A stage lighting real-time control method and system based on semantic mapping and mixed constraints

Abstract

The application provides a stage lighting real-time control method and system based on semantic mapping and mixed constraints, relates to the fields of stage performance control and digital lighting technology, and comprises the following steps: acquiring MIDI beat time stamps and semantic prompts associated with audio; and according to the semantic prompts, determining a single industrial color card chroma anchor point through deterministic table lookup or keyword matching retrieval in a pre-defined industrial color card library. The application has the following advantages: through deterministic mapping from semantics to industrial color cards and color difference envelope projection, the standard adaptability of color generation is improved; in combination with a kinematic causal limiting mechanism, equipment over-limit operation can be prevented to a certain extent, and mechanical safety is enhanced; through the adoption of limited disturbance and soft saturation strategies, light dynamic changes are smoother and more natural, and visual perception is improved; meanwhile, through the use of a hierarchical time sequence architecture, audio beat accurate alignment is realized, mainstream lighting protocols are compatible, and the engineering feasibility of the system is improved.

Description

Technical Field

[0001] This invention relates to the field of stage performance control and digital lighting technology, and in particular to a real-time stage lighting control method and system based on semantic mapping and hybrid constraints. Background Technology

[0002] The manual pre-wiring and rehearsal calibration console programming method. In stage performance practice, lighting technicians usually manually write cue cards, scenes, and timings on the console during the rehearsal stage, and then calibrate the channel levels and lighting positions during on-site rehearsals. The technical drawback of this method is that it is highly dependent on rehearsals and manual calibration, and it is difficult to automatically realign the control timing with the audio beat under conditions of zero rehearsal or dynamic changes in content.

[0003] Automatic lighting triggering based on fixed rules or a preset scene library. Some systems trigger brightness / color / position changes based on beat detection, volume thresholds, or pre-stored scene tables to reduce the burden of manual filming. However, their technical shortcomings include: color generation lacks semantic anchoring from industrial color charts and implementable color difference envelope constraints; the motion trajectory at inflection points can easily exceed the angular velocity and angular acceleration capabilities of the lighting fixtures, thus posing a risk of exceeding limits or mechanical impact.

[0004] Current Status of Standards and Protocols. Regarding the physical layer and network distribution, DMX512-A (ANSI E1.11) specifies a single-universe serial link and its frame structure; in engineering practice, the single-universe refresh rate typically does not exceed approximately 44Hz. ANSI E1.31 (sACN) and Art-Net™ (a trademark of Artistic Licence) provide Ethernet-based multi-universe distribution capabilities. GDTF / DIN SPEC 15800 provides device descriptions for luminaire capabilities and channel layout.

[0005] Its technical shortcomings are that the above standard only specifies the transmission frame format, network distribution and device capability description, and does not provide a deterministic mapping from semantic prompts to the chromaticity anchor point of the industrial color card, nor does it provide a mechanism for maintaining the color difference envelope projection and kinematic constraints after disturbance.

[0006] Therefore, there is a need for a real-time control method that can perform semantic anchoring, chromatic aberration projection, control frame grid alignment and kinematic limiting within a hard constraint envelope, while allowing limited perturbations to improve visual smoothness. Summary of the Invention

[0007] The purpose of this invention is to provide a real-time stage lighting control method and system based on semantic mapping and hybrid constraints. This solves the technical problems of existing stage lighting control systems in the prior art. Under conditions of zero rehearsal or dynamic content changes, the lack of deterministic mapping from semantics to industrial color cards and kinematic hard constraint mechanisms leads to unanchored color generation, easy movement trajectory overshooting causing mechanical impact, and inability to maintain the alignment of color difference envelope with control frame grid after disturbance.

[0008] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0009] A real-time stage lighting control method based on semantic mapping and hybrid constraints includes the following steps:

[0010] S1: Obtain the MIDI beat timestamp and semantic cues associated with the audio;

[0011] S2: Based on the semantic prompts, a single industrial color card chromaticity anchor point is obtained in the predefined industrial color card library through deterministic table lookup or keyword matching retrieval;

[0012] S3: Generate candidate color mixing channel values ​​in the neighborhood of the anchor point, wherein the neighborhood is a region centered on the anchor point with an Euclidean radius in the CIELab space not exceeding a preset neighborhood radius;

