AI-driven watch movement full life cycle intelligent management system
By using modular processing of light frequency monitoring, exposure comparison, ghost image recognition, and rhythm control, the problem of ghost image superposition caused by the offset between exposure rhythm and illumination period in the optical inspection of watch movements has been solved, thus improving the accuracy and stability of engraving recognition.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- FUJIAN ZHONGCHEN PRECISION MOVEMENT CO LTD
- Filing Date
- 2026-02-10
- Publication Date
- 2026-06-23
Smart Images
Figure CN121684879B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of product lifecycle management technology, specifically to an AI-driven intelligent management system for the entire lifecycle of watch movements. Background Technology
[0002] The AI-driven intelligent management system for the entire lifecycle of watch movements is a comprehensive management system that integrates artificial intelligence and big data processing technologies. It collects data, performs intelligent analysis, and makes dynamic decisions throughout the entire process of watch movement design, manufacturing, assembly, testing, use, maintenance, and even recycling. The system uses multi-source sensors to collect key parameters of the movement in real time, such as assembly accuracy, operating resistance, lubrication status, temperature and humidity environment, and vibration frequency. It relies on big data processing technology to clean, aggregate, and time-series model the multi-dimensional data collected at different stages, and uses artificial intelligence models to identify the movement's operating patterns, wear trends, and potential anomalies. The system reconstructs the movement's working state in a digital twin environment, achieving intelligent monitoring and adaptive management across the entire chain from production quality control to user use and after-sales maintenance. In this way, the system enables visualized tracking of movement status, health assessment, lifespan prediction, and fault early warning, constructing a closed-loop management system covering R&D, manufacturing, and maintenance, providing data-driven precise decision support and quality assurance for high-end watch manufacturing.
[0003] The existing technology has the following shortcomings:
[0004] In existing technologies, during the optical inspection stage of watch movements, the exposure rhythm is often difficult to synchronize strictly with the reflection period of the illumination source, easily causing optical frequency misalignment on highly reflective metal surfaces. When the exposure rhythm deviates from the reflection period, the camera alternates sampling between the bright and dark phases of the light source, resulting in ghosting in the engraving area. Since AI recognition primarily relies on image texture and edge comparison, this ghosting can be misjudged as the actual edge of the engraving, leading to deviations in traceability coding recognition results and even recording incorrect numbers. Such problems are more likely to occur on highly reflective or finely etched surfaces and are difficult to correct using traditional illumination compensation methods, ultimately distorting the movement traceability chain and affecting the accuracy of production quality control and subsequent maintenance.
[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention
[0006] The purpose of this invention is to provide an AI-driven intelligent management system for the entire lifecycle of a watch movement, in order to solve the problems mentioned in the background art.
[0007] To achieve the above objectives, the present invention provides the following technical solution: an AI-driven intelligent management system for the entire lifecycle of a watch movement, comprising a light frequency monitoring module, an exposure comparison and analysis module, a ghost image recognition and retrospection module, a rhythm control core module, and an exposure correction and execution module:
[0008] The light frequency monitoring module collects assembly lighting data, operating temperature data, and maintenance record data during the entire life cycle management of the watch movement. Based on the collected data, it tracks the changing trend of light frequency and generates a light frequency offset record table.
[0009] The exposure comparison and analysis module performs point-by-point comparison and analysis of the camera's exposure rhythm based on the illumination frequency offset record table, identifies time segments where the illumination phase of light and the dark phase of shadow are out of sync, and generates ghost trigger records on the time axis.
[0010] The ghost image recognition and backtracking module records the contour change process of the code image based on ghost image triggering, analyzes the texture features of the code edge, distinguishes between the real code edge and the reflected ghost image edge, determines the starting position of the error recognition, and generates a bitmap of the misjudgment source.
[0011] The rhythm control core module determines the exposure pause sequence and reverse phase sampling window based on the misjudged source bitmap, and generates exposure rhythm adjustment instructions in combination with the bounce suppression measures of the illumination reference point, which are used to dynamically correct the camera exposure process.
[0012] The exposure correction and execution module adjusts the exposure phase online based on the exposure rhythm adjustment command. During the critical period when the light phase and dark phase are misaligned, a reverse sampling grid and a short-term shading window are inserted. The light dissipation operation reduces specular reflection interference and stabilizes the code recognition process.
[0013] Preferably, the steps for generating the illumination frequency offset record table are as follows:
[0014] During the assembly, operation and maintenance of the movement, assembly lighting parameters, operating temperature parameters and maintenance record parameters are collected simultaneously and formed into continuous time series data according to fixed time intervals;
[0015] Based on the formed time series data, a unified time reference alignment process is performed to establish a correspondence between assembly lighting parameters, operating temperature parameters and maintenance record parameters on the same time axis, and form an associated data structure.
[0016] Based on the associated data structure, the light frequency change curve is extracted, and the light frequency state in different time periods is divided to form corresponding records of the light frequency stable stage and the light frequency shift stage.
[0017] Based on the integration of assembly lighting information, operating temperature change information, and maintenance record information during the illumination frequency offset stage, an illumination frequency offset record table indexed by time period is generated for subsequent exposure rhythm control and ghosting suppression.
[0018] Preferably, the illumination frequency shift information for each time period in the illumination frequency shift record table is established in a one-to-one correspondence with the operating temperature shift information and maintenance record information, and is arranged in a continuous time sequence so that the illumination frequency shift status can be continuously updated during the assembly, operation and maintenance process of the core, in order to reflect the dynamic change trajectory of the illumination frequency throughout its entire life cycle.
[0019] Preferably, the steps for generating the ghost trigger record are as follows:
[0020] Based on the light frequency offset record table, light frequency change data is extracted according to time period, and a time correspondence is established between the light frequency change data and the camera exposure start time, exposure duration and exposure end time.
