Real-time noise detection method and system for a photon counting pixel array

Abstract

A single photon counting sensor array includes one or more emitters configured to emit a plurality of energy pulses and a detector array including a plurality of pixels. Each pixel includes one or more detectors, the plurality of one or more detectors configured to receive a reflected energy pulse emitted by the one or more emitters. A mask material is positioned to cover a portion of the detectors of the plurality of pixels rather than all of the detectors to produce blocked pixels and unblocked pixels such that each blocked pixel is prevented from detecting a reflected energy pulse so that each blocked pixel only detects intrinsic noise.

Description

[0001] Cross-references and priority requirements

[0002] This patent document claims priority to U.S. Patent Application No. 17 / 062,856, filed on October 5, 2020, the entire contents of which are incorporated herein by reference. Background Technology

[0003] The recent development of autonomy in transportation, including both fully automated vehicles (AVs) and advanced driver assistance systems (ADAS), has led to an increased demand for high-speed three-dimensional (3D) imaging technologies capable of detecting the position and details of objects in highly dynamic environments. Light detection and ranging (LiDAR) systems have become an extremely important imaging technology for autonomous vehicle applications because they offer the greatest range and highest resolution for 3D imaging compared to other methods such as radar and ultrasound. A LiDAR system is a sensor that emits light into its surroundings and detects reflected light from objects such as landscapes, pedestrians, structures, and vehicles (i.e., moving actors or stationary objects).

[0004] One method for determining the location of such objects is through time-of-flight (TOF), where a light pulse is emitted from a sensor, and the distance to the target is determined by the round-trip time of the reflected pulse, since the speed of light is constant. This time-based data is collected and used to generate a LiDAR spatial point cloud, a three-dimensional (3D) representation of the surrounding environment in space, depicted as discrete points along the vertical, horizontal, and longitudinal axes. For each point generated in the point cloud, a histogram of the collected data is created, where the detector samples all responses that occur during the time interval from pulse emission until the time corresponding to the maximum measurable distance set by system parameters. The histogram is formed by sampling the intensity of the returned pulse and recording the time it takes for the reflected pulse to return to the detector.

[0005] The job of a LiDAR system analyzer is to examine the time-based histogram and identify which intensity peaks originate from the real target—that is, the signal—and which originate from noise sources. Not all light received by a LiDAR system is reflected light originally generated by the system. A wide variety of noise sources can interfere with a LiDAR system, including external noise sources (such as the solar background and other light sources) as well as inherent noise generated within the LiDAR system itself. Inherent noise typically includes any signal not caused by direct photon reception (avalanche counting). Example types of inherent noise include dark counting noise and crosstalk.

[0006] Recent improvements to LiDAR systems utilizing Geiger-mode avalanche photodiodes (GmAPDs) allow for single-photon counting and provide increased light detection sensitivity compared to conventional LiDAR systems using linear-mode avalanche photodiodes (APDs) that require multiple photons for measurable responses. Accurate noise characterization is even more critical in LiDAR systems using Geiger-mode photodiode detectors, as higher sensitivity inherently comes with higher noise components from the surrounding environment and the system itself. Accurately determining the noise components in the time-based intensity histogram is essential for achieving high signal-to-noise ratios (SNR) or for tolerating lower SNRs, and for accurately creating 3D point clouds in the space surrounding the LiDAR system. In the case of autonomous vehicles, this noise characterization must be analyzed at high speeds because the LiDAR system moves in a highly dynamic and constantly changing environment.

[0007] The inherent noise of Geiger-mode detectors includes noise sources such as dark count rate, afterpulse, and early fire, which can occur within each detector. Furthermore, because Geiger-mode LiDAR systems deploy high-density arrays of closely spaced pixels to achieve better spatial resolution, characterizing the inherent noise interference (i.e., crosstalk) between adjacent pixels is crucial for accurate analysis and achieving higher signal-to-noise ratios for wider-area detection at higher spatial resolution.

[0008] This paper describes methods and systems for real-time in-situ characterization of the inherent noise sources of Geiger-mode avalanche photodiode pixels arranged in high-density pixel arrays and / or other related problems. Summary of the Invention

[0009] A single-photon counting sensor array includes: one or more transmitters configured to emit a plurality of energy pulses; and a detector array comprising a plurality of pixels. Each pixel includes one or more detectors, the plurality of said one or more detectors being configured to receive reflected energy pulses emitted by the one or more transmitters. A masking material is positioned to cover only a portion, rather than all, of the detectors of the plurality of pixels to create blocked and unblocked pixels, such that each blocked pixel is prevented from detecting the reflected energy pulses.

[0010] A system that can operate with a sensor array includes a processor and programming instructions to receive characteristic data of signals received by blocked pixels and unblocked pixels, and to compare the characteristic data of signals received by blocked pixels with the characteristic data of signals received by unblocked pixels to determine a measurement of inherent noise in the detector array.

[0011] In various embodiments, the single-photon counting sensor array can be an element of the optical detection and ranging (LiDAR) system in the system, wherein the distance between the center points of adjacent pixels in the detector array is less than the crosstalk length.

[0012] In various embodiments, each detector may include a photosensor having: a surface region of p-type semiconductor material positioned to receive light; a drift region of n-type semiconductor material; and a conductive trace connected to the surface region and positioned to serve as the anode of the photosensor. The detector array may include a substrate on which photosensors are positioned. The substrate may be configured to serve as the cathode of each photosensor in the photosensor array. Each blocked pixel may have a mask material positioned above its surface region to block light from entering the blocked pixel.