[0013] The candidate values ​​are then subjected to CIEDE2000 color difference envelope projection, which includes: transforming the candidate values ​​to the CIELab space; and weighting them accordingly. The color difference between the CIEDE2000 light source (D65) and the standard observer at a 2° angle relative to the anchor point is calculated. If the color difference does not exceed a preset threshold, the candidate value is output; otherwise, the interpolation parameter is used. At the anchor point With candidate points Construct line segments between 1 ,right Perform a one-dimensional binary search until the search interval length is less than a preset tolerance, and then obtain the feasible results that satisfy the threshold. For the corresponding point, after converting it to sRGB, performing color gamut clamping and 8-bit quantization, the color difference between the quantized result and the anchor point is calculated back. If the color difference still exceeds the threshold, then... After multiplying by a preset shrinkage factor, repeat the above transformation, clamping, quantization, and back-substitution verification; when the number of repetitions reaches the preset positive integer upper limit... At that time, the output is the anchor point itself, which is clamped by color gamut and quantized by 8 bits;

[0014] S4: At the first frame rate The control frame grid for the MIDI timestamp Perform quantization and retrieve frame index. And place the peak sample of the light trigger envelope in the first... One control frame;

[0015] S5: Parse from the equipment capability description file or use the preset upper limits of angular velocity and angular acceleration, and perform causal limiting on the azimuth and pitch trajectories of the luminaire with azimuth and pitch degrees of freedom;

[0016] S6: Color difference margin based on the current control quantity The upper bound of the disturbance amplitude of the current control frame is determined by a preset proportional coefficient based on the kinematic margin relative to the upper limits of angular velocity and angular acceleration. ,in, , , The preset positive coefficient is used; the amplitude does not exceed Within the range, a restricted disturbance is superimposed, and the superimposed result is subjected to the color difference envelope projection described in S3 and the causal limiting described in S5 again before outputting a control quantity; the restricted disturbance includes at least one of polynomial easing interpolation and band-limited random disturbance;

[0017] S7: Maps brightness, color, and orientation parameters to multi-channel light control frames and outputs them; the internal control layer uses the first frame rate. Generate control values, and the physical output layer operates at a second frame rate not exceeding 40Hz. 2. Send; the control amount for the second frame rate is obtained by one of the following methods: by integer decimation ratio (M is a positive integer) The first frame rate control flow is downsampled, or the first frame rate control flow is zero-order held sampled at each transmission time of the second frame rate, or it is distributed to the physical layer through the network light protocol multiverse distribution, wherein the physical layer transmission frame rate of each universe is not higher than the second frame rate.

[0018] As an improvement, the deterministic table lookup or keyword matching includes: performing precise table matching after normalizing the semantic prompt string; if no match is found, selecting the entry with the highest score based on the number of keyword matches for the color card entry; when multiple entries have the same number of matches, selecting the first one in the color card library entry order; when all matches are zero, returning the first entry in the color card library; the color card library contains at least one industrial color filter entry with a CIELab anchor point of (45.0, −5.0, −25.0); the color difference determination is calculated only for the color mixing channel values ​​(R, G, B), and the luminance channel is not involved; each entry in the color card library has a predefined fixed index order.

[0019] As an improvement, the color difference envelope projection is a feasibility binary search and back-substitution verification along the ray, rather than an exact projection or minimum shrinkage optimization of the CIEDE2000 non-convex set; the preset color difference threshold is 8.0, according to... D65, 2° observer calculation; the upper limit of the number of outer iterations K is 5, and the bisection tolerance is 1×10. -3 The shrinkage factor is 0.5 and the neighborhood radius is 8.0; the anchor point is located within the sRGB color gamut.

[0020] As an improvement, the polynomial easing interpolation is cubic. Or C 2 The continuous polynomial easing is used, wherein the frequency of the band-limited random perturbation is between 0.5 Hz and 2 Hz; the band-limited random perturbation is obtained by filtering white noise through a second-order Butterworth bandpass filter, or by superimposing multiple sinusoidal components in the range of 0.5 Hz to 2 Hz.

[0021] As an improvement, the upper limit of the angular velocity is 180° / s, and the upper limit of the angular acceleration is 720° / s. 2 The lower limit of the luminance channel normalization is 0.05; in the multi-channel mapping, the first channel is luminance, the second to fourth channels are RGB, the fifth and sixth channels are Pan / Tilt coarse, and the remaining channels are reserved; when used for DMX512-A physical layer transmission, the transmission frame rate is no higher than 40Hz; when transmitted at 40Hz in zero-order hold sampling mode, the upper limit of the time quantization at the transmitting end is 12.5ms.

[0022] As an improvement, the Pan / Tilt consists of a 16-bit mapping between a coarse adjustment channel and a fine adjustment channel.

[0023] As an improvement, the first frame rate is 100 frames per second; the second frame rate is 25Hz and the integer decimation ratio M=4, in which case the upper bound of the time quantization of the physical output layer is 20ms; or the second frame rate is 40Hz and zero-order hold sampling is used, in which case the upper bound of the time quantization of the physical output layer is 12.5ms.