[0021] After establishing the time correspondence, each time segment within the exposure cycle is compared and analyzed point by point to determine whether the exposure time is in the illuminated phase or the shadowed phase, and time segments where the illumination phase and exposure rhythm are inconsistent are marked.
[0022] The time sequence is organized based on the marked time segments, and the starting position and duration of the illumination phase misalignment interval are determined by combining the illumination frequency change trend.
[0023] Based on the illumination phase misalignment interval, a ghost trigger record is generated on the time axis to characterize the correspondence between illumination frequency shift and exposure rhythm deviation.
[0024] Preferably, the ghost trigger record is generated by synchronously associating the exposure start state, exposure duration state and illumination phase state, and mapping the change in reflection brightness and the reflection position on the movement surface to the time axis to form a continuous temporal identification record for subsequent ghost recognition and misjudgment tracing.
[0025] Preferably, the steps for generating the source bitmap that is misjudged are as follows:
[0026] Based on the ghost image triggering record, the etched image frames are traced back in time sequence to reconstruct a continuous image sequence of the etched outline changing with the phase of illumination;
[0027] After completing the backtracking of the etched images, texture analysis was performed on the edge regions of the etched codes in each time period to extract the brightness distribution and contour continuity features of the etched code edges, which were used to distinguish areas affected by illumination.
[0028] Based on the texture analysis results, the spatial backtracking of the engraving contour is performed to establish the spatial correspondence between the real engraving edge and the reflected ghost image edge.
[0029] Based on the spatial correspondence and the ghost trigger record, the starting position of the error identification is determined, and a misjudgment source bitmap is generated within the coordinate range of the engraving area for subsequent exposure rhythm adjustment.
[0030] Preferably, each error identification starting point position in the error source bitmap is associated with corresponding illumination phase information, exposure start time information, and code edge offset direction information. Furthermore, the error source bitmap is arranged according to the time sequence of ghost trigger records, representing the temporal correspondence between code contour changes and illumination phase misalignment, and limiting the formation conditions of the error identification starting point.
[0031] Preferably, the steps for generating the exposure rhythm adjustment instruction are as follows:
[0032] Based on the analysis of the time series and spatial distribution information of the ghost image corresponding to the misjudged source bitmap, the misalignment interval between the illumination phase and the exposure time period is extracted.
[0033] The exposure pause sequence is determined based on the temporal distribution of the misalignment intervals, so that the exposure action is paused during the period of inconsistent illumination phase and resumed in the stable interval;
[0034] After determining the exposure pause sequence, the reverse phase sampling window is determined by combining the ghost trigger record and the illumination phase information in the misjudged source bitmap, which is used to adjust the correspondence between the exposure start time and the illumination period.
[0035] Based on the exposure pause sequence and the backslip suppression measures of the illumination reference point combined with the reverse phase sampling window, an exposure rhythm adjustment command is generated for dynamically correcting the camera exposure process.
[0036] Preferably, the exposure rhythm adjustment command is updated based on the real-time changes in the illumination cycle during execution, so that the exposure start time is continuously adjusted within the reverse phase sampling window, and the exposure start is delayed by the bounce suppression method when brightness fluctuations occur at the illumination reference point, so as to ensure that the exposure process avoids the illumination fluctuation range and maintains the timing stability of the code acquisition process.
[0037] Preferably, the online adjustment steps for the exposure phase are as follows:
[0038] The illumination cycle is divided into time segments based on the exposure rhythm adjustment command, the correspondence between the illumination phase, the dark phase and the exposure start time is determined, and the exposure phase is adjusted synchronously.
[0039] After completing the exposure phase synchronization adjustment, according to the exposure rhythm adjustment command, a reverse sampling grid is inserted within the time interval of the light illumination phase and the dark phase being misaligned, so that the exposure sampling points avoid the period of concentrated light energy change.
[0040] After the reverse sampling grid is inserted, a short-term shading window operation is performed within the illumination misalignment time interval to reduce the local strong reflection light formed by the mirror reflection on the movement surface;
[0041] After the short-term shading window ends, a light dissipation operation is performed to smoothly transition the light intensity on the time axis. Combined with the exposure rhythm adjustment command, the subsequent exposure phase is updated, thereby stabilizing the code recognition process.
[0042] The technical effects and advantages provided by the present invention in the above technical solution are as follows:
[0043] This invention establishes a dynamic correlation control mechanism between the illumination period and the exposure rhythm, enabling the exposure process to respond in real-time to changes in the illumination frequency, thus fundamentally eliminating the ghosting problem caused by the misalignment of the illumination and dark phases. Through the synergistic action of the light frequency monitoring module and the exposure comparison analysis module, continuous tracking of the illumination frequency change trend and adaptive adjustment of the exposure rhythm are achieved, ensuring that the camera sampling process remains within a stable illumination range. This method effectively avoids exposure misalignment caused by illumination period drift, maintaining texture stability and edge continuity in the camera body's marking area during image acquisition, thereby improving the accuracy and reliability of marking recognition.
[0044] This invention achieves self-learning optimization of exposure control by combining a ghost image recognition and backtracking module with a rhythm control core module. This allows the exposure rhythm adjustment command to continuously correct the exposure phase based on the misjudged source bitmap during each detection. Through the coordination of online exposure correction and illumination dissipation operations, reflection interference is effectively reduced, and the brightness distribution of the specular reflection area remains uniform. This method achieves a dynamic balance between illumination and exposure during code recognition, making the image brightness levels more stable and edge features clearer. This ensures the continuity and stability of the mechanism traceability code recognition and improves the optical consistency throughout the entire detection process. Attached Figure Description
[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0046] Figure 1 This is a schematic diagram of the modules of the AI-driven intelligent management system for the entire lifecycle of a watch movement according to the present invention. Detailed Implementation
[0047] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.