[0013] In various embodiments, the detector array may include a substrate configured to serve as the cathode of each photodetector in the photodetector array. Each photodetector may include a metal window extending through the substrate and configured to receive light entering the photodetector. An n-type semiconductor material region may be connected to the metal window of each photodetector. A p-type semiconductor material region may be connected to the n-type semiconductor material region of each photodetector. Conductive traces may be connected to the p-type semiconductor material region of each photodetector and positioned to serve as the anode of that photodetector. A mask material may be positioned to cover the metal window of each blocked pixel.

[0014] In various embodiments, when comparing characteristic data of signals received by blocked pixels with characteristic data of signals received by unblocked pixels to determine a measurement of inherent noise in the detector array, the system may: (i) identify a superpixel comprising a group of pixels in the detector array; (ii) determine the total photon count rate received by the superpixel; (iii) determine the avalanche count rate received by the unblocked pixels of the superpixel; and (iv) determine the noise measurement as a function of the total photon count rate and the avalanche count rate received by the unblocked pixels of the superpixel. Optionally, the function may be:

[0015]

[0016] in:

[0017] λ = avalanche count rate received by the superpixel;

[0018] = The photon count rate generated by reflected signal photons at photoelectric sensor i;

[0019] = Photon count rate generated by background signal photons (i.e., at detector i);

[0020] = avalanche count rate of the inherent noise of detector i; and

[0021] β = Photon detection efficiency.

[0022] In various embodiments, the system may include a data logger configured to receive signals from pixels and store characteristic data corresponding to the received signals.

[0023] Alternatively, the mask material can be formed of metal.

[0024] In various embodiments, the system can be configured to measure the health of the detector array by monitoring spatial variations, temporal variations, or both spatial and temporal variations of noise from the blocked detector. Alternatively, the system can be configured to apply a characterization function to estimate the amount of measurement bias caused by crosstalk in the intensity estimation. Attached Figure Description

[0025] Figure 1 An exemplary component of a LiDAR system is shown.

[0026] Figure 2 This is an exemplary detector array for use in a LiDAR system, in which light is received through the front (anode) side of the detector array.

[0027] Figure 3 It shows that it can be used in, for example Figure 2 Exemplary masked pixels and exemplary unmasked pixels used in the detector array.

[0028] Figure 4 This is an exemplary detector array used in a LiDAR system, in which light is received through the rear (cathode) side of the detector array.

[0029] Figure 5 It shows that it can be used in, for example Figure 4 Exemplary masked pixels and exemplary unmasked pixels used in the detector array.

[0030] Figure 6 This is a flowchart illustrating the process of determining the inherent noise in the detector array.

[0031] Figure 7 Further demonstrated in Figure 6 Certain signal processing steps are performed during the noise estimation and signal processing phases.

[0032] Figure 8It is a block diagram showing the various elements of a possible electronic subsystem for AV and / or external electronic devices. Detailed Implementation

[0033] Unless the context clearly specifies otherwise, the singular forms “a,” “an,” and “the” used herein include plural references. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. As used herein, the term “comprising” means “including, but not limited to.” Definitions of other terms relevant to this document are included at the end of this specific embodiment.

[0034] Figure 1 An example LiDAR system 101 that can be used in various embodiments is shown. For example... Figure 1 As shown, the LiDAR system 101 includes a housing 105 that can rotate 360° about a central axis (such as a hub or axis 118). The housing may include a transmitter / receiver aperture 111 made of a light-transmitting material. Although... Figure 1 The example shown has a single aperture, but in various embodiments, multiple apertures for emitting and / or receiving light may be provided. In either case, as the housing 105 rotates, the system is able to emit light through one or more apertures 111 and receive reflected light reflected back toward one or more apertures 111. In alternative embodiments, the outer shell of the housing 105 may be a fixed dome, at least partially made of a light-transmitting material, with rotatable components inside the housing 105.

[0035] Inside the rotating housing or fixed dome is a light emitter system 104, configured and positioned to generate and emit light pulses via one or more laser emitter chips, an on-chip emitter array, or other light-emitting devices, and through aperture 111 or through the transparent dome of housing 105. Emitter system 104 may include any number of individual emitters, including, but not limited to, eight, 64, or 128 emitters. The emitters may emit light of substantially the same intensity or varying in intensity. The LiDAR system also includes a light detector 108 comprising a photodetector or photodetector array positioned and configured to receive light reflected back into the system. Emitter system 104 and detector 108 will rotate with the rotating housing, or they may rotate within the fixed dome of housing 105. One or more optical element structures 109 may be positioned in front of light-emitting unit 104 and / or detector 108 as one or more lenses or waveplates to focus and guide light passing through optical element structures 109.

[0036] The LiDAR system will include a power unit 121 that provides power to the laser emitter unit 104, motors, and electronic components. The LiDAR system will also include an analyzer 115 with elements such as a processor 122 and a non-transitory computer-readable storage 123 containing programming instructions configured to enable the system to receive data collected by the photodetector unit, analyze it to measure the characteristics of the received light, and generate information that connected systems can use to make decisions about operating in the environment where the data was collected. Optionally, the analyzer 115 may be integrated with the LiDAR system 101 as shown, or it may be partially or entirely located external to the LiDAR system and communicatively connected to the LiDAR system via a wired or wireless communication network or link.