[0024] As an improvement, the upper bound of the disturbance amplitude in S6 is smoothed by a first-order exponential moving average before use, and the restricted disturbance is... The soft saturation method is applied, where This is to disturb the original amplitude.

[0025] A real-time stage lighting control system based on semantic mapping and hybrid constraints includes: an input module for acquiring MIDI beat timestamps and semantic cues associated with audio;

[0026] The semantic retrieval module is used to perform the deterministic table lookup or keyword matching described in S2 of claim 1;

[0027] A color difference projection module is used to perform the CIEDE2000 color difference envelope projection as described in claim 1, S3.

[0028] The timing quantization module is used to perform frame raster quantization and peak setting as described in claim 1, S4;

[0029] A kinematic limiting module for performing causal limiting as described in claim 1, S5;

[0030] A constrained perturbation module is used to perform the margin constraint perturbation and reprojection re-limiting as described in S6 of claim 1;

[0031] The channel mapping and output module is used to perform the multi-channel mapping, first frame rate generation and second frame rate transmission not exceeding 40Hz as described in claim 1, S7.

[0032] The specific limitations of each module are consistent with the corresponding steps described in claims 1 to 8.

[0033] The beneficial effects of this invention are: improved certainty of color generation and better compatibility with industry standards.

[0034] By establishing a deterministic mapping mechanism (such as table lookup or keyword matching) from semantic prompts to a predefined industrial color swatch library, and combining it with the CIEDE2000 color difference envelope projection algorithm, this method can constrain the generated colors within the chromaticity anchor point neighborhood that conforms to industry standards while retaining the flexibility of natural language description. This helps reduce the randomness of color generation, making the output results more consistent with the physical color rendering capabilities of actual stage lighting equipment and industry standards.

[0035] Enhanced the safety and mechanical protection capabilities of equipment movement.

[0036] By introducing a kinematic causal limiting mechanism based on equipment capability descriptions (such as upper limits of angular velocity and angular acceleration), the movement trajectory of the lighting fixtures can be prevented from exceeding physical limits during real-time control. Compared to traditional methods that lack hard constraints, this approach helps reduce the risk of mechanical shock caused by commands exceeding limits, thereby extending the service life of the equipment and improving performance safety to some extent.

[0037] Improved the smoothness and naturalness of dynamic visuals.

[0038] By employing a constrained perturbation strategy (including polynomial easing interpolation and band-limited random perturbation) and a soft saturation application method, this approach avoids the mechanical abruptness and visual stiffness that may result from traditional linear interpolation or hard clipping, while satisfying hard constraints on color difference and kinematics. Experimental data show that this method, while maintaining compliance with constraints, reduces the accelerometer squared time integral by approximately 3.82 × 10³ times, resulting in smoother and more natural light movement and color changes.

[0039] Achieved high-precision timing alignment and protocol compatibility

[0040] By employing a layered architecture of high-frequency frame raster quantization in the internal control layer and downsampling / zero-order hold sampling in the physical output layer, the system ensures precise alignment between the light trigger moment and the audio MIDI beat, while also guaranteeing that the output signal is compatible with the frame rate limits of mainstream lighting protocols such as DMX512-A (e.g., no higher than 40Hz). This design helps improve the accuracy of audio-visual synchronization and the engineering feasibility of the system without breaking through existing hardware transmission bottlenecks. Attached Figure Description

[0041] Figure 1 This is a comparison diagram of the timing alignment of the three methods in the control layer in the embodiments of the present invention;

[0042] Figure 2 This is a color envelope comparison diagram of the three methods in the embodiments of the present invention;

[0043] Figure 3 This is a comparison diagram of Pan's angular velocity and absolute jerk in an embodiment of the present invention;

[0044] Figure 4 This is a flowchart illustrating the method of the present invention. Detailed Implementation

[0045] To make the content of this invention easier to understand, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Identical components are indicated by the same reference numerals.

[0046] like Figure 4 As shown, the present invention provides a real-time stage lighting control method based on semantic mapping and hybrid constraints, comprising the following steps S1 to S7. The steps are described in detail below with reference to specific embodiments.

[0047] I. Semantic Retrieval (S1~S2)

[0048] S1: Get the MIDI beat timestamp and semantic cues associated with the audio.

[0049] In this embodiment, the system acquires a MIDI beat timestamp sequence associated with the audio content and corresponding semantic cue text through an audio analysis module or external input. The MIDI beat timestamps are used to determine the timing anchor point for light triggering, and the semantic cue text is used to determine the color style of the light.

