A chip mounting defect detection method fusing dual-band vision and light field regulation

Abstract

The application discloses a chip mounting defect detection method fusing dual-band vision and light field regulation and control, and relates to the technical field of intelligent detection, and comprises the following steps: collecting a visible light image sequence and a near-infrared image sequence, performing time synchronization on the visible light image sequence and the near-infrared image sequence, generating a light and shadow rhythm script according to a unified time scale, and forming a continuous light intensity change reference. According to the application, through dual-band synchronous collection and light and shadow rhythm control, the light intensity is continuously adjustable on a time axis, a light and shadow rhythm script and a delay list are established, the time correlation between light field changes and reflection paths is realized, ghost images and bright spots caused by multiple reflections are effectively inhibited, and the image signal-to-noise ratio is improved. Through time squeezing index guidance of non-equal-step shutters and reverse half-beat illumination adjustment, combined with dynamic blocking of late reflection signals by sliding dark windows, the light field energy distribution is balanced, the exposure rhythm is adaptively optimized, and therefore the accuracy of defect detection and the image stability are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection technology, specifically to a chip mounting defect detection method that integrates dual-band vision and light field modulation. Background Technology

[0002] Chip mounting defect detection integrating dual-band vision and light field modulation is a high-precision chip packaging inspection technology based on computer vision and artificial intelligence algorithms. This technology integrates visible light and near-infrared dual-band imaging units into the inspection system. Through a light field modulation algorithm within a computer vision framework, it controls the illumination angle and intensity of an adaptive light source array to perform directional compensation and extinction processing on metallic reflective areas, solder overflow areas, and pin reflection areas on the chip surface, thereby improving the signal-to-noise ratio and edge sharpness of the chip image. During imaging, the system relies on a six-axis collaborative positioning platform, combined with nanoscale linear drive and vibration suppression mechanisms, to achieve stable chip positioning and high-precision coordinate consistency during scanning. The detection algorithm employs a dual-band feature fusion model based on AI deep learning to perform multi-scale feature extraction and defect classification training on the chip image, enabling automatic identification and precise location of typical defects such as solder paste misalignment, wafer cracks, and pin misalignment within milliseconds. The system has cloud-based collaborative learning and parameter transfer capabilities, enabling the sharing of visual feature data among different chip mounting devices and achieving adaptive updates of detection parameters. It is suitable for online inspection of various chip packaging forms such as QFN and FCBGA, and builds a fully automated chip mounting quality assurance system that integrates high-resolution imaging, intelligent recognition, and real-time feedback.

[0003] The existing technology has the following shortcomings: In existing technologies, chip mounting defect detection integrating dual-band vision is prone to delayed frame superposition due to multiple reflections from reflective surfaces during light field modulation. When there are slight drifts in the incident angle of the light source, the roughness of the reflective surface, or the sampling timing, some bright reflected light will be repeatedly reflected between metal pads, pins, or frame edges, forming secondary reflection signals that are delayed in entering the imaging chip. This delayed signal will be superimposed on the main image, producing local overexposed bright spots or edge ghosting. The detection device is prone to misidentifying these as actual defects such as solder paste overflow or adhesion abnormalities, thus causing false alarms. This type of reflection interference has transient and random characteristics, and traditional imaging algorithms have difficulty distinguishing between real soldering abnormalities and optical artifacts in the time dimension, which can easily cause confusion in defect identification and affect the stability and accuracy of mounting detection. Summary of the Invention

[0004] The purpose of this invention is to provide a chip mounting defect detection method that integrates dual-band vision and light field modulation to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a chip mounting defect detection method integrating dual-band vision and light field modulation, comprising the following steps: Collect visible light image sequences and near-infrared image sequences and synchronize them in time. Generate a light and shadow rhythm draft based on a unified time scale to form a benchmark for changes in light intensity. Based on the light and shadow rhythm draft, detect the intensity change of bright spots, determine the time point and spatial location of late bright spots, perform inter-frame difference calculation on the dual-band image sequence, obtain the reflection path and generate a delay list; By time coupling the reflection path data in the delay list with the chip handling cycle data, the concentrated time period of the reflection signal is extracted and an error feature table is generated. Based on the error feature table, track the brightness intensity changes, identify overlapping areas in consecutive frames, extract repeated reflection areas, and calculate the time squeeze index; Based on the time-occupancy index, non-synchronous shutter speed and reverse half-shot brightness adjustment are performed, and a sliding dark window is inserted in the exposure window to block late reflection signals.

