arranged to project these light rays to form the emitted light beam F1. By way of example, the optical unit 24 may indifferently comprise one or more reflectors, one or more lenses, one or more diaphragms or one or more collimators or a combination of several of these optical elements.

[0037] The light source 23 includes, for example, a semiconductor generator (not shown). When electrically powered, this light source 23 simultaneously emits blue and yellow light, the resulting light appearing white to the human eye. Only the blue portion of the spectrum will be used for the rangefinding function according to the embodiment of the invention, the yellow portion complementing the blue portion to contribute to the lighting and / or signaling function.

[0038] The receiving module 3 comprises an optical unit 31, downstream of which are provided a plurality of elementary acquisition modules, which may be photodetectors 32. The receiving module 3 also includes a demodulation unit 33. The light beam F2 received by the receiving module 3 is thus concentrated by the optical unit 31 onto one or more of the photodetectors 32. The photodetectors are preferably identical and each is formed, for example, by at least one single-photon avalanche diode (SPAD) of a silicon photomultiplier tube, and preferably by a group of these photodiodes, for example, a group comprising 15 or more photodiodes. These photodiodes are preferably arranged in a matrix. Thus, the acquisition module can be in the form of a matrix of at least 32 photodetectors, and optionally of fewer than 4000 photodetectors.It should be noted that the dimensions of the photodiodes of the photodetectors 32 are on the order of a micrometer in one embodiment. The assembly forms a sensor whose spatial reception resolution, according to one embodiment of the invention, is on the order of 1°, or even 0.1°, and whose detection capabilities, due to the use of avalanche photodiodes, are particularly high, even under degraded acquisition conditions. The receiving module 3 further includes a filter 31 arranged upstream of the acquisition module, the filter having a bandwidth configured to allow the transmission of the spectrum corresponding to the pulse trains of the emitted light beam, in particular blue light, according to the embodiment.

[0039] The light detection system 1 further comprises a processing unit 4, connected by communication means to the transmitting module 2 and the receiving module 3. This processing unit 4 is configured to implement a method for detecting an object, which can be integrated into a driver assistance system for a motor vehicle, particularly a semi-autonomous or even autonomous one. This method for detecting an object, and therefore the operation of the light detection system, will now be described.

[0040] In the first step, the processing unit 4 generates an initial data sequence, Seq1. This initial sequence, Seq1, is, in the example described, a pseudo-random binary sequence composed of "0"s and "1"s. This sequence, Seq1, can be stored in the memory of the processing unit 4. This processing unit 4 can generate such a sequence periodically and continuously.

[0041] In a second step, the modulation unit 22 modulates the emitted light beam F1 from the light module 21, based on this data sequence Seq1, for example by controlling the power supply to the light source 23. In this embodiment, the modulation unit 22 includes a generator of a pulse-frequency modulated control signal. This control signal allows control of a switched-mode power supply (not shown) to the light source 23. Conventionally, the frequency setpoint of this control signal, determined by the modulation unit 22, thus allows control of the average electrical power supplied to the light source 23, and therefore control of the luminous intensity of the emitted light beam F1, so as to satisfy the requirements of the photometric function it performs.

[0042] More specifically, the modulation unit 22 converts the data sequence Seq1 into an output signal Sseq, composed of pulse trains resulting from the conversion of the data sequence Seq1. The pulses follow one another at a predetermined output frequency, which may or may not be variable, sufficiently high, for example, greater than 10 MHz, particularly between 50 MHz and 100 MHz, and more generally between 10 MHz and 100 MHz. At these frequencies, the human eye can no longer distinguish the pulses. Furthermore, the amplitude, width, and / or position of each pulse, with respect to the period, allows for the characterization of a particular sequence of the output signal Sseq, which is reproduced within the emitted light beam F1. According to the described embodiment, each light pulse corresponds to a bit with the value "1" of the data sequence Seq1, the pulse width corresponding to a chosen pulse duration, which can The intensity of the light beam can be predefined, and its luminous power can correspond to a chosen power level, which can also be predefined. For example, a pulse can have a duration of less than 100 ns, or even 10 ns. The emitted light beam F1 thus comprises a succession of elementary pulse trains, each train corresponding to the emission signal Sseq and resulting from the conversion of the data sequence Seq1.