[0050] As an alternative implementation, the semantic prompts can be natural language descriptions extracted from the metadata of the audio file, lyrics text, or external control signals, such as "Desolate / Sadness" or "Warm / Cheerful".

[0051] S2: Based on the semantic prompts, a specific single industrial color card chromaticity anchor point is obtained in the predefined industrial color card library through deterministic table lookup or keyword matching retrieval.

[0052] In this embodiment, the predefined industrial color chart library contains at least one industrial filter entry. As a specific implementation, the color chart library includes the entry for industrial filter #117 (a registered trademark of Lee Filters®), with a CIELab anchor point of (45.0, −5.0, −25.0). This anchor point is located within the sRGB color gamut and can be used for lossless round trips. The color chart library may also include subsequent entries such as industrial filter #119 (CIELab anchor point (28.0, 8.0, −42.0)) and #022 (CIELab anchor point (55.0, 25.0, 45.0)), for keyword matching.

[0053] The specific steps of the deterministic table lookup or keyword matching include:

[0054] (1) After standardizing the semantic prompt string (such as removing punctuation, unifying capitalization, stemming, etc.), perform exact table matching; if a match is found, return the corresponding entry directly;

[0055] (2) If an exact match is not found, calculate the number of keyword matches for each color card entry and select the entry with the highest number of matches.

[0056] (3) When multiple entries have the same number of hits, the first one is selected according to the predefined fixed index order of the color card library entries;

[0057] (4) When the total number of hits is zero, return to the first entry in the color card library (i.e., the above-mentioned industrial filter #117 anchor point).

[0058] In this embodiment, the semantic prompt "Desolate / Sadness" matches industrial color filter #117 after the above retrieval process, resulting in the chromaticity anchor point Lab(45.0, −5.0, −25.0). It should be noted that this embodiment uses single anchor point verification. The VAD (Valence-Arousal-Dominance) interface is reserved but not enabled. The VAD module can be enabled as needed to support richer semantic input.

[0059] II. Color Difference Envelope Projection (S3)

[0060] S3: Generate candidate color mixing channel values ​​in the neighborhood of the anchor point, and perform CIEDE2000 color difference envelope projection on the candidate values.

[0061] The neighborhood is the region centered at the anchor point in the CIELab space whose Euclidean radius does not exceed a preset neighborhood radius. In this embodiment, the neighborhood radius is 8.0.

[0062] The candidate color mixing channel values ​​are generated by sampling candidate points in the neighborhood of the anchor point according to a Gaussian distribution. Specifically, in this embodiment, independent Gaussian noise is superimposed on the three dimensions L*, a*, and b*, with standard deviations of respectively. (Originated by multiplying entropy_ratio = 0.20 by the standard deviations of each Wikipedia entry: 2.0, 1.8, and 2.2), 3 candidate points The range is approximately 1.2 to 1.3, which is much smaller than the neighborhood radius of 8.0, therefore the candidate point always falls within the neighborhood.

[0063] The specific steps of the CIEDE2000 color difference envelope projection include:

[0064] (1) Convert the candidate RGB values ​​to the CIELab space;

[0065] (2) Calculate the CIEDE2000 color difference relative to the anchor point using the weights kL:kC:kH = 1:1:1, D65 standard light source and 2° standard observer;

[0066] (3) If the color difference does not exceed the preset threshold (the preset color difference threshold is 8.0 in this embodiment), the candidate value is directly output;

[0067] (4) Otherwise, use interpolation parameters At the anchor point With candidate points Construct line segments between ,right Perform a one-dimensional binary search until the length of the search interval is less than the preset tolerance (in this embodiment, the binary tolerance is 1×10⁻³), and take the point corresponding to the feasible λ that satisfies the threshold.

[0068] (5) After converting the point to sRGB, performing color gamut clamping and 8-bit quantization, back-substitute the quantization result and calculate the color difference between the anchor point;

[0069] (6) If the color difference after substitution still exceeds the threshold, then multiply λ by a preset shrinkage factor (the shrinkage factor is 0.5 in this embodiment) and repeat the above conversion, clamping, quantization and substitution verification.

[0070] (7) When the number of repetitions reaches the preset positive integer upper limit K (K=5 in this embodiment), the anchor point itself is output after color gamut clamping and 8-bit quantization.

[0071] It is important to emphasize that the color difference envelope projection of this invention is a feasibility binary search and back-substitution verification along the ray, rather than an exact projection or minimum shrinkage optimization of the CIEDE2000 non-convex set. This design effectively reduces computational complexity and meets the requirements of real-time control. The anchor point is located within the sRGB color gamut, so when the outer iteration reaches the upper limit, the output anchor point itself does not violate the color difference threshold constraint.

