Intelligent inspection method and system in an animation production process

Abstract

This invention discloses an intelligent inspection method and system for animation production, relating to the field of intelligent animation inspection technology. It solves the technical problems of disconnected inspection throughout the entire process and independent inspections in pre-production, mid-production, and post-production in existing technologies. The system corresponds to the entire animation production process, from pre-production to mid-production, achieving automated, precise inspection and targeted adjustments at each stage. The front-end inspection module focuses on frame-by-frame errors in the early stages of animation, constructing static error maps and dynamic error trajectories through frame-by-frame processing and cross-scene and cross-frame inspection, enabling precise marking and graded repair of error scenes. The mid-end inspection module focuses on the rationality of the main movements in the animation after pre-production adjustments, identifying and repairing issues such as reverse joint movements and unnatural movement transitions in real time. The back-end inspection module checks the final animation for color consistency and audio-visual synchronization, ensuring the animation production quality meets standards before completion.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection technology for animation, specifically to an intelligent inspection method and system for the animation production process. Background Technology

[0002] The animation production industry is currently developing towards refinement, scale, and efficiency. Whether it is 2D or 3D animation, its production process covers multiple stages, including pre-production storyboard design, character and scene modeling, mid-production motion production, post-production rendering and audio-visual compositing, and each stage has a large number of details that require manual inspection.

[0003] Currently, the mainstream inspection method in the industry is still mainly manual inspection, supplemented by a small number of simple automated tools, which has many pain points:

[0004] First, manual inspection is inefficient. As the number of animation frames increases and the complexity of scenes increases, manual frame-by-frame inspection is not only time-consuming and labor-intensive, but also easily affected by the fatigue and experience of the inspectors, leading to the omission of problems such as continuity errors, non-standard actions, and audio-visual asynchrony.

[0005] Second, the inspection standards are not uniform, and the judgment standards of different inspectors are different, which leads to inconsistent inspection results for the same animation, and there is no clear basis for rework and adjustment.

[0006] Third, the entire process of inspection is disconnected. The early, middle and late inspections are independent of each other. If the problems that are discovered in the early stage are not resolved in time, they will be passed on to the later stages, resulting in a significant increase in rework costs.

[0007] Fourth, there is a lack of targeted, tiered repair mechanisms. A uniform repair method is used for different types and severity of problems, which is inefficient and ineffective.

[0008] To address the aforementioned technical shortcomings, a solution is proposed: an intelligent inspection system that is comprehensive, automated, and precise, enabling efficient inspection and scientific repair at each stage of animation production. Summary of the Invention

[0009] The purpose of this invention is to solve the problems mentioned above by proposing an intelligent inspection method and system for the animation production process.

[0010] The objective of this invention can be achieved through the following technical solutions:

[0011] An intelligent inspection system for animation production includes an intelligent inspection platform, wherein the intelligent inspection platform is connected to the following communication links:

[0012] The front-end inspection module performs preliminary checks on the animation production process and makes adjustments based on the inspection results;

[0013] The mid-stage inspection module performs partial mid-stage inspections on the animation production process after the initial inspection and adjustments have been completed.

[0014] The backend inspection module performs backend checks on the animation production and completes the production process after the checks are passed.

[0015] Furthermore, the process of the front-end inspection module is as follows:

[0016] The real-time animation is divided into frames, and cross-scene and cross-frame checks are performed on each frame.

[0017] The images are sorted in chronological order, and the scenes of the images are determined. Based on the scenes created in each image, they are divided into cross-scene scenes within the same frame and cross-scene scenes outside the same frame.

[0018] In the same frame across scenes, obtain the change in the proportion of characters in the production scene before and after the scene change. If the change in the proportion of characters exceeds the set change threshold, the current same frame across scenes is marked as a scene with a continuity error; if the change in the proportion of characters does not exceed the set change threshold, the current same frame across scenes is marked as a normal scene.

[0019] In non-same-frame cross-scene scenarios, obtain the offset distance of the character and prop positions in the production scene before and after the non-same-frame scene change. If the offset distance of the character and prop positions is not zero, mark the current non-same-frame cross-scene as a non-same-frame glitch scene; if the offset distance of the character and prop positions is zero, mark the current non-same-frame cross-scene as a non-same-frame normal scene.