[0043] The light beam emitted F1 is thus emitted until it reaches an object O, located in the environment of the vehicle, which reflects it towards the receiving module 3. The light beam received F2 by the receiving module 3 is thus composed of a part of the light beam emitted F1 reflected by the object O and noise, for example generated by sources of stray light such as urban lighting, car lighting, or even the sun.

[0044] In a further step, each of the photodetectors 32 of the receiving module 3 converts the portion of the received light beam F2 that it receives into an electrical signal Sel, which it transmits to the demodulation unit 33. The demodulation unit can then extract a demodulated data sequence Seq2 from this signal. The demodulation unit 33 counts, for example, from the electrical signal Sel, the number of photons received by an elementary acquisition module 32 during a predetermined correlation time step. In one example, the correlation time step corresponds to a pulse duration.

[0045] In one example, the demodulation unit determines, by thresholding based on a value determined from the peak power, whether this quantity of photons corresponds to a pulse of the emitted light beam F1, and therefore to a bit with a value of "1" or a bit with a value of "0". If the number of photons detected by the photodetector is less than a certain number of photons corresponding to the threshold, it is very likely that these photons come mainly from ambient light, such as sunlight or light from an artificial light source, such as streetlights. Conversely, if the number of photons detected by the photodetector is greater than a certain number of photons corresponding to the threshold, it is more likely that these detected photons come mainly from the part of the light beam received F2 corresponding to the reflection of the beam emitted F1 from the emission module 2 and reflected by an object O.

[0046] Alternatively, the demodulation unit applies discretization to more than two levels of the received electrical signal Sel, according to a principle analogous to thresholding, which is a discretization to two levels (0 or 1). The Sel signal is thus discretized to obtain a discrete value representing the number of photons received by the transmitting module 2 over the given time interval. Other demodulation alternatives are known from the prior art.

[0047] The electrical signal Sel is thus transformed into a received demodulated data sequence Seq2. It is understood that if the resulting demodulated data sequence Seq2 does indeed originate from the emitted and reflected light beam F1, then it must correspond to the initial data sequence Seq1, shifted in time by a travel time T, which corresponds to the travel time of the beam from its emission by the transmitting module 2 to its reception by the receiving module 3. This travel time allows us, in particular, to deduce the distance of the object O to the vehicle.

[0048] Calculation unit 4 then implements the final part of the calculation, which consists of determining this travel time. To do this, calculation unit 4 first calculates correlation values ​​Vcorr between the received electrical signal Sel and the transmitted signal Sseq, for several values ​​of relative time delay between these two signals. The correlation values ​​are obtained using a cyclic convolution product between these two signals, one of which is delayed according to each of the time delay values, by a chosen correlation time step, for example, 10 ns. As a side note, this choice of time step also corresponds to spatial precision: indeed, for a value of 10 ns, the distance traveled by a light beam is 150 cm, hence such an approach provides a spatial resolution of 150 cm.

[0049] Given the chosen approach, the Vcorr correlation values ​​will be maximum for a time lag value corresponding to the time of flight of the light beam between the moment it is emitted by the module emission time 2 and the instant it is received by the receiving module 3. The computing unit 4 therefore searches for this maximum value among the correlation values ​​Vcorr, to estimate the value T of this time of flight of the light beam emitted F1 between the object O and the vehicle, associated with this maximum value, as illustrated by figures 2a to 2c.

[0050] According to one embodiment, the autocorrelation value Vcorr of a pseudo-random binary sequence is given by the following equation:

[0052] where Vcorr(v) is the autocorrelation value for a time lag v, m1 is the number of bits with a value of "1" in the data sequence of the signals and N is the total number of bits in the sequence corresponding to the compared signals.

[0053] Preferably, correlation values ​​can be obtained by convolving the two signals being compared. Naturally, the invention is not limited to a particular method of calculating correlation values, and any other calculation approach is conceivable. In all cases, correlation values ​​are obtained discretely, according to a chosen correlation time step.