[0037] This paper describes noise compensation methods and systems that can be used with single-photon counting pixel arrays. Single-photon counting pixel arrays can be used in LiDAR systems such as those described above, particularly in LiDAR systems using GmAPDs as sensors. However, the methods and systems described below are not limited to LiDAR systems and can also be used in X-ray detectors and other sensing systems that use single-photon counting pixel arrays and other types of pixel arrays as detectors.

[0038] As described in the background section above, inherent noise generated within a LiDAR system can lead to counts of light pulses that are not derived from reflections from the intended target. Inherent noise can be caused by a variety of factors, including material quality and crosstalk between pixels. It can also vary with temperature, humidity, or other environmental conditions in the area where the system is used. Accurate, real-time noise characterization can help distinguish the true signal count rate from the noise count rate, thereby improving system performance.

[0039] For example, the dark count rate (DCR) of a Geiger-mode detection system is the average rate of counts registered (i.e., avalanche events) when there is no incident light on the pixel, and therefore represents the inherent noise component. The DCR of a detection system is a function of the bias voltage applied to the device, where the higher the applied bias voltage exceeds the breakdown voltage (Vbr), the higher the DCR. Furthermore, for an array of isolated pixels extending over a fairly large spatial area, there will be variations in both Vbr and DCR, and these variations will contribute to noise variations throughout the pixel array. The sensitivity of a single-photon detector is a function of the DCR of its pixels because this determines how much "over-bias" can be applied to the detector while still maintaining a usable signal-to-noise ratio.

[0040] The method described below provides a way to detect and optionally compensate for inherent noise in isolated pixels in real time. In this way, the system's sensitivity can be improved, thereby enhancing performance, either without further increasing the bias voltage or by increasing the signal-to-noise ratio at a higher bias voltage.

[0041] Figure 2 This is a front view of an exemplary front-side detector array 200, which includes N x 2 pixels for a LiDAR system. (In this context, "front" side is the side of the detector array 200 that includes the anode. In this embodiment, the anode is positioned outward from the LiDAR system so that it receives reflected light.) Each pixel 201 includes an active region that serves as a photodetector, and each pixel is connected to an anode contact via a conductive (typically metallic) trace 203. Most pixels are unmasked, meaning their photodetectors are exposed and receive reflected energy, as would be expected in such an array. However, a subset of the pixels and their conductive traces are covered by a mask that prevents the photodetectors from receiving reflected energy. Figure 2 Several exemplary unmasked pixels are designated as 201a, and the masked pixels (also known as blocked pixels) are designated as 201b. Exemplary unmasked traces are also designated as 203a, and the masked traces as 203b. The mask can be formed of metal, such as typical metal stacks like Ti / Pt / Au, Ni, Cr, W, or any material that effectively blocks light of the wavelength of interest. In a LiDAR system, the wavelength of interest can be in the ultraviolet (UV) region (i.e., 180-480 nm), and optionally in the visible light region (i.e., 400-700 nm), and optionally in the near-infrared (IR) region (i.e., 700-2000 nm). Figure 2 The spatial distribution of the blocked detectors shown is just an example; other arrangements are possible.

[0042] Each mask will cover the entire active (photosensitive) area of ​​a pixel, and each mask can be formed during the wafer fabrication process. Masks can also be formed simultaneously with metal traces. Masks can be applied to various locations in an array of any arrangement, such as every nth pixel in a row or column, or even in a random arrangement.

[0043] In such Figure 2 In the arrangement shown, the masked pixel is biased equally with its neighboring pixels but cannot receive direct optical input from external light. Therefore, any counts generated by any masked pixel must originate solely from inherent noise, such as material defects and crosstalk from neighboring pixels. This noise measurement of any active pixel can be used to adjust the value of the signal received from neighboring pixels, or the value of all received signals, to provide real-time noise suppression.

[0044] Figure 3 The image shows elements of unmasked pixels 201a and masked pixels 201b on substrate 311 of the detector array. A surface region 312a of p-type semiconductor material is positioned on top of the photodetector, forming a pn junction maintained at a reverse bias. In a Geiger-mode avalanche photodiode, the reverse bias is maintained above the breakdown voltage Vbr. A drift layer 313a, located directly below the p-type region 312a, extends several micrometers and acts as a multiplication region amplifying excited carriers from the absorber region. Substrate 311 can be formed of n-type semiconductor material and serves as the n+ cathode for each pixel. Geiger-mode devices can also use a p-type substrate and a drift region with an n+ region formed on top, formed with opposite polarities.

[0045] Referring to the unmasked pixel 201a, the incident photon 351a passes through the top of the pixel, the p-region, and the drift layer until it reaches the absorber 314a, which is a semiconductor material with a bandgap energy lower than the incident photon energy (such as InGaAs used for near-infrared photons). In the absorber, the photon generates excitons 316a (i.e., electron-hole pairs). The generated charge carriers are accelerated in the opposite direction by an applied electric field, and one of the charge carriers 316a (in this case, a hole) drifts to the multiplication region 313a. Once the charge carriers reach the multiplication region, they induce electron avalanches through shock ionization and are amplified to produce a measurable current measured via a metal trace 317a, which leads from the surface region 312a of the p-type semiconductor material to a readout integrated circuit (ROIC) (not shown).