[0020] Based on the locations of normal scenes and scenes with continuity errors within the same frame, a static image of the error area is constructed; and based on the locations of error scenes and normal scenes outside the same frame, a dynamic error trajectory is constructed by connecting their positional trajectories.

[0021] Furthermore, the distribution density of the normal scene in the same frame and the distribution density of the scene with the error in the same frame are obtained within the static error area map. If the distribution density is within the corresponding set threshold range, the current static error area map is set as the preliminary repair map; if the distribution density is not within the corresponding set threshold range, the current static error area map is set as the preliminary rework map; and the area map type is sent to the front-end inspection module, which then makes repair and rework decisions upon receiving the map.

[0022] The locations of scenes with continuity errors within the same frame in the previously repaired image are marked, and a region threshold is set. Specifically, a unit area is set. If both normal scene areas and scenes with continuity errors exist within the region threshold, a full repair signal is generated and sent to the front-end inspection module along with the corresponding region location. If only one of the normal scene areas and scenes with continuity errors exists within the region threshold, a partial repair signal is generated and sent to the front-end inspection module along with the location of the only scene with continuity errors. After receiving the signal, the front-end inspection module performs static repair based on the signal type.

[0023] Furthermore, the dynamic continuity error trajectory is analyzed. Based on the location of the area to be repaired, it is determined whether the starting point of the dynamic continuity error trajectory is a non-frame continuity error scene. If so, the frequency increase span of the non-frame continuity error scene within the dynamic continuity error trajectory is identified. When the frequency increase span exceeds the set increase span threshold, the dynamic continuity error trajectory is marked as a continuous continuity error trajectory and sent to the pre-production inspection end. After receiving it, the pre-production inspection end synchronously checks and adjusts the adjacent areas of the production animation where the continuous continuity error trajectory is located. When the frequency increase span does not exceed the set increase span threshold, the dynamic continuity error trajectory is marked as a discontinuous continuity error trajectory and sent to the pre-production inspection end. After receiving it, the pre-production inspection end checks and adjusts the location of the area covered by the discontinuous continuity error trajectory and performs spot checks on the adjacent areas of the production animation.

[0024] If not, the cumulative number of non-same-frame error scenes within the dynamic error trajectory is identified. When the cumulative number exceeds the set cumulative number threshold, the dynamic error trajectory is marked as an error development trajectory and sent to the early inspection end. After receiving it, the early inspection end checks and adjusts the error development trajectory while simultaneously monitoring it according to the trajectory trend and intervening in a timely manner to prevent the number of frames of the error scene from continuing to increase.

[0025] When the cumulative number of occurrences does not exceed the set cumulative number threshold, the dynamic error tracking is marked as the error cumulative tracking and sent to the early inspection end. After receiving it, the early inspection end repairs the corresponding area of ​​the error cumulative tracking.

[0026] Furthermore, the process of the mid-range inspection module is as follows:

[0027] Perform partial checks based on the main elements within the animation;

[0028] Based on the subject's actions in the corresponding scene, actions are divided into static actions and dynamic actions. For example, a person moving or a plant swaying is a dynamic action, while a plant standing still or a person standing still is a static action.

[0029] Obtain the main body motion trajectory nodes in the dynamic action, and determine the movement limit of each trajectory node based on the main body structure and corresponding motion logic, specifically represented as the movement space in the running trajectory where the node is located; obtain the moment when the real-time action position exceeds the movement limit during the execution phase of the dynamic action running trajectory, and mark it as the moment of reverse joint movement;

[0030] Simultaneously, the set points of the main body in the static action are obtained, and when the static action changes, the motion change process of the set points of the main body in the static action is obtained, and the motion change speed of the corresponding main body is identified. According to the type of main body in the animation, the speed movement range is set, and the real-time motion change speed is compared with the speed movement range. If the real-time motion change speed is not within the speed movement range, the corresponding moment is marked as a non-natural movement moment; otherwise, the real-time motion change speed is within the speed movement range, and the corresponding moment is marked as a natural movement moment.

[0031] Furthermore, during the mid-term inspection phase, anti-joint movement moments and unnatural movement moments are identified and immediately repaired upon identification. The process of gradually identifying and repairing these moments is used to check and improve the animation. Once completed, a mid-term inspection completion signal is sent to the intelligent inspection platform.