[0048] This invention provides, for example Figure 1 The AI-driven intelligent management system for the entire lifecycle of a watch movement, as shown, includes a light frequency monitoring module, an exposure comparison and analysis module, a ghost image recognition and retrospection module, a rhythm control core module, and an exposure correction and execution module.
[0049] The light frequency monitoring module collects assembly lighting data, operating temperature data, and maintenance record data during the entire life cycle management of the watch movement. Based on the collected data, it tracks the changing trend of light frequency and generates a light frequency offset record table.
[0050] In the management process of the entire lifecycle of a watch movement, a progressive data acquisition and analysis method is adopted to achieve continuous tracking of light frequency and generation of light frequency offset records. The entire process unfolds chronologically, transforming three types of data—assembly lighting, operating temperature, and maintenance records—into time-correlated light frequency offset records through continuous monitoring, phased processing, correlation analysis, and result modeling. This enables dynamic mapping of light frequency throughout the entire process of movement manufacturing, operation, and maintenance. The specific implementation steps are as follows:
[0051] During the movement assembly stage, multi-source data acquisition nodes are set up. Through lighting detection devices, temperature sensing elements, and information acquisition terminals, assembly lighting parameters, movement operating temperature parameters, and maintenance operation parameters are recorded synchronously. The acquisition of assembly lighting parameters includes the luminous intensity, emission frequency, illumination angle, uniformity of light distribution, reflective surface brightness distribution, and the current-driven fluctuation curve of the light source at the assembly station. The acquisition of movement operating temperature parameters includes the internal temperature rise rate of the movement, the temperature distribution along the surface heat conduction path, changes in ambient temperature in the assembly area, and the thermal diffusion range of the light source's heat output. The acquisition of maintenance operation parameters includes light source replacement time points, reflector cleaning cycles, lubricant replacement records, illuminance changes during parts loading and unloading, and light source output stability test data after each maintenance. During data acquisition, lighting, temperature, and maintenance information are continuously recorded at fixed time intervals to ensure that data from different sources remain synchronized on the same timeline, forming a continuous time-series dataset. This provides a foundation for tracking subsequent trends in illumination frequency changes.
[0052] The collected data on assembly lighting, operating temperature, and maintenance records were time-series processed according to a unified time base. All data were indexed by sampling timestamps, forming a multi-dimensional data matrix. During this process, time axis correction was used to align lighting intensity changes and temperature response data at the same time points, ensuring data correspondence within the same time period. Subsequently, the delayed response relationship between light intensity changes and temperature changes was analyzed to identify the impact of light source power changes on thermal stability. For maintenance operations in each time period, the light frequency fluctuation range corresponding to light source adjustments and reflection path changes was retrieved, establishing a dynamic link between the lighting environment and maintenance actions. In the data processing phase, the periodic variation pattern of light frequency, temperature fluctuation characteristics, and the recording sequence of maintenance operations were linked in chronological order to form an integrated data structure, enabling a complete presentation of light frequency variation trends based on multi-parameter linkage.
[0053] The time-series data is divided into stages and the trend of light frequency variation is extracted. By comparing the light frequency variation curves of different time periods, stable light frequency stages and light frequency deviation stages are identified. A stable light frequency stage indicates that the light source's emission period is consistent with the camera's exposure rhythm, and the light fluctuation amplitude is within the allowable range. A light frequency deviation stage indicates that the light period has experienced a periodic drift or phase change relative to the previous stage. In each deviation stage, the direction of light frequency change, fluctuation amplitude, duration, and corresponding temperature change trend are recorded. Combined with the characteristics of operating temperature changes, periodic light frequency deviations caused by thermal expansion or light source-driven fluctuations can be identified. Simultaneously, by using the time stamps of maintenance records, light source maintenance, replacement, cleaning operations, or lubrication adjustments can be correlated with light frequency deviation events, thereby identifying the impact range of each maintenance action on light frequency changes. In this way, a multi-dimensional correspondence is established between light frequency changes, operating temperature, and maintenance records, enabling the traceability and interpretation of light frequency deviation behavior in each time period.
[0054] An illumination frequency offset record table is generated based on multidimensional correspondences. This table is indexed by time period, integrating data from assembly lighting, operating temperature, and maintenance records into fields, and establishing hourly records of illumination frequency offset values. Each record unit includes the initial value, peak value, average rate of change, relative offset, corresponding temperature response parameters, and a description of the maintenance event for the illumination frequency. During generation, illumination frequency fluctuation curves, changes in light reflection paths, changes in illumination uniformity, changes in heat distribution, and light source maintenance operations are recorded in segments according to time sequence. Each time period also includes illumination reflection intensity, reflective surface brightness distribution coefficient, and light source operating duration information to describe the continuity of illumination stability. By integrating this information, the illumination frequency offset record table clearly displays the trajectory of illumination frequency changes over time and its offset state under different operating conditions, thus providing a data foundation for subsequent exposure rhythm control and ghosting suppression.
[0055] The exposure comparison and analysis module performs point-by-point comparison and analysis of the camera's exposure rhythm based on the illumination frequency offset record table, identifies time segments where the illumination phase of light and the dark phase of shadow are out of sync, and generates ghost trigger records on the time axis.