[0046] for Figure 3 In the polarity case, contact trace 317a serves as the anode of the detector and can be made of metal or any suitable conductive material typical for semiconductor wafer processes. The entire surface of the pixel array and substrate 311 is coated with an anti-reflective dielectric coating 319a, which also serves as an electrically insulating passivation layer to prevent current from leaving the detector at any point other than where the contact trace 317a meets the semiconductor material in the p-region 312a. The anti-reflective coating can be formed using processes such as depositing a thin oxide or nitride layer on the surface of a photodiode, processes suitable for passivating the pixel surface under applied reverse bias and preventing any significant current leakage from the anode to the cathode.

[0047] Referring to the unmasked pixel 201a, the fundamental source of inherent noise arises from shock ionization and avalanche processes. This process also induces emission 315 at higher bandgap energies in the drift layer material (e.g., Eg = 1.33 eV for InP in the near-infrared state). These higher-energy photons are scattered in all directions and can propagate to the absorber layer 314b of adjacent pixels, generating unwanted excitons 316b, and unwanted counts, i.e., noise, in these pixels. This phenomenon, in the form of crosstalk, is a fundamental component of the inherent noise generated in Geiger-mode photodetectors and limits how close the pixel pitch can be as pixels get closer by reducing the signal-to-noise ratio of any particular pixel. Furthermore, drift layer emission generated by avalanche events can induce cascading effects, as unwanted excitons 316b also generate avalanches in adjacent pixels, resulting in more emission events. Such cascading inherent noise sources, if strong enough, can propagate throughout the entire closely spaced pixel array.

[0048] Similar to the unmasked pixel 201a, the masked pixel 201b includes a surface region 312b of p-type semiconductor material, a drift region 313b of n-type semiconductor material, an anti-reflective coating 319b, and a conductive contact trace 317b leading to the ROIC (not shown). However, the exposed surface of pixel 201b (including surface region 312b and conductive contact trace 317b) is coated with a mask 320, which blocks light 351b from reaching the absorber region 314b of the photodetector. Incident photons are absorbed and reflected in the mask material. Since no light reaches the active region, the only signal emitted through the conductive trace 317b of the masked pixel 201b will be a signal generated from inherent noise (including material defects and crosstalk).

[0049] Figure 4 An example of a rear-side detector array 400 for a LiDAR system is shown, comprising N x N pixels. (In this context, the "rear" side is the side of the detector array 400 that includes a substrate 411 and serves as the cathode on which the photodetector is fabricated. In this embodiment, the cathode is positioned outward from the LiDAR system so that it receives reflected light.) Each pixel 401 includes an active region serving as a photodetector, and each pixel is connected to the cathode via a metal layer with an outwardly extending "window" 403 to receive and detect light. Figure 2 The embodiments are the same, in Figure 4 In this embodiment, a subset of the light-receiving area of ​​the pixel will be covered by a light-blocking mask 420 made of a material such as described above.

[0050] Figure 5 It shows in such as Figure 4The rear-side detector array shown has several unmasked pixels 501a and masked pixels 501b on its substrate 511. The materials used in this embodiment may include those described above. Figure 3 Any of those described in the context. In this variant, all pixels can have the same structure, with a photodetector having a surface region 512 of p-type semiconductor material positioned at the bottom of the photodetector to serve as the anode, and a region 513 of n-type semiconductor material positioned between the p-type semiconductor 512 and the N+ substrate 511 to provide a pn junction held under reverse bias. Current is supplied by the surface region 452 of the p-type semiconductor material of the photodetector via a conductive contact trace 517, which leads from the p-type semiconductor material region 512 to the ROIC 533. The entire photodetector and device side of the substrate 511 surface is coated with an anti-reflective dielectric coating 519, which covers all areas except where the trace 517 contacts the p-region 512, as described above for... Figure 3 As described above, this is for passivation purposes. A mask material 520 can be coated on the substrate 511 to block the light-receiving area of ​​the masked pixel 501b. The mask 520 can be a back metal or a cathode. Since no light reaches the active area, the only signal emitted via the conductive traces of the masked pixel 501b will be a signal generated from inherent noise such as DCR and crosstalk as described above. Figure 3 As described, the avalanche event induces luminescence 515 at higher bandgap energies in the drift layer material (e.g., Eg = 1.33 eV for indium phosphide (InP) in the near-infrared state). These higher-energy photons are scattered in all directions and are able to propagate to the absorber layer 414 of adjacent pixels, generating unwanted excitons 516 and unwanted counts, i.e., noise, in these pixels.

[0051] Partially masked photon-counting pixel arrays, as described above, can be used to determine the inherent noise in the array at any given time point. The signal generated by an unmasked (unblocked) detector is a composite signal associated with multiple sources, including reflected pulse signals, solar noise, and inherent detector noise. The signal generated by a blocked (masked) detector, if not entirely, should be essentially a result of inherent noise. The spatial characteristics of this signal (i.e., the location of the noise in the array) can be used to better filter local signals, improve the local signal-to-noise ratio, and increase detector sensitivity. The temporal characteristics of this signal (i.e., the time it is detected) can be used to decouple material-driven DCRs from crosstalk, as DCRs typically have non-temporally statistical characteristics, while crosstalk may have non-stationary temporally characteristics. In-situ noise monitors allow for close pixel spacing to improve the spatial resolution of LiDAR systems and can be used to address effects such as blooming or cascading crosstalk from strong reflections from highly reflective objects, such as retroreflectors used in road signs.