[0032] Furthermore, the backend inspection module's process is as follows:

[0033] The animation is played in random segments according to the plot sequence. The color difference of the corresponding image color space at adjacent moments under the same plot in the random segment playback stage is obtained. At the same time, the time difference between the audio waveform and the timestamp of the corresponding lip-sync animation in the random segment playback stage is collected.

[0034] Furthermore, if the color difference between adjacent moments in the color space of the corresponding image under the same plot within the random segmented playback phase does not exceed the color difference span threshold, and the time deviation between the audio waveform and the timestamp of the corresponding lip-sync animation within the random segmented playback phase does not exceed the deviation duration threshold, then a backend inspection pass signal is generated and sent to the intelligent inspection platform along with the animation to complete the animation.

[0035] If the color difference between adjacent moments in the color space of the corresponding image under the same scene in the random segmented playback phase exceeds the color difference span threshold, or if the time deviation between the audio waveform and the timestamp of the corresponding lip-sync animation in the random segmented playback phase exceeds the deviation duration threshold, a backend inspection failure signal will be generated and sent to the intelligent inspection platform along with the animation to adjust the images inside the animation.

[0036] Furthermore, an intelligent inspection method for the animation production process is provided, the intelligent inspection method being as follows:

[0037] Step 1: Front-end inspection. Conduct a preliminary inspection of the animation production process and make adjustments based on the inspection results.

[0038] Step 2: Mid-term inspection, a partial mid-term inspection of the animation after the preliminary inspection and adjustments have been completed;

[0039] Step 3: Backend check. Perform a backend check on the animation and complete the production process after the check is passed.

[0040] Compared with the prior art, the beneficial effects of the present invention are:

[0041] 1. The front-end inspection module accurately identifies hidden continuity errors in the early stages of animation production. Through image frame segmentation and cross-scene and cross-frame inspection, it can accurately identify hidden errors that are difficult to detect manually, such as abnormal character proportions in the same frame across scenes and misaligned character and prop positions in different frames across scenes. This avoids missing continuity errors and improves the quality of early production. It also enables hierarchical management of continuity errors by constructing static continuity error area maps and dynamic continuity error trajectories to distinguish different types and severity of errors, providing a clear basis for subsequent repairs and avoiding blind repairs. Furthermore, it enhances the targeting of repairs by generating comprehensive and partial repair signals based on the distribution of continuity errors, achieving precise repairs that avoid both over-repairing and wasting resources, as well as incomplete repairs leading to rework. It reduces early production risks by distinguishing between early repair and early rework images based on the static continuity error area map, promptly identifying and reworking serious problems in early production, preventing problem accumulation and reducing subsequent production risks. Finally, it reduces manual workload by replacing traditional manual frame-by-frame inspection with automated frame segmentation, continuity error identification, and marking, significantly reducing manual workload and improving the efficiency of early inspections.

[0042] 2. The mid-stage inspection module addresses the rationality issues of animation movements during the mid-stage, ensuring natural and compliant movements. It accurately identifies reverse joint movements in dynamic actions and unnatural movements during transitions between static actions, providing timely real-time corrections to ensure the main movements conform to physiological structure and motion logic, thus enhancing the naturalness and visual appeal of the animation. It also improves the accuracy of movement inspection, ensuring each movement meets production requirements and shortening the mid-stage production cycle. Through real-time identification and immediate correction, it avoids the accumulation of movement problems, reduces rework in later stages, and shortens the mid-stage production and inspection cycle. Furthermore, it reduces the difficulty of movement production, providing animators with clear guidelines for movement adjustments (such as the timing of reverse joint movements and the speed of unnatural movements), helping them quickly optimize movements and improve production efficiency and quality.

[0043] 3. The backend inspection module addresses color and audio-visual synchronization issues in post-production animation. It ensures color consistency by detecting color differences between adjacent frames in the same scene, preventing sudden color changes and ensuring uniform animation colors to improve the viewing experience. It also achieves audio-visual synchronization by detecting timestamp discrepancies between audio and lip-sync animation, making timely adjustments to avoid mismatches between lip movements and sound, and ensuring the integrity and watchability of the animation. Attached Figure Description

[0044] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0045] Figure 1 This is a system principle block diagram of the present invention;

[0046] Figure 2 This is a flowchart of the backend inspection module in this invention. Detailed Implementation

[0047] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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.

