Adaptive exposure and glare suppression night license plate snapshot camera system and camera

Abstract

The invention discloses an adaptive exposure and glare suppression night license plate snapshot camera system and camera, and relates to the technical field of license plate snapshot, and the method comprises the steps: obtaining power supply capability information and temperature parameters, forming power supply budget parameters and heat margin parameters, and predicting each lane snapshot window; determining a license plate candidate area, generating a saturation risk index and a readability index, and determining a corresponding threshold value; an exposure parameter and an infrared light supplement parameter are jointly determined before a snapshot window arrives, light supplement arbitration is carried out when multiple lane windows are overlapped, and a snapshot plan is generated; and executing snapshot according to the snapshot plan and performing closed-loop correction based on a snapshot result. According to the scheme, exposure and light supplement control can be coordinated under power supply limitation, thermal state change and multi-lane concurrence scenes, license plate overexposure and glare interference are inhibited, license plate character readability is kept, and night snapshot stability and continuous operation capability are improved.

Description

Adaptive exposure and glare suppression nighttime license plate capture system and camera Technical Field

[0001] This invention relates to the field of license plate capture technology, specifically to a nighttime license plate capture imaging system and camera with adaptive exposure and glare suppression. Background Technology

[0002] This technology is applied in scenarios such as road checkpoints, urban intersections, and highway entrances / exits. It captures images of multi-lane vehicles at night to obtain evidence and license plate information. High vehicle speeds and short effective capture times can lead to variations in light and shadow distribution. Headlight glare, license plate retroreflective materials, and rain / fog reflections can cause localized highlights or contrast changes. Additionally, the image may contain vehicles at varying distances and complex reflective backgrounds, resulting in significant differences in installation requirements. Projects often utilize network cameras with infrared illumination and automatic exposure, capturing images through preview / recognition frames and capture frames simultaneously. These systems require low light pollution, glare, PoE power supply, and minimal equipment temperature rise. This approach is commonly used in engineering projects.

[0003] Chinese patent document CN119653245A (publication date: March 18, 2025) discloses a method, apparatus, and computer device for adjusting the exposure of a captured frame. This method acquires video frames through real-time monitoring data. The video frames include recognition frames captured when the supplementary light is off and captured frames captured when the supplementary light is on. If the current recognition frame is the target recognition frame at the pre-capture position, the brightness of the recognition frame is calculated based on the target recognition frame, and a pre-stored equivalent supplementary light brightness is obtained. The equivalent supplementary light brightness is calculated based on previous captured frames. A brightness compensation ratio is calculated based on the current recognition frame brightness and the equivalent supplementary light brightness, and the exposure gain used for the next captured frame is set. The document also describes acquiring the pixel brightness components of the target recognition frame while calculating the recognition frame brightness, and calculating the equivalent supplementary light brightness in conjunction with key area information. The equivalent supplementary light brightness is formed and pre-stored from the brightness statistics of previous captured frames and is called upon in subsequent captures. The corresponding apparatus and computer device are used to store the equivalent supplementary light brightness and execute the above-mentioned exposure gain adjustment process.

[0004] However, the above method still has limitations in multi-lane highway nighttime capture scenarios: First, the equivalent supplementary lighting brightness is obtained from the current capture frame. When lane illumination changes rapidly (due to uneven lighting or oncoming headlights), differences in historical brightness and target reflectivity can cause compensation errors, resulting in local overexposure or underexposure. Second, due to license plate retroreflection, the energy of near-range local echoes increases, and saturation is generally concentrated in the character area. If brightness statistics cannot stably reflect local saturation, it is difficult to avoid areas with white areas. Third, supplementary lighting output may fluctuate or fail to activate due to differences in power supply capacity and temperature protection. Therefore, the above brightness relationship may not yield consistent conclusions under different installation and climatic conditions. Gain compensation is prone to noise under low-light conditions, and this deviation is difficult to correct through multiple adjustments within a short capture window, leading to evidence failure, reduced recognition rate, and increased maintenance costs.

[0005] Therefore, the core technical problem that needs to be solved can be summarized as: how to achieve stable exposure and supplementary lighting control for nighttime multi-lane short-window license plate capture under engineering conditions such as PoE power consumption and thermal constraints, while taking into account both close-range overexposure prevention and long-range readability. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] To address the shortcomings of existing technologies, this invention provides an adaptive exposure and glare suppression nighttime license plate capture system and camera. The system determines candidate license plate regions, generates saturation risk and readability indicators, and determines corresponding thresholds. Before the capture window arrives, exposure parameters and infrared illumination parameters are jointly determined, and illumination arbitration is performed when multiple lanes overlap to generate a capture plan. Capture is executed according to the plan, and closed-loop correction is performed based on the capture results. This solution can coordinate exposure and illumination control under conditions of limited power supply, thermal variations, and concurrent multi-lane scenarios, suppressing license plate overexposure and glare interference, maintaining license plate character readability, and solving the technical problems described in the background art.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] An adaptive exposure and glare suppression nighttime license plate capture system, including:

[0011] Acquire power supply capacity information and temperature parameters to form power supply budget parameters and heat margin parameters, and output the arrival time and width of the lane capture window on the preview video stream; determine the license plate candidate area, calculate the saturation risk index and readability index, and determine the first threshold and the second threshold based on the power supply budget parameters and heat margin parameters;

[0012] Before the capture window arrives, the exposure parameters and infrared supplementary lighting parameters are jointly determined to ensure that the saturation risk index does not exceed the first threshold and the readability index is not lower than the second threshold, while also meeting the power supply budget parameters and heat margin parameters. When the lane capture windows overlap, a capture plan with supplementary lighting arbitration is generated. Before the capture window arrives, the exposure parameters are written, and the infrared supplementary lighting is triggered to acquire the capture image within the window according to the capture plan. The threshold adaptive bias parameters and thermal prediction or energy availability estimation are updated based on the capture image.

[0013] Furthermore, the power supply capacity information and temperature parameters are validated. If the validation fails, the conservative default power supply budget parameters and heat margin parameters are used, and the duty cycle, peak value and pulse count in the infrared supplementary light parameters are limited. The abnormal flag is associated with the corresponding lane number, stored and written into the capture plan.

[0014] Furthermore, vehicle detection and tracking are performed on the preview video stream to obtain target trajectory information. Based on the target trajectory information and the preset trigger line position, the arrival time and width of the capture window are extrapolated. Window event identifiers are generated according to lane number and output to the event scheduling queue.

[0015] Furthermore, in the preview video stream, the license plate candidate region is determined based on the window event identifier, and the highlight candidate region and background candidate region are segmented around the license plate candidate region. The highlight candidate region is generated by the connected bright region, and the background candidate region avoids the highlight candidate region.

[0016] Furthermore, a saturation risk index is generated based on the proportion of saturated pixels and the proportion of saturated connected regions within the license plate candidate area, and a readability index is generated based on the gradient energy and local contrast within the license plate candidate area. The contribution of the highlight candidate area to the saturation risk index is also incorporated into the saturation risk index.

[0017] Furthermore, based on power supply budget parameters, heat margin parameters, and lane distance segmentation, a first threshold and a second threshold are determined for each lane. The capture window width is used as the input for threshold mapping to form a threshold set, thereby outputting a structured result of lane-index-threshold.

[0018] Furthermore, a discrete combination of the exposure parameter candidate set and the infrared illumination parameter candidate set is constructed, and the highest priority combination is selected after filtering according to the first threshold, the second threshold, the power supply budget parameter and the thermal margin parameter. The capture plan entries are generated by looking up the table and performing a one-step correction.

[0019] Furthermore, when overlapping capture windows of two lanes are detected, the controller divides the overlapping time into time slices and performs quota arbitration on the number of pulses and peak values ​​in the infrared supplementary light parameters of each lane, so that the total supplementary light usage in any time slice does not exceed the power supply budget parameter and is limited by the heat margin parameter. When the heat margin parameter is close to the disable threshold, a prediction derating flag is written.

[0020] Furthermore, when there is no parameter combination that simultaneously satisfies the first threshold, the second threshold, the power supply budget parameter, and the heat margin parameter, the controller enters a preset degradation mode. The degradation modes include a license plate protection priority mode and a non-disruption priority mode, and a degradation mode flag and a degradation reason flag are written into the capture plan.

[0021] Furthermore, when the system is configured with an energy buffer unit, the controller charges the energy buffer unit in a controlled manner under the condition that the power supply budget parameters are not exceeded, and outputs pulse supplementary light energy within the capture window. The output is driven by the arrival time of the capture window.

[0022] The system also calculates saturation risk and readability indices for license plate candidate regions in captured images to update threshold adaptive bias parameters and update thermal prediction and energy availability estimates.

[0023] An adaptive exposure and glare suppression nighttime license plate capture camera includes a camera, an infrared illuminator, a temperature acquisition unit, a power supply interface, a controller, and a memory. The memory stores a program that can be executed by the controller, which executes the program to achieve the following:

[0024] The power supply and thermal status acquisition and capture window prediction module is used to acquire power supply capacity information and temperature parameters, form power supply budget parameters and thermal margin parameters, and output the arrival time and capture window width of the lane capture window on the preview video stream; the candidate region analysis and threshold determination module is used to determine the license plate candidate region, calculate the saturation risk index and readability index, and determine the first threshold and the second threshold based on the power supply budget parameters and thermal margin parameters.

[0025] The exposure and supplementary lighting joint planning and arbitration module is used to jointly determine the exposure parameters and infrared supplementary lighting parameters before the capture window arrives, so that the saturation risk index does not exceed the first threshold and the readability index is not lower than the second threshold, and meets the power supply budget parameters and heat margin parameters. When the lane capture windows overlap, a capture plan with supplementary lighting arbitration is generated.

[0026] The timing-based write triggering and closed-loop update module is used to write exposure parameters before the capture window arrives, and trigger infrared supplementary light to acquire capture images according to the capture plan within the window. Based on the captured images, the module updates the threshold adaptive bias parameters and thermal prediction or energy availability estimation.