[0056] To achieve precise correlation analysis between camera exposure rhythm and illumination frequency, and to identify time segments where the illumination phase of light and the shadow phase of shadow are out of sync, thereby generating ghost image trigger records on the timeline, the specific implementation steps are as follows:
[0057] After obtaining the illumination frequency offset record table, the initial, peak, average rate of change, and relative offset of the recorded illumination frequency are extracted in segments, using time periods as the primary index. This establishes a one-to-one temporal correspondence between the illumination frequency data for each time period and the camera's exposure time parameters. In this way, the camera's exposure start time, exposure duration, and exposure end time within each exposure cycle are determined and mapped to the illumination cycle identified in the illumination frequency offset record table. This mapping relationship clarifies the illumination phase at each moment during the exposure process, determining whether the camera shutter opens in the bright phase of illumination or the dark phase of shadow. To ensure temporal matching between illumination and exposure, the entire mapping process is performed with a fixed time resolution, keeping the illumination frequency data and exposure rhythm synchronized on the same time base, thus laying the foundation for subsequent point-by-point comparative analysis.
[0058] After establishing the initial correspondence between illumination frequency and exposure time, each sampling point within the exposure cycle is analyzed point-by-point. Each sampling point corresponds to a specific time segment, which includes four parameters: illumination intensity, illumination phase, reflection intensity, and exposure shutter state. By comparing the illumination intensity in the illumination frequency offset record table with the actual exposure points on the exposure time axis, it is possible to identify whether the illumination phase and the camera exposure time are synchronized. When there is a phase shift between the illumination phase and the exposure rhythm, this shift will manifest as the illumination peak not coinciding with the exposure start point or the illumination trough not coinciding with the exposure end point. In this case, the start and end times of the shift interval are marked on the time axis for subsequent identification and recording of ghost triggers. To ensure the continuity of the analysis, each exposure cycle is compared with the illumination phase of the previous cycle to capture the periodic phase misalignment caused by illumination cycle drift, providing a time reference for ghost trigger recording.
[0059] After completing the point-by-point comparative analysis, all time segments showing asynchronous illumination phases and shadow dark phases were categorized and extended chronologically. Each asynchronous time segment contained four types of data: illumination fluctuation characteristics, camera exposure status, changes in reflected brightness, and reflective characteristics of the camera body surface. By continuously organizing these time segments, a sequence of illumination phase misalignment intervals could be formed to describe the persistence of illumination and exposure asynchrony within a specific time period. During the organization process, the illumination frequency change trend of the corresponding time period in the illumination frequency offset record table was superimposed onto these intervals to identify the direction and magnitude of the illumination offset, thereby further determining the starting point and duration of the illumination phase misalignment. In this way, each illumination phase misalignment interval can be correlated with a specific exposure rhythm, providing a temporal basis for the subsequent generation of ghost trigger records.
[0060] After organizing the asynchronous intervals between illumination phase and shadow phase, ghost trigger records are established on the timeline. These records store the start and end times, illumination frequency offset, exposure start time, exposure duration, reflectance distribution, and camera exposure status for each illumination phase misalignment interval in a time-series format. Each ghost trigger record identifies the relative relationship between the camera's exposure status and the illumination cycle within the illumination misalignment interval, thus determining the probability and trigger time of ghosting during that period. During generation, all ghost trigger events are arranged chronologically, forming a record sequence on the timeline consisting of multiple consecutive ghost trigger points. Each trigger point includes illumination phase information, changes in light source reflectance intensity, and coordinates of the reflective area on the camera's surface, facilitating location and analysis in subsequent ghost identification and misjudgment tracing. In this way, the ghost trigger records not only reflect the temporal distribution of illumination and exposure asynchrony but also demonstrate the correlation between illumination frequency offset and exposure rhythm deviation.
[0061] The ghost image recognition and backtracking module records the contour change process of the code image based on ghost image triggering, analyzes the texture features of the code edge, distinguishes between the real code edge and the reflected ghost image edge, determines the starting position of the error recognition, and generates a bitmap of the misjudgment source.
[0062] To enable retrospective analysis of ghost image trigger records and to identify the difference between the real code edge and the reflected ghost image edge by combining time series and image contour change information, thereby determining the starting position of misidentification and generating a misjudgment source bitmap, the specific implementation steps are as follows:
[0063] After obtaining the ghost trigger records, the movement's marking image sequence is reconstructed retrospectively based on the included time series, illumination phase distribution, exposure start and end times, reflection brightness changes, and illumination shift amplitude. By reading the time period information in the ghost trigger records, each illumination phase record is matched one-to-one with the marking image frames captured by the camera, determining the specific acquisition time of each frame within the illumination phase and shadow phase. Subsequently, the image frame sequence is rearranged on the time axis according to the change in illumination phase, allowing the complete reproduction of the morphological changes of the marking edges within the illumination cycle. During the reconstruction process, the illumination intensity distribution area and reflection highlight area of each frame are extracted, and the illumination curve is superimposed on the marking outline to form a joint time curve of illumination change and edge displacement, thereby revealing the direct impact of illumination change on the marking outline morphology. Through this joint reconstruction of time and illumination, a continuous marking outline change trajectory can be obtained, providing a foundation for in-depth analysis of texture features.
[0064] Based on the image sequence obtained from retrospective reconstruction, texture detail analysis is performed on the code edge region for each time period. First, the grayscale gradient distribution of the code region is extracted in each frame, and the pixel band where the code edge is located is identified by brightness changes, recording its spatial distribution. Then, using the time series as a guide, code edge regions at the same location in consecutive frames are superimposed and compared to analyze the continuity of the edge contour, the smoothness of the edge direction, the uniformity of edge brightness transition, and the distribution density of local reflection points. If a certain edge region exhibits contour overlap, edge splitting, or brightness reversal during the alternation of illumination phase and shadow dark phase, it is determined that the region is affected by reflection interference. By further comparing the synchronization relationship between the illumination shift value in the ghost trigger record and the brightness fluctuation of the edge region, the specific time period of influence of illumination phase shift on edge changes can be confirmed, thus distinguishing the edge region with ghost images caused by illumination interference from the stable real code edge region. Through this process, the texture features, brightness patterns, and spatial continuity of the code edge are fully analyzed, providing data basis for the spatial localization of ghost image regions.