[0048] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0049] Please see Figure 1 As shown, an intelligent inspection system for animation production includes an intelligent inspection platform, wherein the intelligent inspection platform is communicatively connected to a front-end inspection module, a mid-end inspection module, and a back-end inspection module, and the front-end inspection module, mid-end inspection module, and back-end inspection module correspond to the early stage, mid-stage, and late stage of animation production, respectively.

[0050] The intelligent inspection platform generates front-end inspection signals and sends them to the front-end inspection module;

[0051] After receiving the front-end inspection signal, the front-end inspection module performs a preliminary inspection of the animation production process and makes adjustments based on the inspection results.

[0052] The real-time animation is divided into frames, and cross-scene and cross-frame checks are performed on each frame.

[0053] The images are sorted in chronological order, and the scenes of the images are determined. Based on the scenes created in each image, they are divided into cross-scene scenes within the same frame and cross-scene scenes outside the same frame.

[0054] In the same frame across scenes, obtain the change in the proportion of characters in the production scene before and after the scene change. If the change in the proportion of characters exceeds the set change threshold, the current same frame across scenes is marked as a scene with a continuity error; if the change in the proportion of characters does not exceed the set change threshold, the current same frame across scenes is marked as a normal scene.

[0055] In non-same-frame cross-scene scenarios, obtain the offset distance of the character props' positions in the production scene before and after the non-same-frame scene change. If the offset distance of the character props' positions is not zero, then the current non-same-frame cross-scene is marked as a non-same-frame glitch scene. It should be explained that character props refer to the clothes, accessories, and props such as trees in the production animation scene. If the offset distance of the character props' positions is zero, then the current non-same-frame cross-scene is marked as a non-same-frame normal scene.

[0056] Based on the image locations of normal scenes and scenes with flaws within the same frame, a static flaw area map is constructed; and based on the image locations of flaw scenes and normal scenes outside the same frame, a dynamic flaw trajectory is constructed by connecting the positional trajectories.

[0057] The system obtains the distribution density of the normal scene and the scene with the error within the same frame in the static error area map. If both distribution densities are within the corresponding set threshold range, the current static error area map is set as the preliminary repair map. It should be noted that the distribution density of the normal scene within the same frame is within the threshold range, and the distribution density of the scene with the error is also within the threshold range. If the distribution densities are not both within the corresponding set threshold range, the current static error area map is set as the preliminary rework map. The area map type is then sent to the front-end inspection module, which receives the map and makes repair and rework decisions.

[0058] The locations of scenes with continuity errors within the same frame in the previously repaired image are marked, and a region threshold is set. Specifically, a unit area is set. If both normal scene areas and scenes with continuity errors exist within the region threshold, a full repair signal is generated and sent to the front-end inspection module along with the corresponding region location. If only one of the normal scene areas and scenes with continuity errors exists within the region threshold, a partial repair signal is generated and sent to the front-end inspection module along with the location of the only scene with continuity errors. After receiving the signal, the front-end inspection module performs static repair based on the signal type.

[0059] The system analyzes dynamic continuity errors and, based on the location of the area to be repaired, determines whether the starting point of the dynamic continuity error trajectory is a non-frame continuity error scene. If so, it identifies the frequency increase span of non-frame continuity error scenes within the dynamic continuity error trajectory. When the frequency increase span exceeds a set increase span threshold, the dynamic continuity error trajectory is marked as a continuous continuity error trajectory and sent to the pre-production inspection end. Upon receiving the pre-production inspection end, it synchronously checks and adjusts the adjacent areas of the production animation containing the continuous continuity error trajectory. When the frequency increase span does not exceed the set increase span threshold, the dynamic continuity error trajectory is marked as a discontinuous continuity error trajectory and sent to the pre-production inspection end. Upon receiving the pre-production inspection end, it checks and adjusts the location of the area covered by the discontinuous continuity error trajectory and performs spot checks on the adjacent areas of the production animation.

[0060] If not, the cumulative number of non-same-frame error scenes within the dynamic error trajectory is identified. When the cumulative number exceeds the set cumulative number threshold, the dynamic error trajectory is marked as an error development trajectory and sent to the early inspection end. After receiving it, the early inspection end checks and adjusts the error development trajectory while simultaneously monitoring it according to the trajectory trend and intervening in a timely manner to prevent the number of frames of the error scene from continuing to increase.