[0027] (III) Beneficial Effects

[0028] This invention provides an adaptive exposure and glare suppression nighttime license plate capture system and camera, which have the following beneficial effects:

[0029] By acquiring power supply capacity and temperature information, power supply budget parameters and thermal margin parameters are generated. Combined with target trajectory information, the arrival time and width of the capture window are predicted, so that subsequent exposure parameters and infrared illumination parameters can be determined. Furthermore, the subsequent exposure parameters and infrared illumination parameters are established on the actual power supply boundary, thermal boundary, and lane timing boundary, avoiding the problems of delayed illumination time or mismatched execution conditions in short-window nighttime scenes.

[0030] By simultaneously generating saturation risk and readability indices for license plate candidate areas, and establishing first and second thresholds for different lanes, the controller avoids solely adjusting based on local or overall frame brightness. Instead, it places license plate overexposure risk and character legibility on the same decision chain, preventing situations where the license plate area is suppressed and darkened overall, but the characters remain unclear. By using saturation risk, readability, power budget, and thermal margin parameters as jointly determined constraint inputs, and completing the selection and correction of exposure and infrared illumination parameters before the capture window, the capture plan becomes executable, avoiding the instability caused by sequentially capturing and compensating samples. This enhances adaptability to scenarios with uneven lanes and intermittent traffic.

[0031] In step three, multi-lane concurrent arbitration is set by overlapping multi-lane capture windows, the quota of infrared supplementary light parameters is allocated according to time slice, and the boundary conditions are processed by combining prediction derating and degradation modes, thereby avoiding power supply over-limit, thermal state change and long-term mismatch of local lanes caused by the simultaneous superposition of supplementary light.

[0032] By recalculating the saturation risk index and readability index based on the captured images and updating the threshold adaptive bias parameters, thermal prediction, and energy availability, steps one through four connect the license plate candidate area and the capture plan, continuously adapting to the variable state under changes in power supply level, thermal state accumulation, and on-site lighting. Attached Figure Description

[0033] Figure 1 is a schematic diagram of the hardware structure and control connection of the nighttime license plate capture imaging system of the present invention;

[0034] Figure 2 is a schematic diagram of the overall process of the nighttime license plate capture method of the present invention;

[0035] Figure 3 is a schematic diagram of the constraint input formation and the establishment of the event queue for the capture window according to the present invention;

[0036] Figure 4 is a schematic diagram of the vehicle front search window and three types of candidate areas of the present invention;

[0037] Figure 5 is a schematic diagram of the capture plan generation and multi-lane time slice quota arbitration of the present invention;

[0038] Figure 6 is a timing diagram of exposure pre-writing, supplementary light pulse execution, and closed-loop write-back in this invention. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Please refer to Figures 1-6. This invention provides a nighttime license plate capture system with adaptive exposure and glare suppression, comprising:

[0041] The failure of nighttime license plate capture is not simply due to insufficient brightness, but rather to the simultaneous pressure of three types of constraints within the same capture window: one constraint comes from the power supply limit of the Ethernet power interface. If the supplementary lighting device continuously charges or maintains a high duty cycle before and after the window, the entire device will reach the power boundary of the receiving side; another constraint comes from the thermal state of the supplementary lighting device, the housing, and the processor. Heat accumulation will push subsequent windows into the derating zone under continuous traffic flow; the third constraint comes from vehicle movement. The time from when the license plate appears in the preview image to when it enters the effective capture area is very short. If we still rely on continuous capture samples to try and fail, the parameter experience formed by the previous vehicle often cannot be transferred to the next vehicle. Especially in scenarios with lanes of varying widths and multiple lanes in concurrency, the adjustment direction given in the previous frame will instead push the next frame towards overexposure or underexposure.

[0042] Step 1: Before the target enters the effective capture area, converge the power supply boundary, thermal boundary, and lane timing boundary into constraint inputs that can be directly invoked in subsequent steps, so that the capture action is transformed from on-site trial and error to pre-set execution before the window arrives.

[0043] When a nighttime license plate capture system is deployed on-site, the supplementary lighting device is not a power-dependent load. The image sensor, encoding / decoding circuit, network communication circuit, and memory all draw power from the same power path. Simultaneously, the heat released by the supplementary lighting device after operation will conversely limit the duty cycle and peak value of the next supplementary lighting cycle. If the controller only knows that power is currently being supplied or the current temperature is too high, these states remain at the alarm level and cannot provide executable constraints for the license plate candidate area judgment in step two and the parameter linkage in step three.

[0044] Therefore, the measurable power supply capacity information and temperature parameters on site are first compressed into the power supply budget parameters for subsequent deployment. With heat margin parameters And when forming parameters, specify the corresponding supplementary light power consumption and pulse strategy boundary.

[0045] Unavailable power supply information includes power negotiation field timeout failure to update; unreliable power supply information includes power level field verification failure, power level field invalidity, or power level jump before renegotiation is completed. Abnormal temperature parameters include sampling timeout, frozen readings, reading abrupt changes exceeding the predetermined physical slope, and readings exceeding the sensor's range.

[0046] The controller first reads the power supply level, the negotiated available power level, and the current basic power consumption distribution status of the whole machine from the Ethernet power receiving circuit, and then reads the temperature of the supplementary lighting device, the chassis temperature, and the processor temperature from the temperature acquisition unit.

[0047] The order of power supply first, followed by thermal readout, is adopted because the effectiveness of the thermal strategy depends on the currently available supplementary lighting energy: when the power supply margin is already tight, the thermal strategy does not need to wait for the temperature to approach the disable boundary before taking action; conversely, when the power supply margin is sufficient but the thermal margin is limited, the controller places more emphasis on limiting subsequent supplementary lighting based on single peak values ​​and pulse repetition intervals. Therefore, the power supply budget parameters... First, define the upper limit of the supplementary lighting resources and the thermal margin parameter. The release rhythm is further defined, and the two remain coupled during the definition process.

[0048] In the implementation of the integrated gantry camera, the Ethernet power receiving circuit is located at the rear of the housing, while the image sensor and infrared illumination device are located in the front optical compartment. The controller periodically reads the negotiation results and current basic power consumption allocation status from the power receiving side via the power receiving control bus. If the system is equipped with an energy buffer unit, the energy buffer unit is preferably a supercapacitor branch arranged on the same side as the illumination device drive circuit, and its charging and discharging path is controlled by the controller. If the system is not equipped with an energy buffer unit, the corresponding item is set to zero during parameter calculation. The controller then forms the power supply budget parameters accordingly. :

[0049]

[0050] Among them, power supply capacity parameters : Indicates the input power that the Ethernet powered circuit is allowed to allocate to the entire device in the current negotiation state; the value is a power amount greater than 0; base reserved power. This represents the reserved power for the image sensor, processor, network communication circuit, and memory in nighttime license plate capture mode. Its value is not less than 0 and is less than the power supply capacity parameter. Power quantity;

[0051] Controlled charging power : Represents the upper limit of the allowable power allocated to the charging branch of the energy buffer unit, with a value not less than 0. Its function is to limit the charging behavior before the window to prevent it from exceeding the power receiving boundary; in the implementation without an energy buffer unit, the controlled charging power Set to 0;

[0052] Release efficiency : This represents the efficiency coefficient when the energy buffer unit releases energy to the supplementary lighting device via the driving circuit. Its value is greater than 0 and not greater than 1. Its function is to convert the available energy on the buffer side into the available power on the supplementary lighting side; available energy status. : Indicates the amount of energy that the energy buffer unit can use for subsequent capture windows at the current moment, and the value is not less than 0;

[0053] Scheduling Time Window : Indicates the energy allocation time from the current moment to the next recalculation of the capture plan, and the value is greater than 0;

[0054] As a supplement: when the energy buffer unit is a supercapacitor branch, the available energy state... Obtain it in the following form:

[0055]

[0056] capacitance value The nominal capacitance value of the supercapacitor branch is given in the device datasheet or by factory measurement; terminal voltage This indicates the voltage at the energy buffer unit terminals, which is acquired in real time by the controller via an analog-to-digital sampling circuit; minimum discharge voltage. This represents the minimum terminal voltage allowed for the drive circuit to maintain stable discharge, and is given by the hardware design.

[0057] When the energy buffer unit is a battery branch, the battery solution no longer uses the above capacitor energy formula, but the battery management circuit directly outputs the estimated remaining energy value as the available energy state. The controller only limits the boundary amplitude; controlled charging power It is obtained by multiplying the current limit value of the charging control loop by the real-time charging voltage; scheduling time window The time frame is calculated from the current time to the most recent recalculation time of the capture plan and from the current time to the earliest arrival time of the capture window. The smaller of the two.

[0058] This formula calculates the upper bound of the supplementary lighting resources available for subsequent steps, not the instantaneous peak value of the supplementary lighting device: when the power supply side is tighter, the power supply capacity parameter... With base reserved power The difference dominates; when the buffer side is tighter, the controlled charging power... With available energy state Release efficiency and scheduling time window The converted result is dominant. This forms the power supply budget parameters. It is suitable for both direct-powered supplemental lighting and supplemental lighting paths with power receiving and energy buffering. The power receiving control bus can use either register polling or interrupt reporting; the energy buffer unit can use either a supercapacitor branch or a pulse battery branch with a current-limiting interface, as long as the external output is uniformly mapped to the available energy state. That's all.

[0059] Power supply budget parameters After formation, the controller continues to process the thermal state, but does not directly use a certain temperature value as a disable threshold trigger. Instead, it converts the temperatures of multiple heat source locations and the cumulative effect of recent supplemental lighting into the same thermal margin parameter. .