[0065] After identifying the edge regions affected by reflection interference, a spatial backtracking analysis is performed on the contour change trajectory of the code image to determine the spatial relationship between the real code edge and the ghost edge. Specifically, keyframes with alternating illumination and shadow phases are selected in the time series. The edge morphology of the same code region under different phases is superimposed in coordinates, and the offset direction and distance of the edge position in consecutive frames are calculated. If the edge undergoes periodic displacement along a fixed direction in different time periods, and its brightness distribution shows periodic enhancement and weakening, then the edge belongs to the ghost formation region. Through this spatial superposition analysis, the spatial offset path of the ghost edge relative to the real code edge can be plotted, thereby clarifying the generation position and diffusion direction of the ghost in the code image. Subsequently, this spatial offset path is correlated with the time period of ghost trigger recording, so that each illumination phase misalignment event corresponds one-to-one with the specific ghost formation position. Through this dual temporal and spatial correlation, the spatial range of ghost generation can be delineated within the entire code area, forming a clear boundary with the stable area of the real code edge.
[0066] After distinguishing between the real marking edges and the ghost image edges, a misjudgment source bitmap is generated based on spatial coordinates, time series, and lighting conditions. The misjudgment source bitmap uses the two-dimensional coordinate plane of the movement marking area as its foundation, marking the starting position of each misidentification as a point. Each marked point records the lighting phase, exposure start time, reflection brightness value, edge offset direction, lighting frequency offset amplitude, and surface reflectivity of the movement. To maintain temporal continuity, the misjudgment source bitmap is arranged according to the chronological order of ghost image triggering, with misjudgment source points within each time period aggregated and displayed according to spatial coordinates, thus forming a time chain reflecting the ghost image appearance process. The misjudgment source bitmap also includes a layer comparing the distribution of real and ghost image edges in the marking area, allowing the ghost image generation area and the real marking area to be clearly distinguished in the same image. Through this structured recording, the misjudgment source bitmap not only shows the specific location and time of the misidentification but also reflects the dynamic relationship between lighting frequency offset and ghost image generation, providing a precise basis for subsequent exposure rhythm phase adjustment and lighting dissipation processing.
[0067] The rhythm control core module determines the exposure pause sequence and reverse phase sampling window based on the misjudged source bitmap, and generates exposure rhythm adjustment instructions in combination with the bounce suppression measures of the illumination reference point, which are used to dynamically correct the camera exposure process.
[0068] After generating the misjudged source bitmap, in order to dynamically correct the exposure process based on the temporal and spatial distribution characteristics of ghost images, the relationship between the temporal sequence information, illumination phase information, reflection intensity data, and exposure time periods in the misjudged source bitmap is analyzed. This allows for the determination of the exposure pause sequence and the reverse phase sampling window, and the generation of exposure rhythm adjustment instructions in conjunction with bounce suppression measures at the illumination reference point. The specific implementation steps are as follows:
[0069] After obtaining the misjudged source bitmap, the recorded illumination phase, exposure time period, reflection brightness changes, and edge offset direction are analyzed to extract the time marker and spatial coordinate information of each misjudged source point. By analyzing the clustering characteristics of multiple misjudged source points within the same time period in the misjudged source bitmap, the temporal intervals in which ghost images appear can be identified. Each temporal interval represents the duration of the misalignment between camera exposure and illumination phase, and has parameters such as illumination offset, phase direction, and reflection intensity changes. To determine the exposure pause sequence, all temporal intervals are first arranged according to the illumination phase offset direction, so that intervals where the illumination phase is advanced or delayed are identified separately. Then, these intervals are sorted based on time, and the start and end times of the illumination and exposure asynchrony in each exposure cycle are marked. In this process, the reflection intensity change curve under the same illumination phase in the misjudged source bitmap is superimposed on the exposure timeline to determine the time offset relationship between the illumination phase and the exposure start point, providing a precise temporal basis for the subsequent exposure pause sequence.
[0070] After determining the offset relationship between illumination phase and exposure rhythm, the exposure pause sequence is determined based on the correspondence between time intervals and phase directions. The exposure pause sequence is a time sequence for periodically adjusting the camera's exposure action according to the misalignment pattern of light and dark phases within the illumination cycle. By referencing time periods in the source bitmap where ghosting occurs frequently, these time periods are marked as exposure pause nodes. At each exposure pause node, the exposure action is temporarily suspended to avoid the camera capturing images at moments of sudden changes in illumination reflection intensity or unstable illumination phase. The determination of the exposure pause sequence also considers the illumination fluctuation cycle in the illumination frequency offset record table, ensuring that the exposure pause nodes are staggered from the peak intervals of illumination fluctuations, thereby reducing reflection interference caused by sudden changes in illumination energy. In this way, a set of exposure pause periods can be formed on the time axis, allowing the camera to stop sampling in intervals of asynchronous illumination phases and resume normal exposure in stable illumination intervals, thus achieving dynamic matching of illumination and exposure.