[0061] When the cumulative number of occurrences does not exceed the set cumulative number threshold, the dynamic error trajectory is marked as the error cumulative trajectory and sent to the early inspection end. After receiving it, the early inspection end repairs the corresponding area of ​​the error cumulative trajectory. It should be explained that the error development trajectory is a multi-point adjustment in the adjustment phase, while the error cumulative trajectory is a single-point adjustment in the adjustment phase. The former can control the development, while the latter can reduce the cumulative duration.

[0062] Front-end inspection module related thresholds

[0063] The threshold for changes in character proportions is based on the standards and specifications of the animation production industry. It combines the character design requirements of different types of animation (such as cartoon animation and realistic animation) to ensure that the proportions of characters remain consistent across scenes within the same frame, and to avoid obvious proportion discrepancies (such as a sudden change in the head and body proportions of the same character in different scenes within the same frame).

[0064] Acquisition method: By collecting a large amount of cross-scene character ratio data from the same frame of excellent animations of the same type, statistical analysis is performed to calculate the reasonable range of ratio changes, and the upper limit of the reasonable range is taken as the threshold for setting the amount of change; at the same time, combined with the personalized needs of the animation production team, minor adjustments are allowed, and the threshold is finally determined.

[0065] The thresholds for normal scene distribution density and flaw scene distribution density in the same frame are based on the efficiency of front-end inspection and the cost of repair. They ensure the rationality of the static flaw area map. If the normal scene distribution is too sparse and the flaw scene distribution is too dense, it indicates that there are many problems in the early production and rework is required. If both distributions are within a reasonable range, targeted repairs can be carried out.

[0066] Acquisition method: Machine learning training is performed using historical front-end inspection data to statistically analyze the distribution density range of normal scenes and scenes with errors in static error areas of different quality levels. Combined with rework costs and repair efficiency, reasonable ranges for two thresholds are determined. At the same time, the thresholds are dynamically adjusted according to the complexity of animation production (such as the number of scenes and characters).

[0067] Area threshold (unit area): This is used to distinguish between areas that require full repair and areas that require partial repair, ensuring that the repair is targeted and avoiding over-repair or incomplete repair.

[0068] Acquisition method: Combine the animation resolution, frame size, and the size of characters and scenes to set a reasonable unit area (e.g., each unit area is 100×100 pixels); at the same time, refer to historical repair data and optimize the threshold according to the repair effect to ensure that the repair within the unit area can take into account both accuracy and efficiency.

[0069] The frequency increase span threshold is designed to distinguish between continuous and discontinuous flaw tracks, and to develop differentiated inspection and adjustment strategies for different types of dynamic flaw tracks to prevent the continuous spread of continuous flaw scenes.

[0070] Acquisition method: Collect historical dynamic continuity error trajectory data, statistically analyze the frequency increase pattern of continuity errors in different continuity error trajectories, and combine the animation frame rate and scene switching speed to determine the reasonable span threshold of frequency increase; if the frequency increases too fast, it indicates that the continuity error scene is spreading continuously, and it needs to be checked and adjusted.

[0071] Cumulative count threshold: This is designed to distinguish between the development trajectory of a continuity error and the cumulative trajectory of a continuity error, to determine the development trend of a continuity error scene, and to prevent the number of continuity error frames from increasing continuously.

[0072] Acquisition method: By statistically analyzing the cumulative number of non-same-frame continuity errors in historical continuity error trajectories and their correlation with subsequent continuity error trends, and combining this with the overall progress and quality requirements of animation production, a threshold for the cumulative number of errors is set; if the cumulative number exceeds the threshold, it indicates that the continuity error is developing and requires simultaneous monitoring and intervention.

[0073] The intelligent inspection platform generates mid-range inspection signals and sends them to the mid-range inspection module;

[0074] After receiving the mid-term inspection signal, the mid-term inspection module performs a partial mid-term inspection on the animation that has completed the preliminary inspection and adjustment.

[0075] Perform partial checks based on the main subject in the animation, which can be represented by characters such as people, animals, or plants;

[0076] Based on the subject's actions in the corresponding scene, actions are divided into static actions and dynamic actions. For example, a person moving or a plant swaying is a dynamic action, while a plant standing still or a person standing still is a static action.