[0054] Figures 2a to 2c summarize this method for calculating correlation values. Figure 2a represents the emitted signal, which comprises a pulse train emitted as previously explained, generated by the emitting device 2 and contained within the emitted light beam F1. Figure 2b represents the received electrical signal Sel, received by the receiving device 3, and contained within the received light beam F2. This received electrical signal Sel incorporates a received pulse train corresponding to the transmitted one, with a time shift T, which corresponds to the time of flight of the light beam reflected from the object O, as previously explained. Furthermore, it should be noted that the received pulse train exhibits certain differences relative to the emitted pulse train, due to noise generated by ambient light.

[0055] The correlation function is implemented for different offset values ​​and must allow a peak to be reached at the time offset T, as shown in Figure 2c, thus identifying the flight time. As a note, Figure 2c represents a continuous correlation function, as constructed by the method according to the embodiment of the invention.

[0056] As illustrated in Figure 3, the process first calculates correlation values ​​Vcorr, according to a correlation time step. The correlation values ​​are thus distributed discretely. The time of flight most often corresponds to a value T positioned between two of these values, and it is not easy to deduce this value T precisely because, in many situations, the sequence of correlation values ​​has a complex form for which it is not immediately possible to accurately determine, for example, the location of one or more peaks. This accuracy is therefore bounded by the correlation time step, which also indirectly defines the spatial accuracy, as mentioned previously.

[0057] Finally, due to noise, the received light beam contains transformed pulse trains, which generates errors in recognition. Furthermore, depending on the complexity of the objects present in the environment, the correlation values ​​obtained may also exhibit a complex distribution that is not easily usable. The resulting correlation function therefore does not always show a clear and identifiable peak. Moreover, when multiple objects are detected, the function must include several peaks, and it is not always easy to identify these different peaks accurately and reliably.

[0058] To achieve improved object detection accuracy, a natural approach might be to create very short pulse trains to increase their frequency and the amount of usable data. Then, it might be advantageous to choose a very small correlation time step to increase the number of correlation values ​​and thus improve accuracy. However, such an approach faces difficulties. A primary physical limitation is the dead time of the Photodetectors. Indeed, each photodetector has a dead time during which it cannot detect a second photon after detecting a first photon: this necessitates choosing pulse trains whose width is at least equal to this dead time, or even a multiple of it. This therefore limits the repetition of processing calculations to a minimum of this dead time. A second physical limitation arises from the fact that such an approach greatly increases the amount of data to be processed, which would require increasing the computing power available to process this data in real time, up to a threshold that is no longer realistic.

[0059] To improve the accuracy of the detection process under these conditions, the processing unit implements a step to calculate a correlation function from the obtained discrete values. This allows for the identification of the flight time with very high precision, well beyond the correlation time step. In other words, the process reconstructs missing information from the data obtained by the previous calculation, without increasing the amount of data required, and therefore without burdening the light detection system.

[0060] More specifically, the computing unit implements a processing of the correlation values ​​Vcorr so as to generate a continuous correlation function and / or a continuous derivative function of a correlation function, over at least a range of values, so as to accurately identify portions of the correlation function, and in particular one or more maxima, which correspond to a time lag corresponding itself to a time of flight of an emitted light beam F1 until its detection within a received light beam F2 after its reflection on an object in the environment by at least one photodetector, in order to deduce the nature and / or the distance of said object positioned in the environment of the light detection system 1. By the nature of an object, we mean the recognition of the object, among a category such as a post, a wall, a vehicle, a bicycle, etc.

[0061] If we are initially interested in estimating the distance between an object and the light detection system, two complementary approaches are chosen.

[0062] In the case where at least one photodetector 32 of the receiving module 3 is sized to receive a received light beam F2 originating from a small area of ​​the environment, then said area is considered, for simplicity, as a point. In the embodiment described above, each photodetector perceives a spatial resolution angle of reception on the order of 0.1 degrees, which can be considered as the perception of a small area, insofar as this angular resolution corresponds to an object less than 9 cm in diameter located at a distance of 50 m.