[0060] In a preferred embodiment, the temperature of the supplementary lighting device is provided by a temperature sensor near the infrared light-emitting array pads, the housing temperature is provided by a temperature sensor on the inner wall of the housing, and the processor temperature is provided by a temperature register inside the processor; these three types of temperatures respectively characterize the heat source, the heat dissipation channel, and the control board load. The controller generates thermal margin parameters. :

[0061]

[0062] Among them, the upper limit of thermal protection This indicates the upper temperature limit reserved before the supplementary lighting device enters a state of severe derating or disabling. The value is a temperature amount higher than the ambient temperature and at least one currently measured temperature. (Supplyary lighting device temperature) : Represents the temperature collected near the infrared light-emitting array, the value of which is the actual temperature measured on site; casing temperature. : Indicates the temperature collected inside the casing or near the heat dissipation path, and the value is the actual temperature measured on site;

[0063] Processor temperature : Represents the temperature register value on the processor or main control board, which is the actual measured temperature on-site; Weighting coefficient Weighting coefficients With weighting coefficients These represent the thermal margin parameters related to the temperatures of the supplementary lighting device, the housing, and the processor, respectively. The contribution weights are all greater than 0; cumulative duty cycle : Represents the cumulative duty cycle of the supplementary lighting device within a predetermined historical time period, a normalized value not less than 0; where the attenuation coefficient is... : Indicates cumulative duty cycle For heat margin parameters The deduction intensity is a coefficient with a value greater than 0;

[0064] When the heat reserve parameter When reducing the temperature limit, subsequent steps do not wait for the temperature to exceed the limit before abruptly reducing the derating. Instead, they tighten the supplementary lighting duty cycle, peak value, or pulse repetition strategy in advance, allowing thermal limitation to be gradually implemented in the capture plan. If temperature parameters are missing, readings jump abruptly, or the sensor link malfunctions, the controller will activate the thermal protection upper limit. The difference between the current effective temperature and the current temperature is processed according to the most unfavorable combination to optimize the heat margin parameter. It directly falls into the conservative zone. Temperature anomalies include: sampling timeout, consecutive identical values ​​freezing, sudden changes exceeding the physically permissible slope, and readings exceeding the sensor's range; the temperature acquisition cycle is synchronized with or an integer multiple of the capture plan update cycle; the temperatures of the supplementary lighting device, housing, and processor are written to the thermal state cache after being filtered by a first-order low-pass filter; when any temperature enters the forbidden zone, the thermal margin parameter... Directly place in the supplementary lighting disabled zone; when it returns to normal after several consecutive update cycles and drops below the recovery threshold, the thermal margin parameter... Only then will you be allowed to leave the restricted area.

[0065] When using, power supply budget parameters With heat margin parameters Application binding is completed during the formation stage; heat reserve parameters Determined by multiple location temperatures and cumulative duty cycle, it can reveal thermal failure paths in advance.

[0066] Only power supply budget parameters With heat margin parameters This is still insufficient to guide snapshot capture, because supplemental lighting and exposure are not triggered at every moment. What truly determines the timing of the action is when the target enters the effective capture area and how long that area is open to the target. If the controller waits until the target enters the area before calculating, both supplemental lighting charging and exposure register writing will be delayed. If the path correction is continued using continuous capture samples, prolonged periods without vehicles and concurrent multi-vehicle traffic will distort the sample of the previous target for the next target. Therefore, using the preview video stream as the front-end observation, the target trajectory of each lane is extrapolated to the arrival time of the capture window. With the width of the capture window It also outputs the exception flag, so that subsequent steps can be organized around the window.

[0067] The controller first performs vehicle detection, lane assignment, and trajectory continuity maintenance on the preview video stream, and then maps the trajectory to the effective capture area on the road coordinate line. The key here is not simply recognizing the presence of a vehicle, but rather forming an extrapolable position sequence and speed sequence, because only when both exist simultaneously can the controller adjust the power supply budget parameters. and heat margin parameters This corresponds to a specific future moment. In other words, the first half of step one gives whether it is possible to fire and how much fire is possible. The second half of step one needs to apply these two answers to when to fire and which lane to fire on. Only by combining the two can we form the complete starting point for subsequent joint solutions.

[0068] In a preferred embodiment, the controller runs in an embedded processing environment on the camera system motherboard. The preview video stream is output by the image sensor at a fixed preview frame rate. Vehicle detection can employ a bounding box model based on convolutional features, or a target detection process combining background motion and shape constraints. Regardless of the detection path used, the output is uniformly the lane number, target center position, and timestamp. The controller maintains a target center position sequence for each lane and generates a smooth speed based on the position changes between the current and previous frames. It then predicts the window based on the lane trigger line and the length of the effective capture area.

[0069]

[0070] Among them, smoothing speed : Represents the target velocity used for window prediction at the current moment, with a value greater than 0; Precedence smoothing velocity. : Represents the target velocity stored in the previous time step, with a value greater than 0; smoothing coefficient : Indicates the preceding smoothing speed For the current smooth speed The retention ratio, which takes a value greater than 0 and less than 1, serves to balance response speed and trajectory stability.

[0071] Current location With the preceding position These represent the current and previous projected positions of the target center on the road reference axis, respectively. The values ​​are increments along the positive direction of the lane, and their function is to provide the displacement of the target along the lane direction.

[0072] Current moment With the preceding time These represent the current position. With the preceding position The corresponding timestamp, with an incrementing time value, is used to convert position changes into velocity; trigger line distance. : Indicates the calibration distance from the road reference origin to the trigger line at the entrance of the effective capture area, and the value is greater than the current position. Position quantity;

[0073] Region length : The effective capture location is the position along the lane length greater than 0; stable lower limit : A positive compensation term added to the denominator of the smoothed velocity, taking a velocity value greater than 0, to prevent the capture window from being reached due to target start-up, temporary occlusion, or projection jitter. With the width of the capture window A violent throbbing occurred.

[0074] This formula provides the engineering time for snapshot scheduling: smoothing speed. Once formed, the controller can pre-write the fill light charging and exposure registers to align with the time when the capture window arrives. ; Capture window width Once formed, subsequent steps can determine whether a lane is allowed only a short pulse or a longer pulse sequence. The lane target trajectory can be obtained either by extrapolating the road coordinates after monocular perspective calibration or by calculating from binocular depth information, as long as the final output still includes the arrival time of the capture window. With the width of the capture window This will maintain consistency in terminology.

[0075] At the moment the capture window is formed With the width of the capture window Afterward, the controller does not immediately trigger supplemental lighting, but instead writes the results for each lane into the window event queue. Each item in the window event queue contains at least the lane number and the arrival time of the capture window. , capture window width Power supply budget parameters Heat margin parameters And trajectory reliability flags. The reason for this arrangement is that step two needs to calculate the license plate candidate area and highlight candidate area around the predetermined window, and step three needs to arbitrate when the windows overlap. Therefore, the output of step one should be the scheduling objects sorted by lane and time.

[0076] For example, in a two-way multi-lane intersection scenario, if vehicles in the closer lane enter the effective capture area first while vehicles in the farther lane enter later, the controller will first set the capture window arrival time of the closer lane. Position it in the front and retain the arrival time of the capture window for the far lane. As a subsequent candidate; if the heat reserve parameter at this time Approaching the conservative zone, the corresponding item in the window event queue will directly be marked with a thermal derating flag. If vehicle detection is intermittent, severely obstructed, or lane affiliation cannot be maintained stably, the controller will set the trajectory reliability flag to abnormal and simultaneously adjust the power supply budget parameters. Tightening the budget to a conservative level and constraining subsequent exposure time candidates to shorter paths reserves an entry point for the degradation mode in step three; if the power supply capacity information is distorted, the controller uses the most recent valid power supply budget parameters. A conservative lower bound is applied; if the temperature link is distorted, the thermal margin parameter is used. The conservative lower bound.

[0077] When used, the output of step one is not the original observation, but a window event object that can be directly scheduled for subsequent steps; the anomaly is bound to the specific lane and specific window in step one, and subsequent steps do not need to reinterpret the source of the anomaly; the capture action is changed from seeing and then queuing before it arrives, which structurally avoids the path that relies on continuous capture samples to gradually converge.

[0078] Step 2: Based on the capture window and resource boundaries given in Step 1, simultaneously quantify the overexposure risk and character discernibility of the license plate candidate area on each lane, and form a dual constraint for subsequent joint solution calls.

[0079] When the trajectory confidence flag output in step one is abnormal, the controller does not use the target trajectory-guided vehicle head search window, but uses the pre-calibrated lane fixed capture area as the conservative license plate candidate search area, and correspondingly tightens the first threshold and increases the conservative bias corresponding to the second threshold.

[0080] Step one has already given the arrival time of the capture window. and the width of the capture window However, these two quantities only indicate when the target enters the effective capture area, but do not specify which part of the image should be focused on within that window. If the controller still uses the entire frame as the basis for subsequent judgments, headlight highlights, road surface reflections, and background signage reflections will all be mixed into the license plate appearance, causing subsequent saturation risk indicators and readability indicators to lose their specificity.

[0081] The controller receives the arrival time of the capture window for each lane. and the width of the capture window Next, the search range is narrowed down to the forward headlight area of ​​the current target in each lane using the target trajectory information from step one. Then, a headlight search window is generated based on the lane perspective calibration results and trajectory extrapolation results. Within this headlight search window, candidate license plate regions, candidate highlight regions, and candidate background regions are extracted sequentially. The license plate candidate region provides character and panel information, the highlight candidate region provides the propagation path of strong reflections, and the background candidate region provides brightness and texture references for non-license plate locations in the same lane. These three elements form the basis for the subsequent observation of dual-constraint indicators.

[0082] In a preferred embodiment, the controller first projects the target trajectory onto the preview frame according to the lane perspective calibration table, and then extracts rectangular sub-blocks in the front search window according to the forward geometry of the vehicle model. Each rectangular sub-block is not sorted according to brightness, but simultaneously considers the horizontal character texture and the aspect ratio of the panel, excluding headlight covers, brand logos and reflective trim strips.