[0077] Obtain the main body motion trajectory nodes in the dynamic action, and determine the movement limit of each trajectory node based on the main body structure and corresponding motion logic, specifically represented as the movement space in the running trajectory where the node is located; obtain the moment when the real-time action position exceeds the movement limit during the execution phase of the dynamic action running trajectory, and mark it as the moment of reverse joint movement;

[0078] Simultaneously, the set points of the main body in the static action are obtained, and when the static action changes, the motion change process of the set points of the main body in the static action is obtained, and the motion change speed of the corresponding main body is identified. According to the type of main body in the animation, the speed movement range is set, and the real-time motion change speed is compared with the speed movement range. If the real-time motion change speed is not within the speed movement range, the corresponding moment is marked as a non-natural movement moment; otherwise, the real-time motion change speed is within the speed movement range, and the corresponding moment is marked as a natural movement moment.

[0079] During the mid-term inspection phase, anti-joint movement moments and unnatural movement moments are identified and immediately repaired when the corresponding moments are identified. The process of identifying and repairing these moments is carried out step by step to check and improve the animation. Once completed, a mid-term inspection completion signal is sent to the intelligent inspection platform.

[0080] Dynamic motion trajectory node movement limit: It is based on the physiological structure of the subject (such as the range of motion of human joints, the bending range of plant branches) and motion logic (such as the joint movement trajectory when a person walks, the branch movement trajectory when a plant sways), to ensure that dynamic movements conform to natural laws and avoid unnatural movements such as reverse joints.

[0081] Acquisition method: By collecting motion data of real organisms (people, animals, plants), a motion model is established to determine the reasonable movement space (i.e., movement limit) of each trajectory node; at the same time, the limit range is adjusted in combination with the style of the animation (such as cartoon animation can appropriately relax the limit, while realistic animation must strictly follow the real limit) to finally determine the movement limit of each node.

[0082] The speed and range of motion during static action transitions: This is to ensure the naturalness of static action transitions and avoid transitions that are too fast or too slow, which do not conform to the subject's movement characteristics (e.g., when a person moves from stillness to raising their hand, a transition that is too fast will appear stiff, while a transition that is too slow will appear sluggish).

[0083] Acquisition method: Collect speed data of static movement transformation of real organisms, combine with the frame rate of the animation and the subject type (such as the difference in movement speed between people and plants), and calculate a reasonable speed range; at the same time, take into account the artistic requirements of animation production, allow for minor artistic adjustments, and determine the final speed movement range.

[0084] After receiving the signal, the intelligent inspection platform generates a backend inspection signal and sends it to the backend inspection module.

[0085] Please see Figure 2 As shown, the backend inspection module performs a backend inspection on the animation production and completes the production process after the inspection is passed.

[0086] The animation is played in random segments according to the plot sequence. The color difference of the corresponding image color space at adjacent moments under the same plot in the random segment playback stage is obtained. At the same time, the time difference between the audio waveform and the corresponding lip-sync animation timestamp in the random segment playback stage is collected.

[0087] If the color difference between adjacent moments in the color space of the corresponding image under the same plot within the random segmented playback phase does not exceed the color difference span threshold, and the time deviation between the audio waveform and the timestamp of the corresponding lip-sync animation within the random segmented playback phase does not exceed the deviation time threshold, then it is inferred that the animation production backend check has passed, a backend check pass signal is generated and sent to the intelligent inspection platform along with the animation production, and the animation is completed.

[0088] If the color difference between adjacent moments in the color space of the corresponding image under the same scene in the random segmented playback phase exceeds the color difference span threshold, or if the time deviation between the audio waveform and the timestamp of the corresponding lip-sync animation in the random segmented playback phase exceeds the deviation duration threshold, it is inferred that the back-end inspection of the animation production has failed. A back-end inspection failure signal is generated and sent to the intelligent inspection platform along with the animation production for internal image adjustment.

[0089] Color difference span threshold: This is designed to ensure the consistency of colors in animation scenes within the same plot, avoid sudden color changes between adjacent moments that could affect the viewing experience, and comply with industry standards for animation post-production rendering.