[0063] In such a case of a point object, the processing of the correlation values ​​Vcorr according to the embodiment of the invention consists of a mathematical interpolation of the correlation values ​​to generate a continuous correlation function and / or a continuous derivative function, in particular by the least squares method, so that the distance of said surface is calculated by finding the maxima of the continuous correlation function.

[0064] According to the embodiment, this mathematical interpolation considers a predefined mathematical function, such as a cardinal sine, approximated by harmonics based on the predetermined emission frequency, associated with coefficients in the form of third-degree polynomials. This approach makes it possible to take into account the physical reality of the frequencies involved in the signal, as well as the shape observed empirically by classical correlation functions, to reconstruct an optimal correlation function in all situations. On this basis, it is possible to deduce from this continuous function, reconstructed from a few discrete values, the precise distances of objects in the environment. Advantageously, this interpolation is performed on the derivative of the correlation function. From the correlation values ​​obtained, an estimate of the calculated value of the derivative is obtained by the The subtraction of two consecutive values, divided by the time step of the correlation calculation. These calculated derivative values ​​are used to mathematically construct the continuous curve that best approximates this derivative function. In all cases, the most relevant continuous correlation function possible is obtained by this first approach to the invention. Such a function Fcorr is, for example, represented by Figures 2c and 3.

[0065] In other situations or embodiments, the assumption of a point object is no longer applicable. This might be the case, for example, if a smaller number of photodetectors are chosen in the receiving module, so that each photoreceptor perceives light from larger areas of the environment. For instance, an angular resolution of 2° can be considered as resulting in the perception of large areas, since this resolution corresponds to an object larger than 1.7 m located at a distance of 50 m. In such a situation, even though the previous approach would still be applicable, a second approach is considered, based on the use of artificial intelligence principles.

[0066] In addition, if it is desired to obtain more complete information about the environment, which goes beyond just the distance of an object, the same second approach based on the use of artificial intelligence principles is also possible.

[0067] In this second approach, a machine incorporates a neural network to form an embedded artificial intelligence tool. Such a machine first undergoes a learning phase based on the well-known principles of "deep learning." To achieve this, it is placed in a multitude of situations where the environment includes one or more common objects, such as a lamppost, a car, a pedestrian, a bicycle, etc.

[0068] The two approaches are advantageously combined in a single light detection system, the first approach being preferred for objects considered as points and / or when it is the only estimation of the The first approach is based on distance, and the second is preferred for objects of more complex shapes and / or when estimating the precise nature of the detected object. Naturally, either approach remains suitable for providing relevant information in all situations.

[0069] In particular, some sensors can combine large and small photodetectors, thus forming acquisition modules with a plurality of angular resolutions. It is therefore understandable that it is possible to apply both approaches depending on the characteristics of individual pixels, or on the characteristics of groups of several pixels.

[0070] Ultimately, the invention offers the advantage of significantly improving detection accuracy without increasing the amount of data to be processed, thus maintaining the same correlation time step. There is also no need to increase the overall hardware, computing power, the number of photodetectors, or the emission frequency.

[0071] According to one embodiment, the light detection system 1 is used in addition to a conventional LIDAR type detection system to create redundancy between the signals received by the conventional detection system and those received by the light detection system in order to make the data extracted from said signals more reliable.

[0072] The invention also relates to a front headlight for a motor vehicle, which performs the two complementary functions of conventional lighting and detection according to the invention. In other words, the headlight integrates the detection lighting system according to the invention. Advantageously, the same optical components allow both functions to be performed, thus enabling the invention to be implemented without inconvenience.

[0073] The invention also relates to a motor vehicle as such, which includes a light detection system according to the invention, integrated into a projector or positioned according to any other arrangement.

[0074] Finally, the invention also relates to a detection method for a motor vehicle. The light detection system according to the invention implements such a method.