[0083] For example, when a vehicle approaches along a certain lane, the controller sets the capture window to the time it arrives. The first few preview frames are used as candidate frames. The lower half of the front of the vehicle is found in the candidate frames. Then, the rectangular sub-blocks with regular boundaries are searched by sliding horizontally in the lower half of the front of the vehicle. When the position of a rectangular sub-block in consecutive candidate frames is consistent with the target trajectory and the horizontal light and dark alternation feature inside it is maintained, the controller registers the rectangular sub-block as a license plate candidate area.

[0084] The controller calculates the license plate aggregation score for each rectangular sub-block. :

[0085]

[0086] Among them, length and width matching quantity : Represents the degree of fit between the aspect ratio of the rectangular sub-block and the preset license plate width and length range; a dimensionless quantity with a value greater than 0; gradient aggregation quantity. : Represents the degree of energy concentration at the internal edges of a rectangular sub-block, a dimensionless quantity with a value greater than 0; continuous tracking quantity : Indicates whether the displacement of the same rectangular sub-block in adjacent candidate frames is consistent with the target trajectory; a dimensionless quantity with a value greater than 0; occlusion penalty. : Indicates the degree to which the edge of the rectangular sub-block is obscured by the vehicle frame, mud, mounting frame, or the vehicle in front, and is a dimensionless quantity with a value not less than 0;

[0087]

[0088] Among them, the candidate aspect ratio Indicates the pixel aspect ratio of the current rectangular sub-block; reference aspect ratio. This indicates the reference aspect ratio of the license plate corresponding to the current lane distance segment.

[0089]

[0090] Among them, the region Represents candidate rectangular sub-blocks; gradient components With gradient components Obtained by the Sobel or Scharr operator; area of ​​the region This represents the pixel area of ​​the candidate rectangular sub-block.

[0091] Continuous tracking volume The overlap ratio between the current candidate bounding box and the trajectory extrapolation bounding box; occlusion penalty. It is recommended to publish the sum of the effective edge lengths of the four sides of the candidate box as the inverse vector of the candidate box perimeter.

[0092] License plate aggregation score The higher the score, the closer the rectangular sub-block is to the actual license plate location. The controller preferably retains the license plate aggregation score. The largest rectangular sub-block is selected as the license plate candidate area; the license plates of the two adjacent rectangular sub-blocks are aggregated and scored. If they are close, the controller will further check their relative positions to the center line of the target trajectory and prioritize retaining the rectangular sub-block that is closer to the center of the front of the vehicle.

[0093] When using this method, the license plate candidate area is converged using target trajectory and local texture to avoid false detections caused by full-frame brightness-driven detection; License plate aggregation score. By listing boundaries, textures, and continuity as a single sorting criterion, it is easier to use in the next step; when the alternative implementation is a dual preview stream, the output object is still the license plate candidate area, without disrupting the terminology system.

[0094] As a supplement: Dimensions and width matching The gradient aggregation is calculated from the pixel width and pixel height of the candidate rectangle and compared with the license plate aspect ratio range corresponding to the lane distance segment; The gradient is obtained by accumulating the horizontal and vertical gradient magnitudes within the license plate candidate region, with Sobel or Scharr being the preferred gradient operators; continuous tracking quantity. It is calculated by combining the center offset of the candidate box in adjacent candidate frames, the scale change, and the consistency with the center line of the target trajectory.

[0095] Obstruction penalty It is derived from the effective edge coverage of the four sides of the candidate box; the more missing edges, the lower the threshold. Larger; connected area The value is obtained from the ratio of the number of pixels in the candidate connected region of highlights to the area of ​​the front search window; the steepness of the boundary is also considered. It is obtained from the average gray level difference or radial attenuation slope of the two annular bands inside and outside the boundary of the high-light connectivity domain;

[0096] Inter-area spacing The nearest pixel spacing from the boundary of the highlight candidate region to the boundary of the license plate candidate region is obtained by lane perspective calibration normalization; isolation amount It is obtained from the average transmittance difference on the darkest path between the highlight candidate region and the license plate candidate region; background reference value. The noise disturbance is obtained by adding the local texture difference, grayscale difference, and edge density difference between the license plate candidate region and the background candidate region; It is obtained from the amplitude of high-frequency brightness fluctuations in the background candidate region, preferably from the strip of candidate regions that do not contain license plate characters and highlights.

[0097] After the license plate candidate area is determined, the controller then determines the highlight candidate area and the background candidate area within the same front search window. This arrangement is not an additional action, because nighttime glare is not only manifested within the license plate candidate area. In many cases, the highlight first appears on the headlight cover, metal trim, or reflection points on wet road surfaces, and then the brightness is dragged into the edge of the license plate candidate area through optical scattering and image sensor saturation diffusion.

[0098] In terms of specific settings, the controller performs bright area connectivity decomposition on the vehicle front search window, removing bright spots with intact textures within the license plate candidate area, and retaining only a portion of connected regions with steep edges and concentrated peaks that feather outwards as highlight candidate areas; then, outside the license plate candidate area, along the same lane, a small number of strip areas surrounding the vehicle body boundary that are neither covered by the highlight candidate area nor occluded by the vehicle body boundary are registered as background candidate areas, used to measure the highlight expansion score of intruding into the highlight candidate area. :

[0099]

[0100] Among them, the connected area quantity : Represents the proportion of the connected area of ​​the highlight candidate region in the preview frame, a dimensionless quantity with a value greater than 0; boundary steepness. : Represents the steepness of the grayscale decay from the peak to the periphery in the highlight candidate region; a dimensionless quantity with a value greater than 0; region spacing. : Represents the normalized result of the nearest distance from the boundary of the highlight candidate region to the boundary of the license plate candidate region, and is a dimensionless quantity with a value not less than 0;

[0101] The controller is based on the specular extension score. Sort the highlight candidate regions and restrict the background candidate regions to those with highlight expansion scores. In localized areas where the license plate candidate area is lower and not covered.

[0102] Preferably, the background candidate area is arranged along the left and right sides or above the license plate candidate area, so that it does not contain character strokes or fall into the headlight hot zone; when there is cross headlight illumination in adjacent lanes, the controller prioritizes selecting the background candidate area inside the same lane as the license plate candidate area.

[0103] When in use, after the highlight candidate area is explicitly separated, glare propagation is no longer confused with license plate reflection; the background candidate area provides a local reference in the same lane for subsequent readability indicators; the three types of areas are collaboratively determined within the same front search window, which provides a unified observation benchmark for the future.

[0104] Simply identifying the license plate candidate region, highlight candidate region, and background candidate region is insufficient to drive the joint solution of exposure and fill light in step three. Step three needs to filter out executable combinations from the candidate exposure parameters and candidate fill light parameters before the capture window arrives. If the selection criteria are simply based on whether the area is brighter or darker, it cannot answer two more crucial questions: first, will the license plate candidate region become a blank area under the current resource boundary? Second, even if a blank area is avoided, will the character edges still be sufficient to support subsequent recognition?

[0105] Therefore, the three types of regions are compressed into two opposite but simultaneously satisfying quantities: one is a saturation risk index, used to limit the upper limit of overexposure; the other is a readability index, used to limit the lower limit of character exposure. Subsequently, the controller combines this with the power supply budget parameters given in step one. Heat margin parameters , capture window arrival time and the width of the capture window A first threshold and a second threshold are generated for each lane, making the two indicators truly have executable boundaries.

[0106] Incorporate power supply budget parameters during calculation and mapping. and heat margin parameters The same license plate candidate area appearance, under the same power supply budget parameters... Ample and heat margin parameters When sufficient, in accordance with power supply budget parameters Compact and heat margin parameters As the threshold for depreciation decreases approaches, the acceptable safety boundaries for subsequent step three differ; similarly, the capture window width... When the lane is narrower, fewer actions can be taken for fine-tuning in step three. Therefore, the threshold given in step two should be more biased towards the safe zone of a single decision. Based on this, the controller first calculates the saturation risk index and readability index, and then calculates the threshold based on lane position, distance segmentation, and power supply budget parameters. Heat margin parameters and the width of the capture window Generate the first threshold and the second threshold.

[0107] The saturation risk index is not simply a statistical measure of the percentage of pixels reaching full saturation in the license plate candidate region. Instead, it considers simultaneously the strong saturation within the candidate region, the brightness increase at its edges, and the proximity of highlight candidate regions to the license plate candidate region. Many whiteboard failures occur before full saturation is actually achieved: character strokes are first partially overflowed and covered, followed by a rapid collapse in edge contrast. Therefore, the controller constructs a saturation risk index for each lane. :

[0108]

[0109] Among them, license plate saturation : Represents the percentage of pixels exceeding the predetermined highlight threshold within the license plate candidate area; a dimensionless quantity with a value not less than 0; edge lift amount. : Represents the degree of brightness increase within the edge band of the license plate candidate region relative to the background candidate region, a dimensionless quantity with a value not less than 0; Spectrum expansion score Its function is to incorporate the intrusion trend of adjacent highlights into the license plate candidate area into the saturation risk indicator. ;

[0110] Isolation volume : Represents the degree of isolation between the highlight candidate region and the license plate candidate region formed by the dark band or structural boundary, and is a dimensionless quantity with a value not less than 0; weighting coefficient Weighting coefficients With weighting coefficients These represent the license plate saturation levels, respectively. Edge lifting amount With highlight expansion score Saturation risk indicators The contribution weights are all greater than 0;

[0111] Among them, license plate saturation Edge lift was obtained through highlight thresholding and local connectivity statistics. The isolation level is obtained by comparing the edge band of the expanded license plate candidate region with the background candidate region. It is obtained by the average transmittance of the darkest path between the highlight candidate region and the license plate candidate region.

[0112] Preferably, the controller will use the saturation risk index At the moment the capture window arrives The algorithm continuously calculates values ​​on multiple candidate frames and retains the most unfavorable value as the output value for the current window of that lane.