[0006] The preferred steps for generating a light and shadow rhythm draft are as follows: Acquire visible light image sequences and near-infrared image sequences within a unified time scale range, ensuring that the start time and duration of the visible light image acquisition channel and the near-infrared image acquisition channel are consistent in time. Based on a unified time scale, the visible light image sequence and the near-infrared image sequence are synchronized at the frame level, and each pair of images is formed into a corresponding frame under the same time scale and uniformly numbered. Illumination intensity information of consecutive frames is extracted from the time-synchronized joint image sequence. The brightness changes of visible light images and near-infrared images are compared frame by frame to form a light and shadow rhythm draft. Based on the light and shadow rhythm draft, normalization processing is performed to form a continuous light intensity change curve on the time scale, establishing the periodic fluctuation characteristics and phase change relationship of light in the time dimension, which is used as a time reference for light field interference analysis.

[0007] Preferably, during the generation of the light and shadow rhythm draft, the brightness changes of visible light images and near-infrared images are synchronously compared according to the time scale. The light and shadow rhythm curve is formed by the changes in light intensity under the continuous time scale, and the change pattern of brightness rise, decay and transition stages is recorded on the time axis, so that the light intensity forms a continuous distribution in the time dimension, which is used as a reference for the time analysis of light field interference.

[0008] Preferably, the steps for forming the delay list are as follows: Within the time scale of the light and shadow rhythm draft, the intensity of bright spot areas is detected for the visible light image sequence and near-infrared image sequence corresponding to each moment, the brightness distribution information in the continuous image frames is extracted and the light intensity mapping relationship is established. Based on the brightness intensity detection results, the trajectory of the brightness spot changes in a continuous time scale is analyzed to determine the time point and spatial location of the late-arriving brightness spot, and the directionality of the reflection path is determined by the spatial difference of the dual-band brightness spots. Given the time point and spatial location of the late bright spot, inter-frame difference calculation is performed on the visible light image sequence and the near-infrared image sequence to identify the incident angle and return trajectory of the reflection path and establish the spatial distribution relationship. Reflection path data is generated based on the incident angle and return trajectory of the reflection path. The data is then summarized in time and space to form a delay list. The delay list records the start and end times, spatial coordinates, brightness intensity, and direction offset parameters of the reflected signal, which are used for reflection behavior analysis.

[0009] Preferably, during the bright spot intensity detection process, the change in illumination intensity in the bright spot area is continuously monitored according to the time scale of the light and shadow rhythm draft. The reflected light that enters the imaging area late is identified by comparing the brightness change trend under adjacent time scales, and the time starting point and spatial coordinates of the reflection path are determined based on the sudden change in bright spot intensity, thereby ensuring the consistency of time information and spatial information in the reflection path data.

[0010] Preferably, the steps for generating the image error feature table are as follows: After obtaining the delay list, the reflection path data and the motion beat data of the chip handling process are aligned with the time reference, and each reflection path event and the corresponding time period motion beat information are arranged on a unified time scale to form a time sequence correspondence. Based on the time alignment results, analyze the time correspondence between the change in reflection angle in the reflection path data and the displacement phase in the motion beat, and determine the synchronization relationship between the change in reflection path direction and the chip motion phase; Based on the temporal correlation between the change in reflection angle and the beat phase, the coupled data is extracted to determine the time interval in which the reflected signal is concentrated and to correspond it with the corresponding motion phase. A false image feature table is generated based on the time interval in which the reflected signals occur, recording the energy distribution, spatial location distribution, and duration of bright spots, which is used to describe the distribution characteristics of reflection interference in the energy, spatial, and temporal dimensions.

[0011] Preferably, the determination of the time interval in which the reflected signals occur is achieved by tracking the continuity of the reflection angle change and the energy accumulation trend of the bright spot, classifying the energy enhancement period, reflection angle shift period and bright spot dissipation period of the reflected light, so that each group of concentrated reflection events in the false image feature table simultaneously records the start and end time range, the light intensity change curve and the spatial displacement trend, so as to extract the temporal distribution law of light field interference.