[0075] This method comprises the following steps: emission of a light beam whose spectrum includes at least a portion in the visible spectrum, from an emission signal Sseq comprising a train of emission pulses, such that the emitted light beam F1 is modulated on the basis of said emission signal Sseq, comprising the repetition of elementary pulse trains according to a predetermined emission frequency; reception of a received light beam F2 by at least one photodetector 32 and transformation into a received electrical signal Sel; calculation by a computing unit 4 of discrete correlation values ​​Vcorr between the received electrical signal (Sel) and the emission signal Sseq for several relative time offset values ​​of these two signals, these time offset values ​​being defined according to a predetermined correlation time step; processing of the correlation values ​​Vcorr by a computing unit 4,so as to generate a continuous correlation function Fcorr and / or a continuous derivative function of a correlation function Fcorr, over at least a range of values, so as to accurately identify portions of the correlation function from the discrete correlation values ​​calculated in the previous step, such as one or more maxima, which correspond to a time lag corresponding itself to a time of flight of an emitted light beam F1 until the reception by at least one photodetector of a received light beam F2 originating at least partially from its reflection on an object in the environment, and calculation of the nature and / or distance of said object positioned in the environment of the light detection system 1 from said processing of the correlation values ​​Vcorr.,

[0076] The invention also relates to a driving assistance method for a motor vehicle, which includes an object detection phase. positioned within the environment of the motor vehicle, implementing the detection method according to the invention. In other words, a motor vehicle may include a central unit that participates in autonomous driving or assists in driving the motor vehicle. This central unit may integrate the processing unit of the light detection system, or be connected to such a separate processing unit by means of communication.

[0077] The light detection system therefore advantageously includes software and hardware means configured to implement these detection and assistance processes, notably through its computing unit.

[0078] Ultimately, the invention demonstrates that a motor vehicle is automatically informed, in real time, of its distance from an object, particularly an object positioned in front of the vehicle that could potentially obstruct the vehicle's future movement. This information can be automatically used by the motor vehicle to modify or adapt its trajectory accordingly. Alternatively, the motor vehicle may include a human-machine interface that allows it to inform the driver of the distance and / or nature of the object in the environment.

Claims

DEMANDS 1. Light detection system (1) for a motor vehicle comprising: a. an emission module (2) having a light module (21) capable of emitting a light beam whose spectrum has at least a portion in the visible spectrum, the light beam contributing to a lighting or signaling function, and a modulation unit (22) arranged to modulate said light beam emitted by the light module (21) from an emission signal (Sseq) comprising a train of emission pulses, so as to emit a modulated emitted light beam (F1) comprising the repetition of elementary pulse trains according to a predetermined emission frequency; b.a receiving module (3) capable of receiving a received light beam (F2), the received beam originating at least in part from a reflection of the emitted light beam (F1) on an object (O) in the environment, comprising an acquisition module comprising at least one photodetector (32) capable of converting a received light signal into a received electrical signal (Sel); c.a computing unit (4) configured to calculate correlation values ​​(Vcorr) discretely between the received electrical signal (Sel) and the emitted signal (Sseq) for several relative time offset values ​​of these two signals, these time offset values ​​being defined according to a predetermined correlation time step, characterized in that the computing unit (4) is configured to implement a processing of the correlation values ​​(Vcorr), so as to identify specific information from the correlation values ​​(Vcorr), such as one or more maxima, which correspond to a time offset corresponding to a time of flight of an emitted light beam (F1) until the reception by at least one photodetector of the part of the received light beam (F2) reflected on the object (O), in order to deduce the nature and / or distance of said object positioned in the environment of the light detection system (1).

2. A light detection system (1) for a motor vehicle according to the preceding claim, characterized in that the processing of the values ​​of correlation (Vcorr) implemented by the computing unit (4) consists of a mathematical interpolation of the correlation values ​​to generate a continuous correlation function (Fcorr) and / or a continuous derivative function of the correlation function (Fcorr), notably by the least squares method, so that the distance of an object positioned in the environment is calculated by finding the maxima of the continuous correlation function (Fcorr).

3. Light detection system (1) for a motor vehicle according to the preceding claim, characterized in that said mathematical interpolation is based on the least squares method, considers a predefined mathematical function, such as a cardinal sine, approximated by harmonics based on the predetermined emission frequency, associated with coefficients in the form of polynomials, in particular third-degree polynomials.

4. Light detection system (1) for a motor vehicle according to claim 1, characterized in that the processing of correlation values ​​(Vcorr) implemented by the computing unit (4) consists of an estimation of the distance and / or an automatic identification of the nature of an object positioned in the environment associated with the correlation values ​​by an embedded artificial intelligence tool which has undergone learning.