[0052] To this end, the system can measure the noise of individual pixels, as shown above. Alternatively, it can identify clusters of any number of neighboring pixels (referred to as "superpixels" in this paper, which are N x N pixel arrays, a subset of the entire detector array). See also Figure 6 Each superpixel 601a...601k comprises a cluster of adjacent pixels, the cluster including at least one unobstructed (unmasked) pixel 612 and at least one obstructed (masked) pixel 613. The system may include a signal processing chain 608 that receives the output of each pixel and / or superpixel and generates LiDAR histograms or other sensor data from pixel readings. The signal processing chain 608 may include hardware and programming instructions commonly used in autonomous vehicles or other now- or later-known LiDAR and other single-photon counting detection systems. However, the system also includes a noise estimation module 607, which includes a processor, memory, and programming instructions for estimating the inherent noise in each superpixel. The processor and memory of the noise estimation module 607 may be integrated with the signal processing chain 608 and / or Figure 1 The same hardware and / or software components are used in LiDAR systems, or may contain separate hardware and / or software components.

[0053] To determine the inherent noise rate of a single pixel or superpixel, the system can employ a function describing the composite signal generated by the unobstructed (unmasked) detector:

[0054]

[0055] in:

[0056] λ = avalanche count rate received by the superpixel;

[0057] = The photon count rate generated by the reflected signal photons at detector i;

[0058] = The photon count rate generated at detector i by background signal photons (i.e., ambient light and scattered light);

[0059] = The avalanche count rate of the inherent noise of detector (i) (e.g., dark count rate (DCR) and crosstalk from other detectors; and

[0060] β = Photon detection efficiency, the probability that an incident photon will generate photocarriers and a continuous avalanche.

[0061] The photon detection efficiency β is a combination of quantum efficiency, breakdown probability, fill factor, and the reflection and absorption properties of the layer above the absorption layer. Typical quantum efficiencies of InGaAs devices are around 80% to 90%, while β is only on the order of 20% to 30%. Quantum efficiency is the probability that a photon incident on the active region is converted into exciton pairs (i.e., electron-hole pairs). These photogenerated carriers, in addition to dark current carriers, are also called principal carriers, which are likely to pass through the amplification stage and be recorded as counts. In Geiger mode, the reverse bias is a voltage higher than the breakdown voltage to create a very high gain (>>1) region in the semiconductor device. A small fraction of these principal carriers leads to an avalanche event, the arrival time of which is stored by a digital counter. This small fraction is determined by the breakdown probability.

[0062] avalanche count rate Similar to the first-order inherent noise of the Geiger-mode APD, and excluding effects such as after-pulse and early fire, which are assumed to be blocked for a sufficiently long holding time, i.e., the time before the pixel is reset after the count has been registered.

[0063] If the above equation is applied to each pixel, the noise estimation module 601 can identify... And determine its effect on the total compound avalanche count rate λ.

[0064] The determination of the inherent noise rate is not limited to the specific model mentioned above. Other models, including nonlinear models, can be used.

[0065] Figure 7 Yes Figure 6 The flowchart expands upon the publicly available information and describes, in block diagram form, various processing steps for an example of how blocked pixel signals can be combined with the remaining pixels in a superpixel for noise estimation. It also depicts how this information can be used in a typical LiDAR signal processing chain.

[0066] In the noise estimation stage 607, one capability enabled by the presence of blocked pixels is the ability to characterize crosstalk in superpixels. Crosstalk estimation is accomplished by applying a pattern-matching routine, or by using a cross-correlation technique between active and blocked pixel signals, to obtain a score (i.e., a probability value) indicating that a specific sample (i.e., photon count at a specific timestamp) from any unblocked pixel is a result of a crosstalk event rather than steady-state noise (i.e., DCR). This operation is performed via a crosstalk analysis stage 701 using active pixel data 702 and blocked pixel signal data 703, as shown, which can be retrieved from memory. The aggregated samples (from blocked and active pixels) along with their corresponding scores are then passed to a classifier 704, which aims to separate these samples into two streams: one stream 705 consisting of samples caused by crosstalk, and another stream 706 consisting of samples caused by uniform noise (e.g., background light and dark count noise).

[0067] The samples induced by uniform noise are expected to have steady-state statistical properties, meaning they are time-invariant and can be effectively used to estimate the dark count rate in a particular superpixel (by using those samples obtained only from the blocked pixels). Summarizing this dark count rate estimate from this superpixel and other superpixels yields a spatial profile of the dark count rate across the entire detector array, which can be used by the health monitoring function 707 to determine the health of the detector array. If the spatial variation, temporal variation, or both of the dark count rate exceeds a certain tolerance margin, the detector array is considered defective.

[0068] In signal processing stage 608, Figure 7 The application of blocked pixel signals in a typical LiDAR signal processing chain is illustrated. The chain takes a histogram 711 of pixel samples as input. The histogram is then passed to a detection stage 722, which aims to identify specific regions in the histogram 711 expected to contain potential objects. These regions are marked with gray bars in the histogram 711. Once this portion of the histogram is isolated (see 714), it is passed to a waveform analysis stage 723, which performs accurate distance estimation and reflected pulse intensity estimation on that region of the histogram.