[0113] During use, saturation risk indicators This covers issues such as overexposure of the license plate surface and highlight intrusion paths; all saturation risk indicators originate from pre-defined areas, and step three allows for tracing specific locations for suppression; multiple frames are used to determine the most unfavorable value and the capture window width. It is well-coordinated and suitable for short-window scenarios.

[0114] For nighttime license plate capture, simply using the saturation risk indicator Lowering the brightness does not equate to character readability. When fill light is reduced, exposure is decreased, or gain is limited, character strokes enter the weak texture before the background. Therefore, when calculating readability, the controller does not primarily rely on the brightness of the entire candidate area, but rather synthesizes character edges, local contrast, and background separation. The controller constructs readability metrics. :

[0115]

[0116] Among them, edge energy : Represents the cumulative gradient of the character stroke boundaries within the license plate candidate region, a dimensionless quantity with a value greater than 0; contrast separation quantity : Represents the local brightness separation between the character sub-region and the adjacent non-character sub-region of the license plate candidate area, and is a dimensionless quantity with a value greater than 0; background reference value. : Represents the effective texture separation degree between the license plate candidate region and the background candidate region, a dimensionless quantity with a value greater than 0; noise perturbation amount : Represents the intensity of random brightness fluctuations in the preview frame near the license plate candidate area, and is a dimensionless quantity with a value not less than 0;

[0117] The controller obtains readability metrics Subsequently, instead of using a fixed threshold, the power supply budget parameters from step one are applied. Heat margin parameters , capture window width And the lane distance segmentation generates the first and second thresholds.

[0118] Preferably, the first threshold is used to limit the saturation risk index. The upper limit of the power supply budget parameter is expressed by the following mapping rule: when the power supply budget parameter is... In the ample zone and with heat reserve parameters When within the safe zone, the first threshold for close-range lanes is tightened to prevent excessive supplemental lighting from turning reflective license plates into blank surfaces; the first threshold for distant lanes is relaxed, but still limited to the boundary that the high-brightness candidate area does not encroach on the license plate candidate area. (Power supply budget parameters...) Tightening or heat reserve parameters As the threshold decreases, the first threshold continues to tighten, preventing step three from selecting a high-lighting scheme at the source. The second threshold is used to limit readability metrics. The lower limit, whose mapping rule is expressed as: when the capture window width When the threshold is narrower, the second threshold is raised towards the safe zone to ensure that the parameter combination solved in step three can directly leave recognizable characters; when the power supply budget parameter Or heat margin parameter When the threshold is lowered, the second threshold is not rigidly increased, but rather linked to lane distance in segments, prioritizing the preservation of sufficient character edges in mid-to-long-distance lanes. In engineering practice, threshold mapping uses segmented lookup tables or bilinear interpolation. If the system is equipped with a front-end recognizer, the controller uses the character confidence sequence output by the recognizer as an additional verification quantity, used only to correct the second threshold, and does not replace the readability index. The main structure.

[0119] Define two mapping operators and one lookup table operator: the first threshold mapping operator. Inputs include lane number, distance segments, and power supply budget parameters. Gear position, heat reserve parameters Gear position, capture window width The gear position outputs the first threshold; the second threshold mapping operator... Input is the same as above, output is the second threshold; candidate item lookup table operator. Inputs include lane number, distance segments, and power supply budget parameters. Gear position, heat reserve parameters Gear position, camera window width The output is a sorted set of exposure parameter candidates and a sorted set of infrared illumination parameter candidates.

[0120] These entries are generated by factory calibration or on-site calibration; during calibration, samples are collected using standard reflective license plate plates, different distance segments, different power supply levels, and different casing temperature ranges; during operation, only minor corrections to the entries are allowed using the offset obtained in step four, and the entire meter is not directly rebuilt; when the entry exceeds the boundary, the nearest neighbor entry is used, and it is not extrapolated to the uncalibrated area.

[0121] The first and second thresholds are obtained from a pre-calibrated threshold mapping table, which is indexed by lane number, distance segment, power supply budget parameter level, heat margin parameter level, and capture window width level. During operation, only the offset value output in step four is allowed to be used to limit the table entries, and extrapolation is not performed on uncalibrated areas.

[0122] When using, readability metrics Visible limits are defined from three directions: character edges, local contrast, and background separation; the first and second thresholds are influenced by power supply budget parameters. Heat margin parameters With the width of the capture window With common constraints, step three can be directly used for executable filtering.

[0123] Step 3: Before the capture window arrives, generate a capture plan for each lane that simultaneously meets the upper limit of saturation, the lower limit of readability, the power supply boundary, and the thermal boundary, and complete the arbitration of supplementary lighting quota when multiple lane windows overlap.

[0124] The saturation risk index and readability index given in step two are a set of mutually constraining observations: lowering the saturation risk index weakens the character contrast of the license plate candidate area, while raising the readability index pushes the license plate candidate area into the blank space. If the joint solution is delayed until within the capture window, the controller will lose the advance of buffer charging, exposure register pre-writing, and pulse sequence loading, and the capture window width will also be reduced. When the width is too narrow, it is impossible to achieve a stable state through multi-frame asymptotic adjustment.

[0125] Therefore, the solution time point is moved forward to the arrival time of the capture window. Previously, the solution results were solidified into single-lane capture plan items, so that step four only required triggering actions according to the items and no longer required making multiple decisions within the window.

[0126] The controller receives the license plate candidate region, highlight candidate region, background candidate region, and saturation risk index output from step two. Readability metrics After setting the first and second thresholds, the power supply budget parameters are used first. With heat margin parameters Establish an upper limit for the adjustable supplementary lighting resources in this lane, and then base the decision on the width of the capture window. The upper limit of the number of pulses allowed within the window for this lane is determined, and then executable combinations are screened from Cartesian combinations of the exposure parameter candidate set and the infrared illumination parameter candidate set. This screening does not rely on continuous capture samples, but rather on the values ​​calculated in step two. and Based on this, a combination of table lookup and one-step correction is used to converge the combination to a single solution. The candidate set of exposure parameters includes at least exposure time, analog gain, and digital gain, and the candidate set of infrared illumination parameters includes at least peak current, pulse width, and pulse count; both types of candidate sets are based on power supply budget parameters. Heat margin parameters The distance to the lane is segmented and stored as an index, and loaded into the runtime memory after the device is powered on.

[0127] First, the candidate set of exposure parameters × the candidate set of infrared illumination parameters are converged into an executable set. Then, the double constraint is converted into a decidable Boolean condition, so that the subsequent correction step only occurs within the executable set.

[0128] The controller arrives at the capture window. The solution thread is started first. The solution thread reads the first and second thresholds for this lane and sets the power supply budget parameters. With heat margin parameters Write to the same entry context. The controller then performs two layers of filtering in the order of candidate entries: the first layer checks whether the fill light duty cycle and controlled charging power corresponding to the candidate entry are within the power supply budget parameters. Within the allowable average power consumption boundary, those that do not meet the requirement are directly eliminated; the second layer of filtering checks whether the combination of peak current, pulse width, and pulse interval of the candidate items affects the thermal margin parameter. Those that do not meet the requirements will be pushed into the reduction zone, and those that do not meet the requirements will be directly removed.

[0129] After completing the executability filtering, the controller performs a double-constraint decision on the remaining candidate entries: the controller uses the current saturation risk index from step two. With readability metrics Starting with [a specific parameter], a threshold margin is calculated once for each candidate item, and candidates with negative threshold margins are removed. The threshold margin uses the same form to avoid screening bias caused by using different scales for two indicators.

[0130]

[0131] Among them, saturation ratio : Indicates saturation risk indicator The degree of exceeding the first threshold is a dimensionless quantity with a value greater than 0; saturation risk index. The saturation risk index output from step two is a dimensionless quantity with a value not less than 0; saturation threshold. : This represents the first threshold output from step two. It is a dimensionless quantity with a value greater than 0, and its function is to limit the saturation risk index. The upper limit;

[0132] Readability ratio : Represents the second threshold relative to readability metrics The degree of gap, a dimensionless quantity with a value greater than 0; readability threshold. : Represents the second threshold output from step two, a dimensionless quantity with a value greater than 0, whose function is to limit the readability index. The lower limit; readability index The readability index output from step two is a dimensionless quantity with a value greater than 0.

[0133] During the screening process, the controller selects those that meet the criteria. and Candidate items are considered to meet the threshold, and those that meet only one of the thresholds are marked as one-sidedly satisfied. To avoid misselection of one-sidedly satisfied items within a short window, the controller prioritizes selecting candidate items from the threshold-satisfied set; when the threshold-satisfied set is empty, the controller adjusts the one-sidedly satisfied set according to the capture window width. Segmented processing: Capture window width When the width is narrow, prioritize lowering. Entries to prevent whiteboards, capture window width When the width is greater, prioritize reducing. Entries are selected to preserve character edges. When used, candidate entries must first satisfy the power budget parameters. With heat margin parameters The feasibility, and then and The monotonic decision-making process uses double constraints to prevent unexecutable entries from entering the subsequent correction phase.

[0134] Based on the obtained candidate set, segmentation is performed according to lane distance and capture window width. The system performs a convergence based on the threat level of the candidate high-brightness areas and writes the results into a capture plan entry. A window index table is pre-stored in the controller's non-volatile memory. The index fields of the window index table include lane number, lane distance segment, and power supply budget parameters. Gear position, heat reserve parameters The gear and the width of the capture window The index entries correspond to a set of exposure parameter candidate sets and an infrared illumination parameter candidate set. When an index field falls between adjacent index levels, the controller generates a sort by interpolating the adjacent index levels.

[0135] The exposure time candidate set consists of discrete exposure levels supported by the image sensor in the current capture mode, and does not exceed the width of the capture window. The analog gain candidate set and digital gain candidate set are composed of discrete settings allowed by the image sensor register; the peak current candidate set, pulse width candidate set, and pulse interval candidate set are composed of discrete settings allowed by the infrared illumination device drive circuit and thermal design; each candidate item must first pass the power supply budget parameters. and heat margin parameters Constraint screening.