[0090] Acquisition method: Collect a large amount of color data from the post-production rendering of excellent animations, statistically analyze the color difference range of adjacent frames under the same plot, and determine the color difference span threshold by combining the sensitivity of the human eye to color changes; at the same time, dynamically adjust the threshold according to the color style of the animation (such as warm color tone, cool color tone) to ensure color consistency while retaining reasonable color changes.

[0091] Deviation duration threshold: This is designed to ensure the synchronization between audio and lip-sync animation, avoiding mismatches between lip movements and sound, and meeting industry requirements for post-production audio-visual compositing in animation.

[0092] Acquisition method: Referencing industry standards for audio-visual synchronization (such as a deviation duration of no more than 0.1 seconds), and combining the animation's frame rate and audio sampling rate, determine the deviation duration threshold; at the same time, collect historical audio-visual synthesis data, and optimize the threshold based on audience feedback and viewing experience to ensure the naturalness of audio-visual synchronization.

[0093] An intelligent inspection system for the animation production process, the intelligent inspection method is as follows:

[0094] Step 1: Front-end inspection. Conduct a preliminary inspection of the animation production process and make adjustments based on the inspection results.

[0095] Step 2: Mid-term inspection, a partial mid-term inspection of the animation after the preliminary inspection and adjustments have been completed;

[0096] Step 3: Backend check. Perform a backend check on the animation and complete the production process after the check is passed.

[0097] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent inspection system for animation production, characterized in that, This includes an intelligent inspection platform, whose communication connections include: The front-end inspection module performs preliminary checks on the animation production process and makes adjustments based on the inspection results. The process of the front-end inspection module is as follows: The real-time animation is divided into frames, and cross-scene and cross-frame checks are performed on each frame. The images are sorted in chronological order, and the scenes of the images are determined. Based on the scenes created in each image, they are divided into cross-scene scenes within the same frame and cross-scene scenes outside the same frame. In the same frame across scenes, obtain the change in the proportion of characters in the production scene before and after the scene change. If the change in the proportion of characters exceeds the set change threshold, mark the current same frame across scenes as a same frame continuity error scene. If the change in character proportion does not exceed the set change threshold, the current cross-scene frame will be marked as a normal scene within the same frame. In non-same-frame cross-scene scenarios, obtain the offset distance of the character and prop positions in the production scene before and after the non-same-frame scene change. If the offset distance of the character and prop positions is not zero, mark the current non-same-frame cross-scene scenario as a non-same-frame glitch scene. If the offset distance of the character's prop position is zero, then the current non-same-frame cross-scene will be marked as a non-same-frame normal scene; Based on the image locations of normal scenes and scenes with flaws within the same frame, a static flaw area map is constructed; and based on the image locations of flaw scenes and normal scenes outside the same frame, a dynamic flaw trajectory is constructed by connecting the positional trajectories. Obtain the distribution density of normal scenes and flawed scenes within the same frame in the static flaw area map. If both distribution densities are within the corresponding set threshold range, the current static flaw area map is set as the preliminary repair map; if the distribution densities are not both within the corresponding set threshold range, the current static flaw area map is set as the preliminary rework map. The area map type is then sent to the front-end inspection module, which receives the map and makes repair and rework decisions. The locations of scenes with continuity errors within the same frame in the previously repaired image are marked, and a region threshold is set. Specifically, a unit area is set. If both normal scene areas and scenes with continuity errors exist within the region threshold, a full repair signal is generated and sent to the front-end inspection module along with the corresponding region location. If only one of the normal scene areas and scenes with continuity errors exists within the region threshold, a partial repair signal is generated and sent to the front-end inspection module along with the location of the only scene with continuity errors. After receiving the signal, the front-end inspection module performs static repair based on the signal type. The system analyzes dynamic continuity errors and, based on the location of the area to be repaired, determines whether the starting point of the dynamic continuity error trajectory is a non-frame continuity error scene. If so, it identifies the frequency increase span of non-frame continuity error scenes within the dynamic continuity error trajectory. When the frequency increase span exceeds a set increase span threshold, the dynamic continuity error trajectory is marked as a continuous continuity error trajectory and sent to the pre-production inspection end. Upon receiving the pre-production inspection end, it synchronously checks and adjusts the adjacent areas of the production animation containing the continuous continuity error trajectory. When the frequency increase span does not exceed the set increase span threshold, the dynamic continuity error trajectory is marked as a discontinuous continuity error trajectory and sent to the pre-production inspection end. Upon receiving the pre-production inspection end, it checks and adjusts the location of the area covered by the discontinuous continuity error trajectory and performs spot checks on the adjacent areas of the production animation. If not, the cumulative number of non-same-frame error scenes within the dynamic error trajectory is identified. When the cumulative number exceeds the set cumulative number threshold, the dynamic error trajectory is marked as an error development trajectory and sent to the early inspection end. After receiving it, the early inspection end checks and adjusts the error development trajectory while simultaneously monitoring it according to the trajectory trend and intervening in a timely manner to prevent the number of frames of the error scene from continuing to increase. When the cumulative number of times does not exceed the set cumulative number of times threshold, the dynamic error trajectory is marked as the error cumulative trajectory and sent to the early inspection end. After receiving it, the early inspection end repairs the corresponding area of ​​the error cumulative trajectory. The mid-stage inspection module performs partial mid-stage inspections on the animation production process after the initial inspection and adjustments have been completed. The backend inspection module performs backend checks on the animation production and completes the production process after the checks are passed.