[0069] The blocked pixels enable crosstalk and steady-state noise classification, which can be utilized in the signal processing chain as follows: the accurate noise estimate obtained from the steady-state noise estimation stage 607 is fed into detection 722, as it is directly used as the detection threshold. Crosstalk samples have temporal statistical characteristics, which can be characterized by the characterization function 708 to estimate the measurement bias (amplification) in the intensity estimation caused by crosstalk, and this is compensated for in the waveform analysis stage 723.

[0070] Furthermore, when a reflected signal is received, the system can compare the determined inherent noise with the signal algorithm output to determine whether the received signal is just noise or whether it includes a reflected signal plus noise.

[0071] When analyzing waveforms, LiDAR systems can measure the number of received photons (pulse intensity). Noise can amplify the intensity measurement results. By measuring the contribution of noise to the signal content, the system can correct for biases in pulse intensity estimation.

[0072] Finally, as mentioned above Figure 7 As explained in the discussion, the system can measure the health of the LiDAR detector array by monitoring the spatial variation of measurement noise from the blocked detector.

[0073] The above-described system and method can be useful in various types of LiDAR systems, and are particularly useful in systems with pixel densities smaller than the crosstalk length. For example, refer to Figure 2 An exemplary detector array 200, if the pixel pitch of the array (i.e., the distance between the center points of adjacent pixels) is 5 mm or less, can have a significant value of crosstalk length of 5 mm or more by using the mask as described above.

[0074] For example, in some embodiments, the above system can be used in conjunction with the LiDAR system of an autonomous vehicle. Figure 8 An example system architecture 800 for a vehicle, such as an autonomous vehicle, is shown. The vehicle includes an engine or motor 802 and various sensors for measuring various parameters of the vehicle and / or its environment. Operating parameter sensors common to both types of vehicles include, for example: a position sensor 836, such as an accelerometer, gyroscope, and / or inertial measurement unit; a speed sensor 838; and an odometer sensor 840. The vehicle may also have a clock 842, which the system uses to determine vehicle time during operation. The clock 842 can be encoded into the vehicle's onboard computing equipment; it can be a separate device or multiple clocks can be used.

[0075] The vehicle will also include various sensors that operate to collect information about the environment in which the vehicle travels. These sensors may include, for example: a position sensor 860, such as a Global Positioning System (GPS) device; object detection sensors, such as one or more cameras 862; a LiDAR sensor system 864; and / or a radar and / or sonar system 866. Sensors may also include environmental sensors 868, such as precipitation sensors and / or ambient temperature sensors. The object detection sensors enable the vehicle to detect objects within a given distance of the vehicle 899 in any direction, while the environmental sensors collect data about environmental conditions within the vehicle's operating area. The system will also include one or more cameras 862 for capturing environmental images.

[0076] During operation, information is transmitted from sensors to an onboard computing device 820. The onboard computing device 820 may include a processor 851 and a memory device 852 having programming instructions that, when executed, cause the processor 851 to analyze the data captured by the sensors and optionally control the operation of the vehicle based on the analysis results. For example, the onboard computing device 820 may control braking via a brake controller 822; control direction via a steering controller 824; control speed and acceleration via a throttle controller 826 (in a gasoline-powered vehicle) or an electric motor speed controller 828 (such as a current level controller in an electric vehicle); a differential controller 830 (in a vehicle with a transmission); and / or other controllers. Figure 6 and Figure 7 The signal processing and / or noise estimation functions are provided. The memory device 852 of the onboard computing device 820 or another memory device in the system (such as the memory device of a LiDAR system) can provide the function of a data logger, which is configured to receive signals from pixels and store characteristic data corresponding to the received signals.

[0077] Geographic location information can be transmitted from location sensor 860 to onboard computing device 820, which can then access a map of the environment corresponding to the location information to determine known fixed features of the environment, such as streets, buildings, stop signs, and / or stop / go signals. Images captured from camera 862 and / or object detection information captured from sensors such as LiDAR system 864 are transmitted from these sensors to onboard computing device 820. The object detection information and / or captured images can be processed by onboard computing device 820 to detect objects near vehicle 800. Additionally or alternatively, the AV can transmit any data to an external server for processing. Any known or to be known techniques for object detection based on sensor data and / or captured images can be used in the embodiments disclosed herein.

[0078] In the various embodiments discussed herein, the description illustrates that the vehicle or its onboard computing device can implement programming instructions that cause the vehicle's onboard computing device to make decisions and use those decisions to control the operation of one or more vehicle systems. However, the embodiments are not limited to this arrangement, as in various embodiments, analysis, decision-making, and / or operational control can be handled wholly or partially by other computing devices that are in electronic communication with the vehicle's onboard computing device. Examples of such other computing devices include electronic devices (such as smartphones) associated with occupants of the vehicle, and remote servers that are in electronic communication with the vehicle via wireless communication networks. The processor of any such device can perform the operations discussed below.

[0079] The processing and operational steps described herein can also be guided by a computer program product, which includes a memory storing programming instructions that, when executed, will cause one or more processors to perform any of the functions described above.