[0136] After reading the index entry, the controller first generates an initial value entry from the first and second sorted combinations. This initial value entry is then written to the running capture plan cache. The controller then performs a correction step: when the saturation ratio... When approaching the threshold boundary and the threat level of the highlight candidate region is high, the controller, without changing the sorting position of the exposure parameter candidate set, first lowers the peak current in the infrared illumination parameter by one discrete level, and then raises the digital gain in the exposure parameter by one discrete level, in order to maintain the readability index. Not suddenly depressed; when the ratio is readable When approaching the threshold boundary and the lane distance segment is in the far-distance segment, the controller, without changing the sorting position of the infrared fill light parameter candidate set, first increases the exposure time by one discrete level, and then increases the number of pulses in the infrared fill light parameters by one discrete level, in order to concentrate the effective energy within the capture window. Each level change in this one-step correction re-checks the power supply budget parameters. With heat margin parameters If the boundary is reached, it will revert to the previous gear and adjust another type of parameter, thus maintaining the order of execution first and judgment later.

[0137] For example, at one installation point of the gantry equipment, the controller is located on a metal heat sink plate inside the housing, and the infrared supplementary lighting devices are located on both sides of the optical chamber and are attached to the heat sink plate through thermal pads; when the processor temperature rises, it causes the thermal margin parameter to be affected. When lowering, the controller reaches the capture window at the specified time. First, increase the pulse interval in the infrared illumination parameters, and then increase the exposure time in the same item to improve the readability index of the license plate candidate area. It still falls above the second threshold.

[0138] The direct and visible result of this embodiment is that the infrared illumination device still outputs pulse sequences item by item within the window, the license plate candidate area does not appear as a completely blank area, and the character edges remain continuous. In use, the lookup table path provides a stable initial value, and a one-step correction completes one convergence with discrete level changes, avoiding multiple trials within the window and optimizing the power supply budget parameters. With heat margin parameters Boundary checks are written into the same decision chain.

[0139] When using it, the time reaches the capture window The pre-output capture plan entry ensures that the exposure register is pre-written and the fill light pulse is loaded with a specific target. The entry is also subject to power supply budget parameters. Heat margin parameters The first and second thresholds are used to limit the mismatch caused by temporary rewriting within the window.

[0140] The capture plan entries obtained from solving a single lane can be executed directly when traffic flow is sparse, but in multi-lane concurrent scenarios, the arrival time of the capture window for each lane varies. The infrared illumination parameters of each lane tend to concentrate energy release within their respective windows, resulting in higher electrical and thermal loads during the overlapping time slices. If the controller simply executes multiple entries in parallel, the power supply budget parameters... When the circuit is momentarily filled, the power receiving side protection will be triggered; thermal margin parameters. Being pushed into the disabled zone will trigger the disabling of fill light, both of which will cause subsequent windows to suddenly become ineffective.

[0141] Therefore, items are rearranged and quotas are allocated on the timeline to ensure that the total illuminator usage within the same time slot does not exceed the power supply budget parameter. Then ensure the heat margin parameters It does not enter the restricted area due to continuous overlap. Finally, when it is impossible to satisfy all conditions at the same time, it explicitly enters the degradation mode and records the degradation reason flag, so that step four is executed only as planned and does not temporarily preempt during the execution phase.

[0142] The controller maintains a window event queue, with each queue item corresponding to the arrival time of the capture window for one lane. With the width of the capture window It also carries the generated capture plan entries. The controller uses the capture window arrival time as the reference. The queues are sorted, and an arbitration process is triggered when adjacent queue item windows overlap. The arbitration process divides the overlapping window into several time slices and allocates pulse quotas for the infrared illumination parameters of each lane within each time slice. When the system is configured with an energy buffer unit, the arbitration process uses the available energy state of the energy buffer and the controlled charging power as time slice constraint inputs. After arbitration, the controller outputs a global capture plan, which consists of an illumination trigger sequence arranged by time slices and an exposure register pre-write sequence arranged by lanes, with a predicted derating flag and a degradation mode flag attached to each time slice.

[0143] Transforming the multi-lane overlap problem into an intra-time-slice quota allocation, thus adjusting the power supply budget parameters. With heat margin parameters The supplementary lighting trigger sequence is defined in an executable manner. After detecting an overlapping window, the controller first expands the number of pulses in each lane's capture plan entry into a pulse demand column, which is then sorted according to the arrival time of the capture window. Align with the global timeline. The controller then divides the overlapping area into fixed-width time slices, the width of which is equal to the width of the capture window. The minimum value is determined together with the minimum pulse interval of the infrared illumination device.

[0144] For each time slice, the controller first calculates the power supply contention limit for that time slice. The power supply contention limit is determined by the power supply budget parameters. This is converted to the maximum allowed number of pulses for that time slice. When the system is configured with an energy buffer unit, the controller maps the available energy state of the energy buffer to the maximum additional number of pulses and sets the controlled charging power to zero for that time slice to avoid the superposition of charging and discharging within the same slice. The controller then calculates the thermal contention limit for that time slice, which is determined by the thermal margin parameter. This is converted to the upper limit of the peak current level allowed in that time slot. If the peak current of a lane entry exceeds this upper limit, the pulse demand column for that lane in that time slot will be split into multiple low-peak pulses and re-queued.

[0145] After the quota limit is set, the controller allocates pulse quotas within the time slice in the following order: threshold satisfaction priority, narrower window priority, and longer lane priority. This means that when the readability ratio of a lane... Larger and saturated ratio A lower saturation ratio indicates that the lane requires more energy to maintain the character edges, and the controller prioritizes pulse allocation during that time slice; when the saturation ratio of a lane is lower... A larger value indicates that the lane is more prone to blanking out, so the controller prioritizes pre-writing the exposure register and reduces the pulse quota for that lane. After the pulse quota is determined, the controller generates a supplementary lighting trigger sequence. This sequence is based on a hardware timer, triggering the infrared supplementary lighting device drive channel for the corresponding lane at the start of each time slice, and writing the duty cycle limit and pulse width within the same sequence. In practice, multi-lane overlap is broken down into time slice quotas and power supply budget parameters. With heat margin parameters The maximum number of pulses and the maximum peak level are converted into executable limits to prevent power protection and thermal disablement from being interrupted during the execution phase.

[0146] By addressing boundary conditions and shifting the risk of nearing the peak forward, the global capture plan includes derating actions and degradation paths before execution, avoiding inconsistencies caused by temporary trimming within the window in step four. When generating the trigger sequence for each time slice, the controller simultaneously checks the number of consecutive time slices preceding it and the peak current level allocated to each slice. If the number of consecutive time slices is large and the peak current level remains high for an extended period, the controller sets the predicted derating flag and progressively lowers the peak current level for subsequent time slices, while simultaneously increasing the exposure time in the corresponding lane's exposure parameters for each slice, improving readability indicators. The rate reduction will not cause a sudden drop. If the power supply budget parameters cannot be met simultaneously within a certain overlapping area... Heat margin parameters The controller enters degradation mode and writes the degradation mode flag when the first and second thresholds are met.

[0147] The degradation modes include a license plate preservation priority mode and an uninterrupted flow priority mode: In the license plate preservation priority mode, the controller prioritizes ensuring the readability index of the license plate candidate area. The brightness must not fall below the second threshold, at the cost of reducing the brightness of the background candidate area and compressing the supplementary lighting quota for some lanes; in the non-stop power priority mode, the controller prioritizes ensuring the power supply budget parameters. To avoid triggering protection and ensure continuous output of the network stream, the cost is to compress the exposure time of some lanes to a shorter level and reduce the infrared illumination parameters to a lower level, while writing a degradation reason flag in the capture plan entry.

[0148] As a supplement: the charging control switch and the discharging control switch are controlled by the controller through a mutual exclusion latch; before the capture window begins, the charging control switch is turned off and confirmed to be off, then the discharging control switch is turned on; after the capture window ends, the discharging control switch is turned off first, then the power supply budget parameters are used. Resume charging; if any switch status readback is abnormal, this window will cancel pulse illumination and write an execution error flag.

[0149] For example, at a two-way four-lane intersection, a car with a highly reflective license plate appears in the near lane, while a truck with bright headlights appears in the far lane; the controller captures the image at the moment the capture window arrives. First, the single-lane solution for both lanes is completed. Upon detecting window overlap, arbitration is initiated. During arbitration, the controller prioritizes the threat level of the candidate high-brightness area where the truck's headlights are located, reducing the peak current level and increasing the exposure time for the truck lane, and reducing the number of pulses and increasing the digital gain for the car lane. Then, a predictive derating flag is written into the global capture plan, ensuring that the windows for the subsequent two vehicles are executed at the same pace. Visible actions on-site include the infrared supplementary lighting devices on both sides flashing alternately within the overlapping area, the housing temperature not triggering supplementary lighting disabling, and the license plate candidate area in the captured image maintaining character edges. In practice, predictive derating shifts the thermal boundary risk forward, and the degradation mode explicitly writes unsatisfactory conditions into the plan entries, ensuring consistent actions during execution and retaining verifiable evaluation metrics. Time-slice quotas limit the total supplementary lighting usage within the overlapping window, and predictive derating shifts the thermal boundary risk forward. The degradation mode converts unsatisfactory states into executable entries, allowing step four to be triggered sequentially according to the global capture plan to complete concurrent capture.

[0150] Step 4: Accurately execute the exposure parameters and infrared illumination parameters into the capture window of each lane according to the global capture plan, and perform closed-loop correction on the threshold bias, thermal state estimation and energy availability estimation based on the capture results, thereby forming a cyclic control chain that does not depend on continuous capture samples.