2. The intelligent inspection system for animation production process according to claim 1, characterized in that, The process of the mid-range inspection module is as follows: Perform partial checks based on the main elements within the animation; Based on the subject's actions in the corresponding scene, actions are divided into static actions and dynamic actions. For example, a person moving or a plant swaying is a dynamic action, while a plant standing still or a person standing still is a static action. Obtain the main body motion trajectory nodes in the dynamic action, and determine the movement limit of each trajectory node based on the main body structure and corresponding motion logic, specifically represented as the movement space in the running trajectory where the node is located; obtain the moment when the real-time action position exceeds the movement limit during the execution phase of the dynamic action running trajectory, and mark it as the moment of reverse joint movement; Simultaneously, the set points of the main body in the static action are obtained, and when the static action changes, the motion change process of the set points of the main body in the static action is obtained, and the motion change speed of the corresponding main body is identified. According to the type of main body in the animation, the speed movement range is set, and the real-time motion change speed is compared with the speed movement range. If the real-time motion change speed is not within the speed movement range, the corresponding moment is marked as a non-natural movement moment; otherwise, the real-time motion change speed is within the speed movement range, and the corresponding moment is marked as a natural movement moment.

3. The intelligent inspection system for animation production process according to claim 2, characterized in that, During the mid-term inspection phase, anti-joint movement moments and unnatural movement moments are identified and immediately repaired upon identification. The process of identifying and repairing these moments is carried out step by step to check and improve the animation. Once completed, a mid-term inspection completion signal is sent to the intelligent inspection platform.

4. The intelligent inspection system for animation production process according to claim 1, characterized in that, The backend inspection module process is as follows: The animation is played in random segments according to the plot sequence. The color difference of the corresponding image color space at adjacent moments under the same plot in the random segment playback stage is obtained. At the same time, the time difference between the audio waveform and the timestamp of the corresponding lip-sync animation in the random segment playback stage is collected.

5. The intelligent inspection system for animation production process according to claim 4, characterized in that, If the color difference between adjacent moments in the color space of the corresponding image under the same scene in the random segmented playback phase does not exceed the color difference span threshold, and the time deviation between the audio waveform and the timestamp of the corresponding lip-sync animation in the random segmented playback phase does not exceed the deviation time threshold, then a back-end inspection pass signal is generated and sent to the intelligent inspection platform along with the animation to complete the animation. If the color difference between adjacent moments in the color space of the corresponding image under the same scene in the random segmented playback phase exceeds the color difference span threshold, or if the time deviation between the audio waveform and the timestamp of the corresponding lip-sync animation in the random segmented playback phase exceeds the deviation duration threshold, a backend inspection failure signal will be generated and sent to the intelligent inspection platform along with the animation to adjust the images inside the animation.

6. An intelligent inspection method in the animation production process, characterized in that, The intelligent inspection system applied to the animation production process according to any one of claims 1-5 includes the following intelligent inspection method: Step 1: Front-end inspection. Conduct a preliminary inspection of the animation production process and make adjustments based on the inspection results. Step 2: Mid-term inspection, a partial mid-term inspection of the animation after the preliminary inspection and adjustments have been completed; Step 3: Backend check. Perform a backend check on the animation and complete the production process after the check is passed.

Images