[0012] Preferably, the time-crowding index calculation steps are as follows: Based on the bright spot energy distribution information and spatial location distribution information in the false image feature table, the change of bright spot intensity over time is tracked frame by frame in the real-time detection image to construct a time chain of bright spot intensity change; Based on the time tracking results of the bright spot intensity, combined with the spatial distribution information of the previous frame image, the region overlapping with the previous frame image in the real-time detection image is identified and the trend of light intensity change is determined. Based on the identified overlapping areas, the existence time of bright spot areas in consecutive frames is statistically analyzed, repeated reflection areas are extracted, and the duration on the time axis is determined. Based on the temporal distribution of the repeated reflection areas, a time occupancy index is calculated to reflect the proportion of normal exposure time occupied by the reflected signal, and the time occupancy index is used as the control basis for light field rhythm adjustment.

[0013] Preferably, the calculation of the time occupancy index is based on the ratio of the duration of the reflected signal on the time axis to the normal exposure cycle. By accumulating the duration of the repeated reflection area, the occupancy ratio of the reflected signal within the exposure cycle is obtained, and the time occupancy index is used as a parameter for adjusting the light field rhythm to guide the synchronization and coordination of the exposure time interval and the illumination phase.

[0014] Preferably, the steps of performing non-synchronous shutter control based on the time-occupancy index, adaptively adjusting the exposure interval according to the intensity of reflection interference, and performing reverse half-shot brightness adjustment, inserting a sliding dark window to block late-arriving reflection signals during the interference period are as follows: After obtaining the time occupancy index, non-synchronous time control is performed on the exposure rhythm. The time interval between adjacent exposure cycles is adjusted according to the changing trend of the time occupancy index, so that the exposure interval is adaptively adjusted according to the intensity of reflection interference. Based on the adjusted exposure cycle, reverse half-photograph brightness rhythm adjustment is performed, and the phase of the illumination light and the time distribution of the reflected interference signal are reversed, so that the illumination light illuminates the detection area with the opposite phase during the reflected interference period. Based on the results of reverse half-photograph brightness rhythm adjustment, a sliding dark window is dynamically inserted in the exposure window to block reflected interference signals from entering the imaging area during the interference period; Based on the results of non-synchronous shutter rhythm control, reverse half-shot bright rhythm adjustment, and sliding dark window insertion, the light intensity distribution and exposure time allocation within the exposure cycle are coordinated and adjusted to form a dynamic balance of light field rhythm.

[0015] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention introduces dual-band synchronous acquisition and light-shadow rhythm control during the imaging process, enabling a continuous and adjustable exposure rhythm of light intensity over time. This allows for pre-adjustment and rhythm correction of the light field before reflection interference occurs. By establishing a light-shadow rhythm template and a delay list, changes in the light field and the reflection path are synchronously correlated in the time dimension, effectively suppressing multiple reflection signals in high-reflection areas. This prevents delayed light rays from superimposing on the main image, ensuring the sharpness and brightness balance of the imaging edges. This method fundamentally suppresses optical artifacts such as bright spots and ghosting in the detection image, improves the image signal-to-noise ratio, and provides a stable and reliable optical foundation for subsequent defect identification.

[0016] This invention guides non-synchronous shutter speed control with time-squeezing indexing and reverse half-shot illumination coordination, achieving a dynamic balance between exposure time and illumination phase. By inserting a sliding dark window within the exposure window, delayed reflection signals are effectively blocked, and the light field energy distribution becomes more uniform, thereby reducing false positives and improving defect identification accuracy. This light field control method can adaptively adjust the exposure speed according to the intensity of reflection interference, achieving temporal illumination optimization during the detection process, enabling the system to maintain stable image quality and consistent judgment during continuous detection. Attached Figure Description

[0017] 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.

[0018] Figure 1 This is a flowchart of a chip mounting defect detection method that integrates dual-band vision and light field modulation according to the present invention. Detailed Implementation

[0019] 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.