[0071] After determining the exposure pause sequence, the setting method for the reverse phase sampling window is determined based on the ghost trigger record and the illumination phase information in the misjudged source bitmap. The purpose of the reverse phase sampling window is to ensure that the exposure sampling point falls into the stable phase of illumination when the illumination phase is misaligned, by delaying or advancing the camera exposure start time. In specific implementation, firstly, the alternation boundary between the light and dark phases is located within each illumination cycle. Based on the direction and magnitude of the illumination shift in the misjudged source bitmap, the time interval for adjusting the exposure phase is calculated. Then, this time interval is defined as the start and end range of the reverse phase sampling window. Within the reverse phase sampling window, the camera exposure action operates inversely to the illumination fluctuation cycle; that is, when the illumination phase is advanced, the exposure start time is delayed; when the illumination phase is delayed, the exposure start time is advanced. Through this phase inversion, the relative stability of exposure sampling can be maintained during the illumination fluctuation phase. To ensure the continuity of the reverse phase sampling window, at the end of each exposure cycle, the illumination phase shift of the previous cycle is smoothly connected to the exposure start time of the next cycle, thus forming a coherent time sequence for the reverse sampling window.
[0072] After establishing the exposure pause sequence and the reverse phase sampling window, an exposure rhythm adjustment command is generated by combining the backslip suppression measures of the illumination reference point. Illumination reference point backslip refers to the instantaneous brightness change caused by drive waveform disturbances during periodic switching of the light source. To prevent such sudden brightness changes from affecting exposure stability, a backslip suppression operation is embedded in the exposure rhythm adjustment command to achieve a smooth illumination transition. Specifically, an illumination reference point is selected at the beginning of each illumination cycle, and its brightness change curve is monitored. When a sudden increase or decrease in brightness is detected at the illumination reference point before or after phase transition, the exposure rhythm adjustment command automatically inserts a short delay, postponing the camera's exposure start time to avoid acquiring images during brightness fluctuations. Simultaneously, a smooth transition segment for illumination intensity is also embedded in the exposure rhythm adjustment command, gradually extending the exposure start window to weaken the instantaneous impact of illumination changes. The output of the exposure rhythm adjustment command is generated in the form of a time series, including the exposure pause node, the start and end times of the reverse phase sampling window, the illumination reference point brightness adjustment segment, and the backslip suppression delay amount. The entire instruction is invoked in real time during the camera exposure control process, enabling the exposure process to be dynamically corrected according to the light fluctuation state, thereby ensuring that the marking area maintains clear texture and stable edges under light reflection interference.
[0073] The exposure correction and execution module adjusts the exposure phase online based on the exposure rhythm adjustment command. It inserts a reverse sampling grid and a short-term shading window during the critical time period when the light illumination phase and dark phase are misaligned. It weakens specular reflection interference through light dissipation operation and stabilizes the code recognition process.
[0074] In the exposure control stage of optical inspection of watch movements, to achieve real-time coordination between the illumination period and the camera exposure phase, and to dynamically correct the exposure process based on the misalignment of the illuminated and dark phases, the interference caused by specular reflection is gradually reduced by executing exposure rhythm adjustment commands, combined with illumination period division, insertion of reverse sampling grids, adjustment of short-term light-blocking windows, and continuous execution of light dissipation operations. This ensures the stability and edge sharpness of the engraving recognition image. The specific implementation steps are as follows:
[0075] Upon receiving the exposure rhythm adjustment command, the current illumination cycle is analyzed in real time. Based on the start and end times of the illumination phase, the transition time of the dark phase, and the exposure pause node information marked in the command, a complete illumination cycle is divided into continuous time segments. Each time segment corresponds to a specific illumination range, light source intensity, light incident angle, and surface reflection state. By establishing the correspondence between the illumination phase and the dark phase on the time axis, the offset between the camera's exposure start point and the illumination cycle can be clearly defined. After determining the offset, the camera's exposure start time and exposure end time are gradually adjusted according to the phase correction parameters provided in the exposure rhythm adjustment command, so that the exposure process is realigned with the illumination cycle in the time dimension. During this stage, the change curve of the light source brightness over time is recorded by continuously monitoring the illumination reference point. Using this curve as a reference, the exposure start time is ensured to be in the middle region of the stable illumination phase, thereby avoiding interference from unstable light intensity at the rising or falling ends of the illumination phase on the exposure uniformity. After this adjustment, the camera's exposure phase and the illumination cycle are synchronized, providing a time reference for subsequent backsampling and the insertion of the shading window.
[0076] After completing the synchronous adjustment of the exposure phase, a reverse sampling grid is deployed within the critical interval where the illumination and dark phases are misaligned, according to the start and end times of the reverse phase sampling window defined in the exposure rhythm adjustment command. The reverse sampling grid alters the temporal distribution of exposure sampling points, placing them in the reverse region of the illumination cycle, thus avoiding the periods of strongest or most drastic changes in light energy. In practice, the peak and trough positions of the light energy curve are first determined based on the light cycle fluctuation patterns recorded in the light frequency offset log. When the illumination phase precedes the camera's exposure cycle, the reverse sampling grid shifts towards the dark phase, covering the stable region where light energy gradually decreases. When the illumination phase lags behind the exposure cycle, the reverse sampling grid shifts towards the exposure position, placing the sampling points in the stage where light energy is rising but has not yet reached its peak. During grid distribution, the time interval between each sampling point is set according to the rate of change in light energy to ensure that the exposure sampling covers the energy stability segment within the entire misalignment interval. By using this reverse distribution method, the exposure process can avoid high-energy reflections during the period of most intense light fluctuations, thereby reducing the impact of reflection interference on the etched image and ensuring consistent image brightness distribution.