[0072] In this embodiment, the anchor point Lab(45.0, −5.0, −25.0), after sRGB conversion, has a floating-point RGB value of approximately (62.9433, 110.9510, 147.7362), which is (63, 111, 148) after 8-bit quantization. The CIEDE2000 color difference calculated by substituting the anchor point itself after 8-bit quantization is approximately 0.0703, which is much smaller than the threshold of 8.0.

[0073] III. Control Frame Raster Quantization (S4)

[0074] S4: Quantize the MIDI timestamp t using the control frame grid of the first frame rate f1, take the frame index n = round(t·f1), and place the peak sample of the light trigger envelope in the nth control frame.

[0075] In this embodiment, the first frame rate f1 = 100 frames / second (100FPS). For each MIDI beat timestamp t, the corresponding control frame index n = round(t × 100) is calculated, and the peak sample of the light trigger envelope (such as the brightness envelope or color switching envelope) is precisely placed in the nth control frame.

[0076] This design ensures precise alignment between the light triggering moment and the music beat. In this embodiment, with BPM=120, the beat interval is 500ms, and the frame grid resolution of 100FPS is 10ms. The beat moment falls exactly on the integer control frame, so the root mean square error (RMSE) of the control layer timing alignment is 0.0000ms.

[0077] IV. Kinematic Causal Limitation (S5)

[0078] S5: Parse from the device capability description file or use the preset upper limits of angular velocity and angular acceleration, and perform causal limiting on the azimuth and pitch trajectories of the luminaire with degrees of freedom of azimuth and pitch.

[0079] In this embodiment, a preset upper limit for angular velocity ω_max = 180° / s and an upper limit for angular acceleration α_max = 720° / s² are used. These upper limits for angular velocity and angular acceleration can also be obtained from a device capability description file in GDTF (General Device Type Format) format.

[0080] For the Pan (azimuth) and Tilt (tilt) trajectories of each control frame, causal limiting is performed—that is, the output of the current frame is limited only based on the current and historical states, without relying on information from future frames, in order to ensure the causality of real-time processing.

[0081] V. Constrained Disturbances and Constraint Preservation (S6)

[0082] S6: Determine the upper limit of the disturbance amplitude based on the color difference margin and kinematic margin of the current control quantity, superimpose the restricted disturbance within the range where the amplitude does not exceed the upper limit, and output the control quantity after performing color difference envelope projection and causal limiting on the superposition result again.

[0083] Specifically, the upper bound A_max of the disturbance amplitude in the current control frame is determined as follows:

[0084]

[0085] in, The preset color difference threshold is 8.0 in this embodiment. The current control variable is the CIEDE2000 color difference, ω and α are the absolute values ​​of the current angular velocity and angular acceleration, respectively, and kc, kω, and kα are preset positive coefficients. In this embodiment, kc = 0.07, kω = 0.0031, and kα = 0.00078.

[0086] The upper bound of the disturbance amplitude A_max is used after smoothing by a first-order exponential moving average (EMA), with an EMA coefficient α_A = 0.05.

[0087] The constrained perturbation includes at least one of polynomial easing interpolation and band-limited random perturbation. In this embodiment:

[0088] (1) The polynomial easing interpolation is a cubic Bézier easing, with four control points P0 = (0,0), P1 = (0.42,0), P2 = (0.58,1), and P3 = (1,1), which are evaluated using cubic Bernstein basis functions;

[0089] (2) The frequency of the band-limited random disturbance is between 0.5 Hz and 2 Hz, specifically obtained by superimposing multiple sinusoidal components (the Pan channel contains 0.95 Hz and 2.0 Hz components, and the Tilt channel contains an additional 1.7 Hz component), and is superimposed with white noise smoothed by EMA (α=0.05) (Pan coefficient 0.04, Tilt coefficient 0.03), with a nominal amplitude scale amp = 0.56°. As another optional implementation, the band-limited random disturbance can also be obtained by filtering white noise through a second-order Butterworth bandpass filter.

[0090] The restricted perturbation is applied in a soft saturation manner:

[0091] p_clipped = A_max,used · tanh(p / A_max,used)

[0092] Where p is the original amplitude of the perturbation, and A_max,used is the upper bound of the amplitude after EMA smoothing. This soft saturation method avoids the discontinuity of the first derivative of the hard clipping (min(p, A_max)) at the clipping transition point, thereby reducing the mechanical abrupt stop sensation while preserving the dynamics.

[0093] For the control quantity after superimposed perturbation, the CIEDE2000 color difference envelope projection described in S3 and the causal limiting described in S5 are executed again to ensure that the final output quantity always falls within the envelope of the color difference hard constraint and the kinematic hard constraint.