[0020] This invention provides, for example Figure 1The chip mounting defect detection method shown includes the following steps: (The method integrates dual-band vision and light field modulation) Visible light image sequences and near-infrared image sequences are acquired, and the visible light image sequences and near-infrared image sequences are synchronized in time. A light and shadow rhythm draft is generated according to a unified time scale to form a continuous light intensity change benchmark, which is used to establish a time series reference for light field interference analysis. To establish a temporal reference for light field interference analysis, an image acquisition and time synchronization method integrating dual-band vision is adopted. Visible light image sequences and near-infrared image sequences are continuously acquired, and a light and shadow rhythm draft is generated according to a unified time scale to form a temporal benchmark for continuous changes in illumination intensity, thereby enabling subsequent temporal dimension analysis of light field interference. The specific implementation steps are as follows: Visible light and near-infrared image sequences were acquired synchronously. Before acquisition, the time scale range and sampling interval were pre-determined to ensure that the start time and duration of the visible light and near-infrared image acquisition channels were consistent in time. During acquisition, a constant exposure rhythm was used to maintain stable illumination, ensuring that each frame of visible light and its corresponding near-infrared image was recorded at the same time point. Each frame contained the same light field incident conditions and environmental reflection state, making the dual-band images completely corresponding in the time dimension. The core of this step lies in achieving strict time unification, ensuring that the acquisition rhythms of the two types of imagery form a completely overlapping sequence in time, providing the original input data for generating the light and shadow rhythm draft.

[0021] The acquired visible light and near-infrared image sequences are synchronized at the frame level. During synchronization, using the unified time scale obtained in the previous step as a reference, the visible light and near-infrared image sequences are arranged chronologically, ensuring that each pair of images forms a corresponding set of frames at the same time scale. This one-to-one frame-level synchronization method ensures that the visible light and near-infrared images at any given time point share a common source of illumination change. During the synchronization process, each set of image frames is uniformly numbered with a time index, enabling subsequent illumination intensity change analysis to be tracked along a continuous time axis. By merging the images from the two bands into a single, consecutively numbered joint image sequence in the time dimension through frame-level synchronization, a consistent acquisition rhythm for the dual-band information is maintained throughout the entire time domain.

[0022] A light and shadow rhythm profile is generated based on the time-synchronized joint image sequence. In this process, illumination intensity information of consecutive frames is extracted according to a unified time scale, and the brightness changes of visible light and near-infrared images are compared frame by frame. By comparing the illumination intensity changes at adjacent time scales, the continuous distribution trend of illumination on the time axis is depicted, forming a light and shadow rhythm curve. This curve reflects the temporal rhythm of illumination changes in the dual-band images, including the patterns of brightness increases, decreases, and transition phases. Subsequently, this light and shadow rhythm curve is organized chronologically into a light and shadow rhythm profile, ensuring a continuous temporal distribution of illumination throughout the acquisition process. The light and shadow rhythm profile includes not only brightness change information from the visible light image sequence but also intensity change information from the near-infrared image sequence, thus fully expressing the rhythmic relationship of the dual-band light field within a unified time frame. This step enables a stable rhythmic baseline for illumination intensity in the time dimension, providing a continuous temporal reference for identifying light field interference.

[0023] A temporal reference for continuous illumination intensity variation is established based on the light and shadow rhythm draft. In this process, the light and shadow rhythm draft is normalized as a time series reference, resulting in a continuous and traceable variation curve of illumination intensity at different time scales. This temporal reference describes the periodic fluctuation characteristics and phase change relationship of illumination in the time dimension, providing a unified reference time axis for light field interference analysis. Based on this, the temporal reference can be used to locate the changes in bright spots appearing in subsequent images, thereby determining the incident period of interfering light and the occurrence time of reflection delay phenomena. Through this process, illumination forms a continuous and alignable time scale relationship throughout the entire acquisition cycle, allowing each frame of image to be analyzed under the same illumination intensity background. This temporal reference for continuous illumination intensity variation serves as the basic data for the entire detection process, providing an accurate time reference for subsequent temporal mapping and path analysis of light field interference.