[0080] Terms related to the disclosures provided above include:

[0081] The term "vehicle" refers to any form of mobile transport capable of carrying one or more human passengers and / or goods and powered by any form of energy. The term "vehicle" includes, but is not limited to, automobiles, trucks, vans, trains, autonomous vehicles, aircraft, and unmanned aerial vehicles. An "autonomous vehicle" is a vehicle with a processor, programmed instructions, and drivetrain components that can be controlled by the processor without a human operator. An autonomous vehicle may be fully autonomous, meaning that no human operator is required for most or all driving conditions and functions, or it may be semi-autonomous, meaning that a human operator may be required under certain conditions or during certain operations, or that a human operator can override the vehicle's autonomous system and gain control of the vehicle. Autonomous vehicles also include vehicles whose autonomous systems enhance human operation, such as vehicles with driver-assisted steering, speed control, braking, parking, and other systems.

[0082] As used herein, the term "light" refers to electromagnetic radiation associated with optical frequencies, such as ultraviolet, visible, infrared, and terahertz radiation. Exemplary light emitters include laser emitters and other emitters that emit focused light. In this document, the term "emitter" will be used to refer to a light emitter, such as a laser emitter that emits infrared light.

[0083] "Electronic device" or "computing device" refers to a device that includes a processor and memory. Each device may have its own processor and / or memory, or the processor and / or memory may be shared with other devices, as in a virtual machine or container device. The memory will contain or receive programming instructions that, when executed by the processor, cause the electronic device to perform one or more operations according to the programming instructions.

[0084] The terms “memory,” “memory device,” “data storage,” “data storage facility,” etc., refer to a non-transitory device on which computer-readable data, programming instructions, or both are stored. Unless otherwise specified, the terms “memory,” “memory device,” “data storage,” “data storage facility,” etc., are intended to include embodiments of a single device, embodiments of multiple memory devices together or collectively storing a set of data or instructions, and individual sectors within such devices.

[0085] The terms "processor" and "processing device" refer to hardware components of an electronic device configured to execute programmed instructions, such as a microprocessor or other logic circuitry. Processors and memories can be microcontrollers, custom-configurable integrated circuits, programmable system-on-a-chip (SoC), or other elements of an electronic device capable of being programmed to perform various functions. Unless otherwise specified, the singular terms "processor" or "processing device" are intended to include embodiments of a single processing device as well as embodiments in which multiple processing devices together or collectively perform processing.

[0086] The terms "photodetector" and "photoelectric sensor" can be used interchangeably in this article. There is no difference in meaning between the two terms.

[0087] The features and functions disclosed above, as well as alternatives, can be combined into many other different systems or applications. Various components can be implemented in hardware, software, or embedded software. Those skilled in the art can make various alternatives, modifications, variations, or improvements that are not currently foreseen or anticipated, each of which is also intended to be included in the disclosed embodiments.

Claims

1. A system comprising: A single-photon counting sensor array, comprising: One or more transmitters are configured to emit multiple energy pulses; A detector array comprising multiple pixels, wherein each pixel includes one or more detectors, and the plurality of said one or more detectors are configured to receive reflected energy pulses emitted by said one or more transmitters; and A mask material is positioned to cover part of the detectors, rather than all of the detectors, of the plurality of pixels to produce blocked and unblocked pixels, such that each blocked pixel is prevented from detecting the reflected energy pulse. Processor; and The memory contains programming instructions configured to instruct the processor to: receive data of signals generated by the blocked pixels and the unblocked pixels, and compare the data of signals generated by the blocked pixels with the data of signals generated by the unblocked pixels using a similarity measurement technique to characterize the crosstalk features within the generated signals, and determine the measurement results of the inherent noise in the detector array.

2. The system according to claim 1, wherein the similarity measurement technique includes at least one of pattern matching routines and cross-correlation techniques.

3. The system according to claim 1, wherein, The single-photon counting sensor array is a component of a light detection and ranging LiDAR system in which the distance between the center points of adjacent pixels in the detector array is less than the crosstalk length.

4. The system according to claim 1, wherein, Each detector includes a photoelectric sensor, the photoelectric sensor comprising: The surface region of the p-type semiconductor material is positioned to receive light. The drift region of n-type semiconductor materials, and A conductive trace, the conductive trace being connected to the surface region and positioned as the anode of the photoelectric sensor; The detector array includes a substrate on which photodetectors are positioned, wherein the substrate is configured to serve as the cathode of each photodetector in the photodetectors; and Each blocked pixel has a mask material positioned to cover its surface area to block light from entering the blocked pixel.

5. The system according to claim 1, wherein: Each detector includes a photoelectric sensor, and the detector array includes a substrate configured to serve as the cathode of each photoelectric sensor. Each of the photoelectric sensors includes: A metal window extending through the substrate and configured to receive light entering the photoelectric sensor. An n-type semiconductor material region, wherein the n-type semiconductor material region is connected to the metal window. p-type semiconductor material region, the p-type semiconductor material region being connected to the n-type semiconductor material region, and A conductive trace, the conductive trace being connected to the p-type semiconductor material region and positioned as the anode of the photoelectric sensor; and The mask material is positioned as a metallic window covering each blocked pixel.