[0094] VI. Channel Mapping and Output (S7)

[0095] S7: Maps brightness, color, and orientation parameters into multi-channel lighting control frames and outputs them.

[0096] The internal control layer generates control values ​​at a first frame rate f1 = 100 FPS. The physical output layer sends the data at a second frame rate f2, not exceeding 40 Hz.

[0097] In this embodiment, the convention for mapping the seven-dimensional control quantity to a DMX512-A single universe (512 channels) is as follows:

[0098]

[0099] As an optional implementation, the brightness channel normalization lower limit is 0.05, meaning that when the value is below this, it is output as 0.05 to avoid the lamps from completely turning off. The Pan / Tilt can be configured with a 16-bit mapping between the coarse adjustment channel and the fine adjustment channel to improve the angle control accuracy.

[0100] The control amount for the second frame rate is obtained in one of the following ways:

[0101] (1) Downsample the first frame rate control flow according to the integer decimation ratio M = f1 / f2 (M is a positive integer);

[0102] (2) Perform zero-order hold sampling on the control flow of the first frame rate at each transmission time of the second frame rate;

[0103] (3) After being distributed through a network light protocol (such as ANSI E1.31 / sACN or Art-Net™) in multiple universes, the frames sent by the physical layer of each universe are no higher than the second frame rate.

[0104] In this embodiment, the internal control layer is 100FPS, and the physical transmission uses zero-order hold sampling up to f2 = 40Hz. At this time, f1 / f2 = 2.5 is a non-integer ratio, so zero-order hold instead of integer decimation is used. The upper bound of the time quantization of the physical output layer is 12.5ms.

[0105] As another optional implementation, the second frame rate is 25Hz and the integer decimation ratio M = 4, in which case the upper bound of the time quantization of the physical output layer is 20ms.

[0106] The data export module exports a pseudo-universe byte stream of 3,072,000 bytes (@f1 = 100Hz control archive), which does not constitute a claim of a single DMX512 universe being physically transmitted at a frequency higher than 40Hz (let alone higher than 44Hz). When distributed via the network lighting protocol's multi-universe model, the multi-universe model expands the channel capacity but does not increase the physical layer frame rate of a single universe; the frame rate of each universe remains no higher than the second frame rate.

[0107] VII. Summary of Parameters for Complete Implementation Examples

[0108]

[0109]

[0110] VIII. Ablation Configuration Instructions

[0111] To illustrate the function of each constraint module, the two ablation configurations of this invention are set as follows:

[0112] Ablation Configuration A (the configuration of this invention with chromatic aberration envelope projection and kinematic causal limiting disabled): Employs MIDI hard-trigger + fixed anchor point + linear interpolation, without chromatic aberration envelope projection and kinematic limiting.

[0113] Ablation Configuration B (the configuration of this invention with all hard constraints disabled): This configuration replaces deterministic retrieval and projection with an untrained, randomly initialized network, and artificially injects a delay of 120–180 ms (166.437 ms obtained from uniform[120,180]ms with seed=42, quantized to 17 frames (170.0 ms) at 100 FPS). It only performs soft pruning and does not include any hard constraint mechanisms. The above configuration is the internal ablation setting of this invention and does not represent any publicly disclosed end-to-end generative lighting control system, nor does it constitute a technical assessment of the prior art.

[0114] IX. Verification of Technical Effects

[0115] Under the parameters of the above embodiments, the comparison results between the present invention and ablation configurations A and B are as follows:

[0116]

[0117] The above results indicate that:

[0118] (1) In terms of control layer timing alignment (Temporal RMSE = 0.0000ms), the present invention is on par with ablation configuration A and superior to ablation configuration B;

[0119] (2) In terms of color envelope constraint (Color OOB = 0.0000%, CIEDE2000 OOB = 0.0000%), the present invention is on par with ablation configuration A and superior to ablation configuration B;

[0120] (3) The present invention is superior to ablation configuration A (1.4500%) and ablation configuration B (96.0500%) in terms of equipment capability constraints (GDTF Violation Rate = 0.0000%).

[0121] (4) The square time integral of the accelerometer in this invention is reduced by approximately 3.82 × 10³ times relative to ablation configuration A and by approximately 2.25 × 10³ times relative to ablation configuration B. 8 times;

[0122] (5) The RGB Variance of the present invention is 12.3433, and the ablation configuration A is 5.49×10⁻²³, indicating that the present invention retains rich dynamic changes within the color difference envelope and achieves a combination effect that is both smooth and non-static.