[0024] Based on the light and shadow rhythm pattern, the intensity changes of bright spots at each moment are detected to determine the time and spatial location of late bright spots. Inter-frame difference calculation is performed on the visible light image sequence and the near-infrared image sequence to obtain the incident angle and return trajectory of the reflection path. Reflection path data is generated and a delay list is formed. The delay list is used as the basis for subsequent reflection behavior analysis. The spatiotemporal distribution information of late-arriving bright spots is obtained from the light and shadow rhythm draft. Continuous inter-frame difference processing is performed on visible light and near-infrared image sequences to identify the variation patterns of reflected light in time and space. Based on this, the incident angle and return trajectory of the reflection path are determined, thereby generating reflection path data and forming a delay list, providing a basis for subsequent temporal characteristic analysis of reflection behavior. The specific implementation steps are as follows: Based on the time scale of the light and shadow rhythm draft, the intensity of bright spot regions is detected in the visible light image sequence and near-infrared image sequence corresponding to each moment. By extracting brightness distribution information frame by frame in continuous image frames, the trend of light intensity change over time is observed to determine the formation and decay process of bright spots in the light and shadow rhythm draft. In this process, a continuous light intensity mapping relationship is established for the light distribution at each moment, so that each bright spot region has a unique intensity identifier on the time axis. In this way, the brightness changes in the light and shadow rhythm draft can be transformed into traceable time series features, thus providing a continuous illumination baseline for subsequent time series analysis. The key point of this step is to extract the intensity of bright spot regions at each moment, so as to establish a correspondence between the light and shadow rhythm draft and the image frames, providing a continuous illumination feature reference for the next stage of bright spot change identification.

[0025] Based on the brightness intensity detection results, the trajectory of the bright spots across continuous time scales is analyzed to determine the time and spatial location of late-arriving bright spots. By comparing the rising and falling trends of bright spot intensity at adjacent time scales, regions with abrupt and persistent intensity changes are identified, which are determined to be reflected light that enters the imaging range late. Combined with the time scale information from the light and shadow rhythm draft, the specific time point of each late-arriving bright spot is calibrated, and its spatial location is determined in both visible and near-infrared images. Due to the differences in reflectivity between visible and near-infrared light, late-arriving bright spots typically exhibit different spatial brightness distributions in the two bands. By comparing the spatial differences between the two bands, the directionality of the reflection path can be further determined. Through this process, the time and spatial coordinates of each late-arriving bright spot are determined, laying the data foundation for subsequent reflection path calculations. The result of this step is a set of late-arriving bright spot feature sets containing both temporal and spatial information, ensuring that the formation, persistence, and dissipation of bright spots are fully recorded in both temporal and spatial dimensions.

[0026] After obtaining the time point and spatial location of the late-arriving bright spot, the following steps are used to determine the incident angle and return trajectory of the reflection path from the visible light image sequence and the near-infrared image sequence: After determining the spatial position of each late-arriving bright spot, a spatial coordinate mapping relationship is established using pixel coordinates in the image plane. By pre-setting the proportional relationship between image coordinates and the physical plane of the detection area, the pixel positions in the image can be mapped to two-dimensional spatial coordinates in the actual detection plane. Based on this spatial coordinate mapping relationship, the pixel coordinates of the late-arriving bright spots at different time scales are converted into actual spatial position coordinates, thereby obtaining the motion trajectory of the bright spots in the detection plane. Through the spatial position changes at continuous time scales, the bright spot movement direction vector can be obtained, which is used to represent the propagation direction of reflected light in the imaging plane.

[0027] The spatial positions of bright spots at the same time scale are extracted from visible light and near-infrared images, respectively. By comparing the relative offsets of the bright spot center positions in the two bands, a spatial offset relationship between the two bands is established. Since visible light and near-infrared light propagate differently along the same reflection path, the bright spot positions in the two bands are synchronous in time but have a directional difference in space. By calculating the direction of the line connecting the two bright spot center positions, the projection direction of the reflected light onto the detection plane can be determined. Combined with the obtained bright spot motion direction vector, the projection angle of the incident light direction onto the detection plane can be determined, thus obtaining the directional expression of the incident angle in the two-dimensional plane.

[0028] After obtaining the planar projection of the incident direction, the reflection path's return trajectory is inferred by analyzing the variation of the bright spot intensity over time. When the reflected light enters the reflection stage from the incident stage, the bright spot intensity undergoes a change from concentrated to dispersed over continuous time, while the spatial position of the bright spot shifts in the opposite direction. By comparing the changes in the direction of movement of the bright spot position over continuous time, the turning point of the light in space can be determined. The direction vector before the turning point is defined as the incident path direction, and the direction vector after the turning point is defined as the return path direction, thus forming a complete return trajectory. This trajectory, connected by the spatial positions over continuous time, forms a broken line structure used to describe the propagation process of light within the detection area.