5. Light detection system (1) for a motor vehicle according to claim 1, characterized in that at least one photodetector (32) of the acquisition module is sized to receive a received light beam (F2) from a surface of the environment of dimension less than a predefined threshold, so that said surface is considered in a simplified way as a point and in that the processing of the correlation values ​​(Vcorr) implemented by the computing unit (4) consists of a mathematical interpolation of the correlation values ​​to generate a continuous correlation function and / or a continuous derivative function of the correlation function.

6. Light detection system (1) for a motor vehicle according to claim 1, characterized in that at least one photodetector (32) of the acquisition module is sized to receive a received light beam (F2) from a surface of the environment with a dimension greater than a predefined threshold, so that said surface is considered three-dimensional, and in that the processing of the correlation values ​​(Vcorr) implemented by the computing unit (4) consists of an estimation of the distance and / or an automatic identification of the nature of an object positioned in the environment associated with the correlation values ​​by an embedded artificial intelligence tool that has undergone learning.

7. Light detection system (1) for a motor vehicle according to any one of the preceding claims, characterized in that the predetermined emission frequency is between 10 MHz and 100 MHz.

8. Light detection system (1) for a motor vehicle according to any one of the preceding claims, characterized in that the predetermined correlation time step is greater than or equal to 500ps, or greater than or equal to 1 ns, and optionally less than or equal to 4ns.

9. Light detection system (1) for a motor vehicle according to any one of the preceding claims, characterized in that the acquisition module of the receiving module (3) comprises an array of at least 32 photodetectors, and optionally of less than 4000 photodetectors.

10. Light detection system (1) for a motor vehicle according to any one of the preceding claims, characterized in that the emission module (2) emits a beam in a visible spectrum comprising a first part, in particular in the blue spectrum, comprising said emitted light beam (F1) modulated for the detection of at least one object in the environment, and comprising a second part intended to complement the first part so that the beam contributes to the function of lighting or signaling, such as a daytime running light, a high beam or a low beam.

11. Front headlight for a motor vehicle, characterized in that it comprises a light detection system (1) according to one of the preceding claims.

12. A detection method (1) for a motor vehicle, comprising the following steps: - emission of a light beam whose spectrum has at least a portion in the visible spectrum, from an emission signal (Sseq) comprising a train of emission pulses, so that the emitted light beam (F1) is modulated on the basis of said emission signal (Sseq) and includes the repetition of elementary pulse trains according to a predetermined emission frequency; - reception of a light beam received (F2) by at least one photodetector (32) and transformation into a received electrical signal (Sel); - calculation by a computing unit (4) of correlation values ​​(Vcorr) in a discrete manner between the received electrical signal (Sel) and the emitted signal (Sseq) for several relative time offset values ​​of these two signals, these time offset values ​​being defined according to a predetermined correlation time step, and characterized in that it includes a step of processing the correlation values ​​(Vcorr) by a computing unit (4), over at least a range of values, so as to identify specific information from the correlation values ​​(Vcorr), such as one or more maxima, which correspond to a time offset corresponding to a time of flight of an emitted light beam (F1) until the reception by at least one photodetector of a received light beam (F2) originating at least partially from the reflection of this emitted light beam (F1) on an object in the environment,and in that it implements a calculation of the nature and / or distance of said object positioned in the environment of the light detection system (1) from said processing of correlation values ​​(Vcorr).

13. A detection method (1) for a motor vehicle according to any one of the preceding claims, characterized in that the step of processing the correlation values ​​(Vcorr) by a computing unit (4) comprises: - a mathematical interpolation of the correlation values ​​to generate a continuous correlation function (Fcorr) and / or a continuous derivative function of the correlation function (Fcorr), in particular by the least squares method, so that the distance of an object positioned in the environment is calculated by searching for the maxima of the continuous correlation function (Fcorr), and / or - an estimation of the distance and / or an automatic identification of the nature of an object positioned in the environment of the motor vehicle associated with the correlation values ​​by an embedded artificial intelligence tool that has undergone learning.

Images