[0077] After the reverse sampling grid is inserted, to further reduce the specular reflection of the metal surface of the camera body under strong light, a short-term shading window operation is performed within the critical area of illumination misalignment. The short-term shading window temporarily reduces the light intensity or blocks part of the incident light path, causing an instantaneous reduction in light energy density on the camera body surface, thereby changing the distribution of reflected light. Specifically, the peak time of the illumination phase is detected within each illumination cycle, and the start and end points of the shading window are set at its leading and trailing edges, respectively. The duration of the shading window is determined based on the illumination cycle length and the rate of change of light energy, typically covering the transition period before and after the light peak. During the shading window's operation, the light intensity gradually decreases according to a preset ratio, causing the reflected light from the camera body surface to transition from a concentrated reflection state to a diffused state. To avoid image sampling breaks caused by the shading operation, the exposure control section maintains the continuity of the exposure rhythm, allowing the camera to continuously acquire images during the shading window operation. By dynamically adjusting the light-blocking window, the brightness of the reflective area on the movement surface is reduced, the light transition at the edge of the engraving is smoother, and the grayscale level of the image is kept stable, creating continuous transition conditions for subsequent light dissipation operations.
[0078] After completing the short-term shading window operation, the light dissipation phase begins. By gradually controlling the distribution of light source intensity, the light energy transitions naturally in time and space, thus completely reducing specular reflection interference. The light dissipation operation is performed within the transition period after the shading window is closed. Its main purpose is to ensure a smooth transition of light from the dark phase to the bright phase. Specifically, the end time of the dark phase is extended within the lighting cycle, resulting in a linear increase in light intensity during the recovery phase, with the light source brightness gradually transitioning from the end of the dark phase to the beginning of the bright phase. During the light recovery period, the energy density of the reflective areas on the movement surface gradually disperses, expanding locally reflected beams into wider reflective surfaces, and the reflection angle distribution tends to be more even. Simultaneously, the exposure rhythm adjustment command updates the exposure start time of the next cycle during the light dissipation operation, ensuring that the exposure start time avoids the period before the light energy is fully stable. In this way, light intensity fluctuations are weakened, and the exposure rhythm and lighting cycle achieve a coordinated dynamic balance. As the light dissipates, the distribution of reflected light on the movement surface becomes more uniform, the brightness transition in the marking area is natural, the edge features remain stable, the image texture is clear, and the marking recognition accuracy is improved.
[0079] This invention establishes a dynamic correlation control mechanism between the illumination period and the exposure rhythm, enabling the exposure process to respond in real-time to changes in the illumination frequency, thus fundamentally eliminating the ghosting problem caused by the misalignment of the illumination and dark phases. Through the synergistic action of the light frequency monitoring module and the exposure comparison analysis module, continuous tracking of the illumination frequency change trend and adaptive adjustment of the exposure rhythm are achieved, ensuring that the camera sampling process remains within a stable illumination range. This method effectively avoids exposure misalignment caused by illumination period drift, maintaining texture stability and edge continuity in the camera body's marking area during image acquisition, thereby improving the accuracy and reliability of marking recognition.
[0080] This invention achieves self-learning optimization of exposure control by combining a ghost image recognition and backtracking module with a rhythm control core module. This allows the exposure rhythm adjustment command to continuously correct the exposure phase based on the misjudged source bitmap during each detection. Through the coordination of online exposure correction and illumination dissipation operations, reflection interference is effectively reduced, and the brightness distribution of the specular reflection area remains uniform. This method achieves a dynamic balance between illumination and exposure during code recognition, making the image brightness levels more stable and edge features clearer. This ensures the continuity and stability of the mechanism traceability code recognition and improves the optical consistency throughout the entire detection process.
[0081] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.
Claims
1. An AI-driven watch movement full life cycle intelligent management system, characterized in that, It includes a light frequency monitoring module, an exposure comparison and analysis module, a ghost image recognition and retrospection module, a rhythm control kernel module, and an exposure correction and execution module: The light frequency monitoring module collects assembly lighting data, operating temperature data, and maintenance record data during the entire life cycle management of the watch movement. Based on the collected data, it tracks the changing trend of light frequency and generates a light frequency offset record table. Among them, the characteristics of operating temperature change are used to identify periodic light frequency offsets caused by thermal expansion or light source driving fluctuations. The exposure comparison and analysis module performs point-by-point comparison and analysis of the camera's exposure rhythm based on the illumination frequency offset record table, identifies time segments where the illumination phase and the illumination dark phase are out of sync, and generates ghost trigger records on the time axis. The ghost image recognition and backtracking module records the contour change process of the code image based on ghost image triggering, analyzes the texture features of the code edge, distinguishes between the real code edge and the reflected ghost image edge, determines the starting position of the error recognition, and generates a bitmap of the misjudgment source. The rhythm control core module determines the exposure pause sequence and reverse phase sampling window based on the misjudged source bitmap, and generates exposure rhythm adjustment instructions in combination with the bounce suppression measures of the illumination reference point, which are used to dynamically correct the camera exposure process. The exposure correction and execution module adjusts the exposure phase online based on the exposure rhythm adjustment command. It inserts a reverse sampling grid and a short-term shading window during the critical time period when the illumination phase and the dark phase are misaligned. It weakens specular reflection interference through illumination dissipation operation and stabilizes the code recognition process. The steps for generating exposure rhythm adjustment instructions are as follows: Based on the analysis of the time series and spatial distribution information of the ghost image corresponding to the misjudged source bitmap, the misalignment interval between the illumination phase and the exposure time period is extracted. The exposure pause sequence is determined based on the temporal distribution of the misalignment intervals, so that the exposure action is paused during the period of inconsistent illumination phase and resumed in the stable interval; After determining the exposure pause sequence, the reverse phase sampling window is determined by combining the ghost trigger record and the illumination phase information in the misjudged source bitmap. This window is used to adjust the correspondence between the exposure start time and the illumination period. Within the reverse phase sampling window, the camera exposure action operates in the opposite direction to the illumination fluctuation period. That is, when the illumination phase is advanced, the exposure start time is delayed; when the illumination phase is delayed, the exposure start time is advanced. After establishing the exposure pause sequence and the reverse phase sampling window, a lighting reference point is selected. When a sudden increase or decrease in brightness is detected at the lighting reference point before and after the phase transition, a bounce suppression operation is embedded to achieve a smooth lighting transition. Finally, an exposure rhythm adjustment command is generated to dynamically correct the camera's exposure process. The output of the exposure rhythm adjustment command is generated in the form of a time series, which includes the exposure pause node, the start and end time of the reverse phase sampling window, the brightness adjustment segment of the lighting reference point, and the bounce suppression delay. 