[0029] The incident direction and return direction data are recorded chronologically to form reflection path data. This data includes start time, incident direction vector, turning time, return direction vector, and end time. Each reflection path is numbered and organized to create a delay list. This delay list uses time as the primary index, categorizing reflection paths occurring within different time periods, ensuring each path has a clear temporal and spatial directional information. This delay list comprehensively reflects the propagation direction and return process of late-arriving reflected light in the detection area, providing a spatially directional data foundation for subsequent reflection behavior analysis.

[0030] After obtaining the time and spatial location of the late-arriving bright spots, inter-frame difference calculations are performed on the visible light image sequence and the near-infrared image sequence to identify the incident angle and return trajectory of the reflection path. By calculating the difference in brightness changes between consecutive frames, the transfer trend of illumination in the time dimension can be captured, thereby inferring the temporal process of light from incidence to return. Combining the changes in the spatial location of the bright spots obtained in the previous sub-step, the movement direction of the reflected signal in the image frame can be tracked in time sequence to determine the degree of offset of the incident angle of the light relative to the main direction of illumination. By comparing and analyzing the bright spot offsets in dual-band images, the return path of the light in different bands can be determined, thereby establishing a continuous distribution relationship of the reflection path in spatial coordinates. The implementation of this step makes each bright spot traceable in both time and space dimensions, and the incident direction and return trajectory of the reflection path can be determined, thus reconstructing the light propagation process in time sequence. Through continuous calculation of inter-frame difference, the static illumination changes in the light and shadow rhythm draft are transformed into dynamic light trajectories, giving the propagation path of light field interference a quantifiable delay characteristic on the time axis.

[0031] Based on the incident angle and return trajectory of the reflection path, reflection path data is generated and a delay list is formed. The reflection path data includes information such as the start time, end time, spatial start point, spatial end point, incident direction, and return direction for each reflection path. By organizing the reflection path data in chronological order, a distribution map of the delay signal throughout the entire acquisition period is constructed. Then, this data is systematically summarized according to the time and spatial axes to form a delay list. The delay list uses time as the primary index, grouping the delay characteristics of different reflection paths so that each reflection event has a unique correspondence in the time series. The delay list not only contains the temporal and spatial information of the reflection path data but also parameters such as bright spot intensity, reflection duration, and directional offset, used to analyze the temporal distribution characteristics of reflection behavior in subsequent steps. By generating the delay list, the light field interference process can be transformed from random illumination disturbances into a set of regular temporal events, providing basic data for the statistical analysis of reflection behavior. The result of this step establishes a closed-loop correspondence between the reflection path data and the light and shadow rhythm draft, thus fully quantifying the formation, duration, and attenuation processes of the delay signal in both time and space dimensions.

[0032] By using the reflection path data in the delay list and the motion beat data of the chip handling process, time coupling calculation is performed to analyze the correlation between the reflection angle change and the beat phase, extract the time periods in which the reflection signal is concentrated, and generate an error feature table. The error feature table includes bright spot energy distribution, spatial location distribution and duration frame information. This study reveals the formation patterns of reflected interference signals from both temporal and motion rhythm dimensions. By using reflection path data from the delay list and motion beat data from the chip handling process for time-coupled calculations, the correlation between reflection angle changes and beat phase is continuously analyzed. This extracts the time periods in which reflected signals are concentrated, further generating an image distortion feature table to describe the bright spot energy distribution, spatial location distribution, and duration frame information. The specific implementation steps are as follows: After obtaining the delay list, the reflection path data and the motion beat data of the chip handling process are aligned using a time reference. The delay list includes the start time, end time, incident direction, return direction, and changes in the spatial position of the bright spot for each reflection path, while the motion beat data records the chip's displacement sequence, rotation rhythm, and positioning duration during handling. By arranging both types of data uniformly on the same time scale, each reflection path event is mapped one-to-one with the motion beat information of the corresponding time period. At this point, each time node corresponds to a set of reflection path parameters and handling beat parameters, forming a time series correspondence. In this process, the time index in the delay list is expanded to cover the complete period range of the handling beat data, thereby establishing a continuous correlation chain between reflection events and motion rhythm. This time alignment method enables the reflection path and the handling process to interact in the time dimension, laying a unified time reference foundation for subsequent coupling analysis.