[0151] Step three has provided the capture plan entries for each lane and the supplementary lighting trigger sequence for cross-lane traffic. However, these entries can only truly correspond to the license plate candidate area when the image sensor, infrared supplementary lighting device drive circuit, optional energy buffer unit, and result output link are executed according to a unified time base. If the controller only saves parameters at the software level without specifying the execution order, the capture window arrival time will be... Staying only at the planning level cannot guarantee that exposure register writing, pulse triggering, image buffer latching, and result frame output occur within the same capture window width. Therefore, the planned items are converted into execution actions, and the resulting image data is output in a structured form for subsequent closed-loop correction.

[0152] The controller first reads the arrival times of the most recent capture windows from the global capture plan in chronological order. Then, based on the lane's exposure parameter pre-writing time, infrared illumination device drive trigger time, and image buffer latch time, a window execution chain is generated. This window execution chain is driven by a hardware timer, rather than being temporarily deduced from the arrival interrupt of the preview frame. This is because the preview frame time base is affected by bitstream load and line delay, while the capture action requires a more stable timing reference. In the execution chain, the controller first completes the exposure register pre-writing, then completes the charging and discharging switching of the optional energy buffer unit, and then... The system triggers an infrared fill light pulse and latches the captured image, ultimately outputting the captured result data with lane number, timestamp, degradation mode sign, and predicted depreciation sign.

[0153] Accurately load the exposure parameters selected in step three onto the image sensor so that the capture window reaches the specified time. The corresponding frame actually uses this set of exposure parameters. The controller saves the exposure time, analog gain, and digital gain in each capture plan entry and writes these parameters into the image sensor's pending register area via the image sensor control bus.

[0154] To ensure that the parameters stabilize before the capture window arrives, the controller does not... Instead of writing to the register at a specific time, the write start point is calculated based on the write advance. The write advance is composed of the register loading time, the image sensor line synchronization waiting time, and the drive command propagation time. The controller calculates the write advance time. :

[0155]

[0156] Among them, the pre-write time : Indicates the start time when the controller writes the capture exposure parameters to the image sensor, and the value is earlier than the arrival time of the capture window. Time quantity; time of arrival of the capture window This is the window event time referenced in both steps one and three, and its value is greater than the current control time.

[0157] Register settling time : Represents the time required for the image sensor to stabilize its internal timing after receiving exposure parameters; a value greater than 0 indicates the time required for this process. (Link propagation time) : Represents the time for command issuance and acknowledgment propagation between the controller, drive bus, and image sensor; a value greater than 0 indicates the time elapsed between these two points. After the trigger register is written, it will also check the lane number in the capture plan entry again when the image sensor returns the loading completion sign, so as to avoid the entry being written in the wrong order due to continuous windows of multiple lanes.

[0158] For example, in a camera with infrared supplementary lighting devices arranged on both sides, the controller is located in the middle of the motherboard, and the image sensor is connected to the motherboard via a short line. After the most recent line synchronization is completed, the controller writes the exposure parameters into the register area to be activated. When the target enters the capture window, the new exposure parameters are already in effect, and the brightness of the license plate candidate area in the captured image no longer uses the old parameters from the preview stage.

[0159] When using it, the capture exposure parameters and the capture window arrival time are... To achieve front alignment, frame errors within the window caused by register switching lag are avoided; at the same time, reviewing entries by lane number can suppress parameter cross-referencing when multiple lanes overlap.

[0160] Furthermore, the supplementary lighting trigger sequence from step three is applied to the drive circuit and the optional energy buffer unit, and the captured results are organized into a data structure that can be directly called in subsequent closed-loop operations. After completing the pre-writing of the exposure register, the controller continues to read the infrared supplementary lighting parameters corresponding to the time slice, including peak current, pulse width, pulse count, and pulse interval.

[0161] When an energy buffer unit exists in the system, the controller opens the controlled charging path and then opens the discharge path from the energy buffer unit to the infrared illumination device drive circuit; if no energy buffer unit exists, the controller directly uses the power supply budget parameters. The infrared illumination device is driven by an allowable instantaneous power consumption boundary. To ensure that the illumination energy released within the window does not exceed the time slot quota in step three, the controller calculates the window output energy. :

[0162]

[0163] Among them, the window output energy : Indicates the width of a single capture window The total electrical energy released into the optical channel by the infrared supplementary lighting device is an energy quantity that is not less than 0; forward voltage drop. : Represents the on-state voltage of the infrared supplementary lighting device in the current driving state, and is a voltage value greater than 0; the first pulse current : indicates the first The peak current of the first pulse, taking a value greater than 0; the peak current of the second pulse. pulse width : indicates the first The duration of each pulse, taking a value greater than 0; number of pulses. : Indicates the number of pulses actually executed within the current capture window, a positive integer;

[0164] The controller synchronously activates the capture buffer latch switch during the execution of the supplementary light pulse and outputs the capture result data after the window ends. The capture result data includes at least the captured image, lane number, and the time the capture window arrives. Energy output in the window after execution Degradation pattern indicators and predicted depreciation indicators.

[0165] Preferably, the capture result data also includes exposure parameter readback values ​​and infrared illumination parameter readback values, so as to determine whether the execution layer has deviated.

[0166] In practice, the capture plan entries are transformed into actual exposure writes, fill light pulses, and image latching actions that occur within the window, so that the planning in step three no longer remains at the parameter list level. The capture result data also includes execution parameters and flag fields, facilitating direct backcalculation and correction in the latter half of step four.

[0167] The output of the captured image does not signify the end of the control loop. If the controller does not utilize the captured image result to correct the threshold bias, thermal state estimation, and energy availability estimation in steps two and three, subsequent windows will continue to use the old entries. When encountering changes in power supply level, accumulation of residual heat in the chassis, changes in lane markings, or changes in license plate reflection, the system will gradually deviate from the current on-site state. At the same time, the closed-loop correction cannot revert to a path that relies on continuous captured samples to asymptotically converge; otherwise, it will lose stability during prolonged periods of no traffic or sparse traffic.

[0168] Therefore, by adopting a single-result back-calculation, parameter-level correction, and exiting upon anomaly, the bias and estimate are updated only once, and the captured image is not used as a control prerequisite that the next frame must rely on.

[0169] After receiving the captured image data, the controller first recalculates the saturation risk index on the captured image according to the license plate candidate region defined in step two. With readability metrics However, the images used here are captured images rather than preview frames, so they can more accurately reflect the final execution effect.

[0170] Subsequently, the controller compares the recalculated post-capture saturation risk index and post-capture readability index with the first and second thresholds referenced before step three, respectively, to generate a threshold bias correction amount; simultaneously, the controller also adjusts the output energy based on the window. The controlled charging power and the current pulse execution record are used to correct the available energy state of the energy buffer, and the thermal state estimate is corrected based on the latest sampling of the supplementary lighting device temperature, housing temperature, and processor temperature. After the correction is completed, the controller determines whether the exit condition is met. If it is, it returns to step one and waits for the next target; if the power supply or thermal state is found to have entered the unrecoverable zone, it switches to safe mode and stops the supplementary lighting output.

[0171] Furthermore, the dual-indicator definition from step two is directly integrated, and the capture result is transformed into the threshold bias for the next capture. The controller relocates the license plate candidate area in the captured image, using the area mapping table from step two and the lane number of this capture for localization, without performing a full-frame search. The controller then obtains the post-capture saturation risk index. Readability index after snapshot The threshold bias is then adjusted based on the deviation between the two values ​​and the threshold value.

[0172]

[0173] Among them, the updated saturation bias : Represents the saturation bias amount used for the next round of first threshold correction, with a value that is a bounded real number; original saturation bias : Represents the currently effective saturation bias, with a value that is a bounded real quantity; saturation correction coefficient : Indicates the proportion of saturation risk deviation written to the bias after capture, with a value greater than 0 and less than 1;

[0174] Saturation risk index after snapshot : Represents the saturation risk index recalculated on the captured image, a dimensionless quantity with a value not less than 0; saturation threshold : Represents the first threshold called in steps two and three, a dimensionless quantity with a value greater than 0; the updated bias is readable. : Represents the readable bias amount used for the next round of second threshold correction, and its value is a bounded real number;

[0175] Raw readable bias : Represents the currently effective readable bias, with a value of a bounded real number; readable correction coefficient : Represents the proportion of readability deviation written to the bias after capture, with a value greater than 0 and less than 1; Readability threshold : Represents the second threshold called in steps two and three, a dimensionless quantity with a value greater than 0; readability index after snapshot. : Represents the readability index recalculated on the captured image, a dimensionless quantity with a value greater than 0;

[0176] Furthermore, the updated saturation bias and updated readable bias All will be limited to the set offset range; when the degradation mode flag is continuously prioritized mode and the execution anomaly flag is valid, this window will not update the threshold offset; when there are no capture results for several consecutive weeks, the offset will return to the factory baseline according to the decay rule; when the lane number changes, the offset is stored by lane and is not shared across lanes.

[0177] After the correction is complete, the controller does not immediately recalculate the current window, but instead applies the updated saturation bias. With updated readable bias Write the threshold mapping table to be called in the next round of step two.

[0178] For example, when the capture result of a certain lane shows that the edge of the license plate candidate area has been slightly whitened, but the characters are still readable, the controller will increase the updated saturation bias. Suppressing direction while maintaining readable bias after update The basic structure remains unchanged, causing the first threshold to be more compact in the next round while the second threshold remains within the original range.

[0179] Define the coordinate mapping operator from preview to snapshot. ; The calibration is obtained from two operating modes of the same image sensor; if the preview stream and the capture stream come from different readout windows of the same sensor, then... It consists of scaling factor and crop offset; if the preview stream and the snapshot stream come from different sensors, then The location of the license plate candidate region in step four is obtained from the output region in step two via a dual-camera calibration matrix. The image is mapped, and then local fine-tuning is performed on the captured image.