2.The AI-driven watch movement full life cycle intelligent management system according to claim 1, characterized in that, The steps for generating the illumination frequency offset record table are as follows: During the assembly, operation and maintenance of the movement, assembly lighting parameters, operating temperature parameters and maintenance record parameters are collected simultaneously and formed into continuous time series data according to fixed time intervals; Based on the formed time series data, a unified time reference alignment process is performed to establish a correspondence between assembly lighting parameters, operating temperature parameters and maintenance record parameters on the same time axis, and form an associated data structure. Based on the associated data structure, the light frequency change curve is extracted, and the light frequency state in different time periods is divided to form corresponding records of the light frequency stable stage and the light frequency shift stage. Based on the integration of assembly lighting information, operating temperature change information, and maintenance record information during the illumination frequency offset stage, an illumination frequency offset record table indexed by time period is generated for subsequent exposure rhythm control and ghosting suppression. 3.The AI-driven watch movement full life cycle intelligent management system according to claim 2, characterized in that, The illumination frequency offset record table establishes a one-to-one correspondence between the illumination frequency change information for each time period, the operating temperature change information, and the maintenance record information. It is arranged in a continuous time sequence so that the illumination frequency offset status can be continuously updated during the assembly, operation, and maintenance of the core, reflecting the dynamic change trajectory of the illumination frequency throughout its entire life cycle. 4.The AI-driven watch movement full life cycle intelligent management system according to claim 2, characterized in that, The steps for generating the ghost trigger record are as follows: Based on the light frequency offset record table, light frequency change data is extracted according to time period, and a time correspondence is established between the light frequency change data and the camera exposure start time, exposure duration and exposure end time. After establishing the time correspondence, each time segment within the exposure cycle is compared and analyzed point by point to determine whether the exposure time is in the bright phase or dark phase of the light, and the time segments where the light phase and the exposure rhythm are inconsistent are marked. The time sequence is organized based on the marked time segments, and the starting position and duration of the illumination phase misalignment interval are determined by combining the illumination frequency change trend. Based on the illumination phase misalignment interval, a ghost trigger record is generated on the time axis to characterize the correspondence between illumination frequency shift and exposure rhythm deviation. 5.The AI-driven watch movement full life cycle intelligent management system according to claim 4, characterized in that, The ghost trigger record is generated by synchronously associating the exposure start state, exposure duration state and illumination phase state, and mapping the change in reflection brightness and the reflection position on the movement surface to the time axis to form a continuous time sequence identifier record for subsequent ghost identification and misjudgment tracing. 6.The AI-driven watch movement full life cycle intelligent management system according to claim 4, characterized in that, The steps for generating the misjudged source bitmap are as follows: Based on the ghost image triggering record, the etched image frames are traced back in time sequence to reconstruct a continuous image sequence of the etched outline changing with the phase of illumination; After completing the backtracking of the etched images, texture analysis was performed on the edge regions of the etched codes in each time period to extract the brightness distribution and contour continuity features of the etched code edges, which were used to distinguish areas affected by illumination. Based on the texture analysis results, the spatial backtracking of the engraving contour is performed to establish the spatial correspondence between the real engraving edge and the reflected ghost image edge. Based on the spatial correspondence and the ghost trigger record, the starting position of the error identification is determined, and a misjudgment source bitmap is generated within the coordinate range of the engraving area for subsequent exposure rhythm adjustment.
7. The AI-driven watch movement full life cycle intelligent management system according to claim 6, characterized in that, Each error identification starting point in the misidentification source bitmap is associated with corresponding illumination phase information, exposure start time information, and code edge offset direction information. Furthermore, the misidentification source bitmap is arranged according to the time sequence of ghost image trigger records, representing the temporal correspondence between code outline changes and illumination phase misalignment, and limiting the formation conditions of the error identification starting point.
8. The AI-driven watch movement full-lifecycle intelligent management system of claim 7, wherein, The exposure rhythm adjustment command is updated based on the real-time changes in the illumination cycle during execution, so that the exposure start time is continuously adjusted within the reverse phase sampling window. When brightness fluctuations occur at the illumination reference point, the exposure start is delayed through a bounce suppression method to ensure that the exposure process avoids the illumination fluctuation range and maintains the timing stability of the code acquisition process.
9. The AI-driven watch movement full-lifecycle intelligent management system of claim 8, wherein, The steps for adjusting the exposure phase online are as follows: The illumination cycle is divided into time segments based on the exposure rhythm adjustment command, the correspondence between the illumination phase, the illumination dark phase and the exposure start time is determined, and the exposure phase is adjusted synchronously. After completing the exposure phase synchronization adjustment, according to the exposure rhythm adjustment command, a reverse sampling grid is inserted in the time interval when the illumination phase and the darkness phase are misaligned, so that the exposure sampling points avoid the period when the light energy changes are concentrated. After the reverse sampling grid is inserted, a short-term shading window operation is performed within the illumination misalignment time interval to reduce the local strong reflection light formed by the mirror reflection on the movement surface; After the short-term shading window ends, a light dissipation operation is performed to smoothly transition the light intensity on the time axis. Combined with the exposure rhythm adjustment command, the subsequent exposure phase is updated, thereby stabilizing the code recognition process.