[0123] After 8-bit quantization and recalculation, the ΔE of this invention 00Both OOB and HSV OOB remain at 0.0000%. The maximum CIEDE2000 color difference after quantization is 3.367823 (frame 2612), which is below the threshold of 8.0, leaving a margin of 4.632. The color difference envelope constraint remains strictly valid after 8-bit DMX quantization.

[0124] In summary, the claimable gains of this invention relative to ablation configuration A are: the device capability out-of-bounds rate is reduced from 1.4500% to 0.0000%, the jerk square time integral is reduced by approximately 3.82 × 10³ times, and the color difference envelope constraint remains strictly valid after 8-bit quantization. Under the parameters of this embodiment (first frame rate 100 FPS, tick interval 500 ms), the event timing falls exactly in the integer control frame; therefore, the root mean square error of the control layer timing alignment of this invention and ablation configuration A is the same at 0.0000 ms, and this does not constitute a difference between the two.

[0125] 10. System Implementation Examples

[0126] This invention also provides a real-time stage lighting control system based on semantic mapping and hybrid constraints, comprising:

[0127] (1) Input module, used to obtain the MIDI beat timestamps and semantic cues associated with the audio;

[0128] (2) Semantic retrieval module, used to perform deterministic table lookup or keyword matching in S2;

[0129] (3) Color difference projection module, used to perform the CIEDE2000 color difference envelope projection of S3;

[0130] (4) Timing quantization module, used to perform frame raster quantization and peak setting in S4;

[0131] (5) Kinematic limiting module, used to perform the causal limiting of S5;

[0132] (6) A constrained disturbance module, used to perform the margin constraint disturbance and reprojection and re-limiting of S6;

[0133] (7) Channel mapping and output module, used to perform the multi-channel mapping, first frame rate generation and second frame rate transmission not higher than 40Hz in S7.

[0134] XI. Storage Media Examples

[0135] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in any of the above embodiments.

[0136] The computer-readable storage media include, but are not limited to, non-volatile storage media such as ROM, RAM, magnetic disks, optical disks, and flash memory.

[0137] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A real-time stage lighting control method based on semantic mapping and hybrid constraints, characterized in that, Includes the following steps: S1: Obtain the MIDI beat timestamp and semantic cues associated with the audio; S2: Based on the semantic prompts, a single industrial color card chromaticity anchor point is obtained in the predefined industrial color card library through deterministic table lookup or keyword matching retrieval; S3: Generate candidate color mixing channel values ​​in the neighborhood of the anchor point, wherein the neighborhood is a region centered on the anchor point with an Euclidean radius in the CIELab space not exceeding a preset neighborhood radius; The candidate values ​​are then subjected to CIEDE2000 color difference envelope projection, which includes: transforming the candidate values ​​to the CIELab space; and weighting them accordingly. The color difference between the CIEDE2000 light source (D65) and the standard observer at a 2° angle relative to the anchor point is calculated. If the color difference does not exceed a preset threshold, the candidate value is output; otherwise, the interpolation parameter is used. At the anchor point With candidate points Construct line segments between 1 ,right Perform a one-dimensional binary search until the search interval length is less than a preset tolerance, and then obtain the feasible results that satisfy the threshold. For the corresponding point, after converting it to sRGB, performing color gamut clamping and 8-bit quantization, the color difference between the quantized result and the anchor point is calculated back. If the color difference still exceeds the threshold, then... After multiplying by a preset shrinkage factor, repeat the above transformation, clamping, quantization, and back-substitution verification; when the number of repetitions reaches the preset positive integer upper limit... At that time, the output is the anchor point itself, which is clamped by color gamut and quantized by 8 bits; S4: At the first frame rate The control frame grid for the MIDI timestamp Perform quantization and retrieve frame index. And place the peak sample of the light trigger envelope in the first... One control frame; S5: Parse from the equipment capability description file or use the preset upper limits of angular velocity and angular acceleration, and perform causal limiting on the azimuth and pitch trajectories of the luminaire with azimuth and pitch degrees of freedom; S6: Color difference margin based on the current control quantity The upper bound of the disturbance amplitude of the current control frame is determined by a preset proportional coefficient based on the kinematic margin relative to the upper limits of angular velocity and angular acceleration. ,in, , , The preset positive coefficient is used; the amplitude does not exceed Within the range, a restricted disturbance is superimposed, and the superimposed result is subjected to the color difference envelope projection described in S3 and the causal limiting described in S5 again before outputting a control quantity; the restricted disturbance includes at least one of polynomial easing interpolation and band-limited random disturbance; S7: Maps brightness, color, and orientation parameters to multi-channel light control frames and outputs them; the internal control layer uses the first frame rate. Generate control values, and the physical output layer operates at a second frame rate not exceeding 40Hz.