6. The system according to claim 1, wherein, The programming instructions configured to instruct the processor to compare data of the signal generated by the blocked pixel with data of the signal generated by the unblocked pixel to characterize the crosstalk within the generated signal and determine the measurement results of the inherent noise in the detector array include instructions for the following: Identify superpixels that include a set of pixels in the detector array; Determine the total photon count rate received by the superpixel; Determine the avalanche count rate received by the unblocked pixels of the superpixel; and The noise measurement result is determined as a function of the total photon count rate and the avalanche count rate received by the unblocked pixels of the superpixel.

7. The system according to claim 6, wherein, The function is: in: λ =Avalanche count rate received by the superpixel; =In photoelectric sensor i The photon count rate generated by reflected signal photons at the location; =In the detector i Photon count rate generated by background signal photons; =detector i The avalanche count rate of the inherent noise; and β =Photon detection efficiency.

8. The system of claim 1 further includes a data logger configured to receive signals from pixels and store characteristic data corresponding to the received signals.

9. The system according to claim 1, wherein, The mask material includes metal.

10. The system of claim 1, further comprising additional programming instructions configured to instruct the processor to measure the health status of the detector array by monitoring spatial variations, or temporal variations, or both spatial and temporal variations of noise measured from the blocked detector.

11. The system of claim 1, further comprising additional programming instructions configured to instruct the processor to apply a characterization function to estimate the amount of measurement bias caused by crosstalk in the intensity estimation.

12. A method of operating a single photon counting sensor array, the method comprising: Operating a LiDAR system, the system comprising: One or more transmitters are configured to emit multiple energy pulses; A detector array comprising multiple pixels, wherein each pixel includes one or more detectors, the plurality of said one or more detectors being configured to receive reflected energy pulses emitted by said one or more transmitters, and A mask material is positioned to cover a portion, rather than all, of the detectors of the plurality of pixels to generate blocked and unblocked pixels, thereby preventing each blocked pixel from detecting reflected energy pulses; and By processor: Receive data of signals generated by the blocked pixels and the unblocked pixels, and Similarity measurement techniques are used to compare the data of the signal generated by the blocked pixels with the data of the signal generated by the unblocked pixels to characterize the crosstalk features within the generated signals and determine the measurement results of the inherent noise in the detector array.

13. The method according to claim 12, wherein: Each detector includes a photoelectric sensor, the photoelectric sensor comprising: The surface region of the p-type semiconductor material is positioned to receive light. The drift region of n-type semiconductor materials, and A conductive trace, the conductive trace being connected to the surface region and positioned as the anode of the photoelectric sensor; The detector array includes a substrate on which photodetectors are positioned, wherein the substrate is configured to serve as the cathode of each photodetector in the photodetectors; and Each blocked pixel has a mask material positioned to cover its surface area to block light from entering the blocked pixel.

14. The method according to claim 12, wherein: Each detector includes a photoelectric sensor, and the detector array includes a substrate configured to serve as the cathode of each photoelectric sensor. Each of the photoelectric sensors includes: A metal window that extends through the substrate and is configured to receive light entering the photoelectric sensor; An n-type semiconductor material region, the n-type semiconductor material region being connected to the metal window; A p-type semiconductor material region, wherein the p-type semiconductor material region is connected to the n-type semiconductor material region; and A conductive trace, the conductive trace being connected to the p-type semiconductor material region and positioned as the anode of the photoelectric sensor; and The mask material is positioned as a metallic window covering each blocked pixel.

15. The method according to claim 12, wherein, Comparing data from signals generated by the blocked pixels with data from signals generated by the unblocked pixels to characterize crosstalk within the generated signals and to determine measurements of inherent noise in the detector array includes: Identify superpixels that include a set of pixels in the detector array; Determine the total photon count rate received by the superpixel; Determine the avalanche count rate received by the unblocked pixels of the superpixel; and The noise measurement result is determined as a function of the total photon count rate and the avalanche count rate received by the unblocked pixels of the superpixel.

16. The method according to claim 15, wherein, The function is: in: λ =Avalanche count rate received by the superpixel; =In photoelectric sensor i The photon count rate generated by reflected signal photons at the location; =In the detector i Photon count rate generated by background signal photons; =detector i The avalanche count rate of the inherent noise; and β =Photon detection efficiency.

17. The method of claim 12, further comprising using a data logger: Receive signals from the pixels; and Store the characteristic data corresponding to the received signals.

18. The method of claim 12, further comprising: The processor measures the health of the detector array by monitoring spatial or temporal variations in noise measured from the blocked detectors, or both spatial and temporal variations.

19. The method of claim 12, further comprising: The processor applies a characterization function to estimate the measurement bias caused by crosstalk in the intensity estimation.

20. A computer-readable storage medium storing programming instructions, which, when executed, cause one or more processors to: Operating a LiDAR system, the system comprising: One or more transmitters, the transmitters being configured to emit multiple energy pulses; A detector array comprising multiple pixels, wherein each pixel includes one or more detectors, the plurality of said one or more detectors being configured to receive reflected energy pulses emitted by said one or more transmitters, and A mask material is positioned to cover part of the detectors, rather than all of the detectors, of the plurality of pixels to produce blocked pixels and unblocked pixels to prevent each blocked pixel from detecting reflected energy pulses. Receive data of signals generated by the blocked pixels and the unblocked pixels, and Similarity measurement techniques are used to compare the data of the signal generated by the blocked pixels with the data of the signal generated by the unblocked pixels to characterize the crosstalk features within the generated signals and determine the measurement results of the inherent noise in the detector array.

Images