[0180] As a supplement: the controller is also configured to establish a coordinate mapping operator between the preview image coordinates and the captured image coordinates. When the preview video stream and the captured image come from different readout modes of the same image sensor, the coordinate mapping operator... It consists of scaling factors and cropping offsets; when from different imaging units, the coordinate mapping operator It consists of a camera calibration matrix; the license plate candidate area in step four is obtained by mapping the output area in step two to the captured image through a coordinate mapping operator.

[0181] When in use, the capture results are directly written back as the threshold bias, without relying on continuous sample accumulation to take effect; at the same time, the saturation bias and readable bias are corrected separately, which can avoid a single error from pulling the two constraint directions at the same time.

[0182] Furthermore, the impact of this execution on the energy buffer unit and thermal state is written back to step one, and it is determined whether the system returns to the loop or enters safe mode. After completing the threshold bias correction, the controller continues to read the temperature of the supplementary lighting device, the housing temperature, and the processor temperature after this capture, and combines this with the windowed output energy... Update the available energy state and thermal state estimates for the next round.

[0183] If the system is equipped with an energy buffer unit, the controller updates the available energy state based on the current discharge energy and the allowed charging time after the window. :

[0184]

[0185] Among them, the updated available energy state : Indicates the available energy state of the energy buffer for the next round of step one, with a value not less than 0; energy limit. : Represents the maximum energy that the energy buffer unit is allowed to store under the current circuit limitations, with a value greater than 0; the original available energy state. : Indicates the available energy status of the energy buffer recorded in step one before this capture, with a value not less than 0;

[0186] Window output energy Energy quantity with a value not less than 0; controlled charging power : Indicates the upper limit of charging power jointly defined by steps one and three, with a value not less than 0; recovery time This represents the charging time from the end of the current window to the next allowed discharge, and its value is not less than 0. Its function is to control the charging power. This is converted into replenished energy.

[0187] Regarding thermal status updates, the controller writes the temperatures of the supplementary lighting device, the casing, and the processor after the snapshot back to the thermal status cache from step one, and simultaneously checks for two types of exit scenarios: the first is a normal exit scenario, where the snapshot result has been output, the target has left the effective snapshot area, and the power supply budget parameters are within acceptable limits. With heat margin parameters If the system has not entered the unrecoverable zone, the controller will terminate the current target process and return to step one. The second type is a safe exit scenario, where power supply capacity information is lost and the default budget cannot meet the basic reserved power consumption, or temperature sampling indicates that the thermal margin parameter is insufficient. The system has entered the infrared illumination disabled zone. At this time, the controller stops the output of the infrared illumination device, and only retains the preview video stream and the result reporting link, waiting for the power supply or thermal status to be restored.

[0188] As a supplement: Ethernet power control messages that have not been updated for several consecutive cycles are considered unavailable; power supply level field verification failures and power level field invalidations are considered untrustworthy; if the current power level jumps more than one preset level compared to the previous valid power level without a renegotiation completion flag, it is considered untrustworthy; in the case of untrustworthiness, the power supply budget parameters will be... Set the minimum safe supplemental lighting budget above the base reserved power.

[0189] In practice, the threshold system in step two is reconnected with the execution results in step four by back-calculating the indicators after the snapshot is captured; the input state in step one is rewritten to the latest field state by updating the energy and thermal state. This forms a parameter-level closed loop, rather than an asymptotic compensation path that relies on continuously captured samples.

[0190] Please refer to Figures 1-6. This invention provides a nighttime license plate capture camera with adaptive exposure and glare suppression, including a camera body, an infrared filler, a temperature acquisition unit, a power supply interface, a controller, and a memory. The memory stores a program executable by the controller. When the controller executes the program, it implements: a power supply and thermal status acquisition and capture window prediction module, used to acquire power supply capacity information and temperature parameters, form power supply budget parameters and thermal margin parameters, and output the arrival time and width of the lane capture window on the preview video stream; and a candidate region analysis and threshold determination module, used to determine the license plate candidate region, calculate the saturation risk index and readability index, and based on the power supply... The power budget parameters and thermal margin parameters determine the first and second thresholds; the exposure and supplementary lighting joint planning and arbitration module is used to jointly determine the exposure parameters and infrared supplementary lighting parameters before the capture window arrives, so that the saturation risk index does not exceed the first threshold and the readability index is not lower than the second threshold, and meets the power budget parameters and thermal margin parameters. When the lane capture windows overlap, a capture plan with supplementary lighting arbitration is generated; the timing write trigger and closed-loop update module is used to write the exposure parameters before the capture window arrives, and trigger infrared supplementary lighting to acquire capture images according to the capture plan within the window. Based on the capture images, the threshold adaptive bias parameters and thermal prediction or energy availability estimation are updated.

[0191] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0192] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0193] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0194] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0195] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A nighttime license plate capture system with adaptive exposure and glare suppression, characterized in that: This includes acquiring power supply capacity information and temperature parameters, forming power supply budget parameters and heat margin parameters, and outputting the arrival time and width of the lane capture window on the preview video stream; The candidate license plate area is determined, the saturation risk index and readability index are calculated, and the first threshold and the second threshold are determined based on the power supply budget parameter and the heat margin parameter. Before the capture window arrives, the exposure parameter and the infrared supplementary light parameter are jointly determined so that the saturation risk index does not exceed the first threshold and the readability index is not lower than the second threshold, and the power supply budget parameter and the heat margin parameter are satisfied. When the lane capture windows overlap, a capture plan with supplementary light arbitration is generated. Before the capture window arrives, the exposure parameters are written, and the infrared supplementary light is triggered to capture the image within the window according to the capture plan. The threshold adaptive bias parameters and thermal prediction or energy availability estimation are updated based on the captured image.

2. The nighttime license plate capture system according to claim 1, characterized in that: The power supply capacity information and temperature parameters are validated. If the validation fails, the conservative default power supply budget parameters and heat margin parameters are used. The duty cycle, peak value and pulse count in the infrared supplementary light parameters are limited. The abnormal flag is associated with the corresponding lane number, stored and written into the capture plan.

3. The nighttime license plate capture system according to claim 2, characterized in that: Vehicle detection and tracking are performed on the preview video stream to obtain target trajectory information. Based on the target trajectory information and the preset trigger line position, the arrival time and width of the capture window are extrapolated. Window event identifiers are generated according to lane number and output to the event scheduling queue.

4. The nighttime license plate capture system according to claim 3, characterized in that: In the preview video stream, the license plate candidate region is determined based on the window event identifier, and the highlight candidate region and background candidate region are segmented around the license plate candidate region. The highlight candidate region is generated by the connected bright region, and the background candidate region avoids the highlight candidate region.

5. The nighttime license plate capture system according to claim 4, characterized in that: A saturation risk index is generated based on the proportion of saturated pixels and the proportion of saturated connected regions within the license plate candidate area. A readability index is generated based on the gradient energy and local contrast within the license plate candidate area. The contribution of the highlight candidate area to the saturation risk index is also incorporated into the saturation risk index.

6. The nighttime license plate capture system according to claim 5, characterized in that: Based on power supply budget parameters, heat margin parameters, and lane distance segmentation, a first threshold and a second threshold are determined for each lane. The capture window width is used as the input for threshold mapping to form a threshold set, and then a structured result of lane-index-threshold is output.

7. The nighttime license plate capture system according to claim 6, characterized in that: Discrete combinations of exposure parameter candidate sets and infrared illumination parameter candidate sets are constructed. After filtering according to the first threshold, the second threshold, power supply budget parameters, and thermal margin parameters, the highest priority combination is selected. The capture plan entries are generated by looking up a table and performing a one-step correction.

8. The nighttime license plate capture system according to claim 7, characterized in that: When overlapping capture windows of two lanes are detected, the controller divides the overlapping time into time slices and performs quota arbitration on the number of pulses and peak values ​​in the infrared supplementary light parameters of each lane, so that the total supplementary light usage in any time slice does not exceed the power supply budget parameter and is limited by the heat margin parameter. When the heat margin parameter is close to the disable threshold, a prediction derating flag is written.

9. The nighttime license plate capture system according to claim 8, characterized in that: When no parameter combination simultaneously satisfies the first threshold, the second threshold, the power supply budget parameter, and the heat margin parameter exists, the controller enters a preset degradation mode. The degradation modes include a license plate priority mode and a non-disruptive power supply priority mode. A degradation mode flag and a degradation reason flag are written into the capture plan. When the system is configured with an energy buffer unit, the controller charges the energy buffer unit in a controlled manner under the condition of not exceeding the power supply budget parameter, and outputs pulse supplementary light energy within the capture window. The output is driven by the arrival time of the capture window. The controller also calculates saturation risk index and readability index for the license plate candidate area of ​​the captured image to update the threshold adaptive bias parameter and update the thermal prediction and energy availability estimate.

10. A nighttime license plate capture camera with adaptive exposure and glare suppression, comprising a camera body, an infrared illuminator, a temperature acquisition unit, a power supply interface, a controller, and a memory, wherein the memory stores a program executable by the controller, characterized in that... When the controller executes the program, it implements the following module: power supply and thermal status acquisition and capture window prediction module, which is used to acquire power supply capacity information and temperature parameters, form power supply budget parameters and thermal margin parameters, and output the arrival time and width of the lane capture window on the preview video stream. The candidate region analysis and threshold determination module is used to determine the candidate license plate region, calculate the saturation risk index and readability index, and determine the first threshold and the second threshold based on the power supply budget parameter and the heat margin parameter. The exposure and supplementary lighting joint planning and arbitration module is used to jointly determine the exposure parameters and infrared supplementary lighting parameters before the capture window arrives, so that the saturation risk index does not exceed the first threshold and the readability index is not lower than the second threshold, and meets the power supply budget parameter and heat margin parameter constraints, and generates a capture plan for supplementary lighting arbitration when the lane capture windows overlap. The timing write trigger and closed-loop update module is used to write the exposure parameters before the capture window arrives, and trigger the infrared supplementary lighting to acquire the capture image according to the capture plan within the window, and update the threshold adaptive bias parameter and thermal prediction or energy availability estimate based on the capture image.

Images