2. Send; the control amount for the second frame rate is obtained by one of the following methods: by integer decimation ratio (M is a positive integer) The first frame rate control flow is downsampled, or the first frame rate control flow is zero-order held sampled at each transmission time of the second frame rate, or it is distributed to the physical layer through the network light protocol multiverse distribution, wherein the physical layer transmission frame rate of each universe is not higher than the second frame rate.

2. The real-time stage lighting control method based on semantic mapping and hybrid constraints according to claim 1, characterized in that, The deterministic table lookup or keyword matching includes: performing precise table matching after normalizing the semantic prompt string; if no match is found, selecting the entry with the highest score based on the number of keyword matches for the color card entry; when multiple entries have the same number of matches, selecting the first entry in the color card library; when all matches are zero, returning the first entry in the color card library; the color card library contains at least one industrial color filter entry with a CIELab anchor point of (45.0, −5.0, −25.0); the color difference determination is calculated only for the color mixing channel values ​​(R, G, B), and the luminance channel is not involved; each entry in the color card library has a predefined fixed index order.

3. The real-time stage lighting control method based on semantic mapping and hybrid constraints according to claim 1, characterized in that, The color difference envelope projection is a feasibility binary search and back-substitution verification along the ray, rather than an exact projection or minimum shrinkage optimization of the CIEDE2000 non-convex set; the preset color difference threshold is 8.0, according to... D65, 2° observer calculation; the upper limit of the number of outer iterations K is 5, and the bisection tolerance is 1×10. -3 The shrinkage factor is 0.5 and the neighborhood radius is 8.0; the anchor point is located within the sRGB color gamut.

4. The real-time stage lighting control method based on semantic mapping and hybrid constraints according to claim 1, characterized in that, The polynomial easing interpolation is cubic. Or C 2 The continuous polynomial easing is used, wherein the frequency of the band-limited random perturbation is between 0.5 Hz and 2 Hz; the band-limited random perturbation is obtained by filtering white noise through a second-order Butterworth bandpass filter, or by superimposing multiple sinusoidal components in the range of 0.5 Hz to 2 Hz.

5. The real-time stage lighting control method based on semantic mapping and hybrid constraints according to claim 1, characterized in that, The upper limit of the angular velocity is 180° / s, and the upper limit of the angular acceleration is 720° / s. 2 The lower limit of the luminance channel normalization is 0.05; in the multi-channel mapping, the first channel is luminance, the second to fourth channels are RGB, the fifth and sixth channels are Pan / Tilt coarse, and the remaining channels are reserved; when used for DMX512-A physical layer transmission, the transmission frame rate is no higher than 40Hz; when transmitted at 40Hz in zero-order hold sampling mode, the upper limit of the time quantization at the transmitting end is 12.5ms.

6. The real-time stage lighting control method based on semantic mapping and hybrid constraints according to claim 5, characterized in that, The Pan / Tilt consists of a coarse adjustment channel and a fine adjustment channel forming a 16-bit mapping.

7. The real-time stage lighting control method based on semantic mapping and hybrid constraints according to claim 1, characterized in that, The first frame rate is 100 frames per second; the second frame rate is 25Hz and the integer decimation ratio M=4, at which time the upper bound of the time quantization of the physical output layer is 20ms; or the second frame rate is 40Hz and zero-order hold sampling is used, at which time the upper bound of the time quantization of the physical output layer is 12.5ms.

8. The real-time stage lighting control method based on semantic mapping and hybrid constraints according to claim 1, characterized in that, The upper bound of the disturbance amplitude described in S6 is used after being smoothed by a first-order exponential moving average, and the restricted disturbance is calculated according to... The soft saturation method is applied, where This is to disturb the original amplitude.

9. A real-time stage lighting control system based on semantic mapping and hybrid constraints, characterized in that, include: The input module is used to obtain the MIDI beat timestamps and semantic cues associated with the audio. The semantic retrieval module is used to perform the deterministic table lookup or keyword matching described in S2 of claim 1; A color difference projection module is used to perform the CIEDE2000 color difference envelope projection as described in claim 1, S3. The timing quantization module is used to perform frame raster quantization and peak setting as described in claim 1, S4; A kinematic limiting module for performing causal limiting as described in claim 1, S5; A constrained perturbation module is used to perform the margin constraint perturbation and reprojection re-limiting as described in S6 of claim 1; The channel mapping and output module is used to perform the multi-channel mapping, first frame rate generation and second frame rate transmission not exceeding 40Hz as described in claim 1, S7. The specific limitations of each module are consistent with the corresponding steps described in claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1 to 8.

Images