[0033] After time alignment, the temporal correspondence between the reflection angle change in the reflection path data and the displacement phase in the motion beat is analyzed. By observing the trend of the reflection angle change over time, the angular fluctuation range of the reflected light under different incident conditions can be obtained. Simultaneously, based on the displacement period and phase shift in the motion beat data, the motion state of the chip in each time segment can be identified. By temporally overlaying and comparing the reflection angle change curve with the motion beat phase curve, it can be determined whether the directional change of the reflection path is synchronous with the chip motion phase. If, within a specific phase interval, the reflection angle shows continuous change or the illumination intensity periodically increases, it indicates a direct coupling between the generation of reflected light and the chip motion. In this process, the incident angle change of each reflection path and the phase information of the motion beat are associated with the same time index, making the spatial shift process of the reflected light and the transport action time-linked. Through this sub-step, the dynamic correlation characteristics between the reflection angle change and the motion rhythm can be revealed, making light field interference no longer an isolated event, but a temporally coupled phenomenon that interacts with the transport beat.

[0034] After obtaining the temporal correlation between the reflection angle change and the cycle phase, time periods are extracted from the coupled data to determine the time intervals in which the reflected signals occur in concentrated bursts. By tracking the continuity of the reflection angle change and the cumulative trend of the bright spot energy, densely distributed intervals of reflection events on the time axis are identified. Combined with the repetition cycle of the transport cycle, these concentrated reflection periods are categorized, with each group of concentrated periods corresponding to a specific motion phase. Each concentrated reflection period includes a period of energy enhancement of reflected light, a period of reflection angle shift, and a period of bright spot dissipation, thus fully reflecting the formation and attenuation process of optical field interference. This time period extraction method clarifies the distribution pattern of reflection interference in the transport cycle and distinguishes between random and periodic reflections. At this point, the reflection path data in the delay list is recombined into a set of reflection events indexed primarily by time. Each set records the specific time periods and energy distribution of the concentrated reflection signals, thus providing basic data for subsequent feature extraction.

[0035] Based on the concentrated occurrence time periods of reflected signals, an image error feature table is generated to comprehensively describe the distribution characteristics of reflection interference in the energy, spatial, and temporal dimensions. In this step, the bright spot energy information, spatial location change trajectory, and duration frame number for each concentrated reflection time period are organized. The bright spot energy distribution describes the change characteristics of reflection intensity over time, the spatial location distribution shows the movement path of the bright spot in the imaging area, and the duration frame number information characterizes the duration of the reflected signal on the time axis. By uniformly summarizing this information, a complete image error feature table is constructed. The image error feature table is based on chronological order, with each record corresponding to a set of concentrated reflection events. The content includes the start and end time range of the reflected signal, the illumination intensity change curve, the spatial displacement trend, and the duration. Through this process, the temporal characteristics of reflection interference are systematically organized, and the formation and attenuation processes of all delayed signals in the light field are transformed into structured temporal data. The image error feature table not only presents the energy distribution pattern of the reflected signal but also reflects the persistence pattern of the interference signal at different spatial locations, enabling a complete quantitative expression of the light field interference phenomenon in both temporal and spatial dimensions.

[0036] Based on the generated error feature table, to reflect the temporal coupling relationship between the periods of concentrated reflection signals and the motion beat data of the chip handling process, while compiling the bright spot energy information, spatial position change trajectory, and duration of each concentrated reflection period, the corresponding motion beat data is simultaneously written into the error feature table. Specifically, for each concentrated reflection period, based on the time start and end range in the delay list, displacement stage information and beat phase information within the same time interval are extracted from the motion beat data, so that a one-to-one correspondence is established between concentrated reflection events and specific motion stages in the chip handling process.

[0037] In this process, each false image feature record includes not only the start and end time range of the reflected signal, the light intensity variation curve, the spatial displacement trend, and the duration, but also the corresponding motion beat phase interval, motion direction state, and beat transition node position. This time interval mapping method allows the spatial offset trend corresponding to the reflection angle change to be synchronized with the motion beat phase change, thereby revealing whether the concentrated occurrence of reflected signals occurs during the acceleration, deceleration, or position-holding phase of the transport process.

[0038] Furthermore, in the error feature table, the reflection angle change curve and the beat phase change curve are arranged side by side according to a unified time scale, so that the reflection angle value at each time scale and the corresponding beat phase value form a corresponding data pair. Through this correspondence, it is possible to observe on the time axis whether the reflection angle change and the beat phase transition node have synchronous characteristics, thereby clarifying whether the formation mechanism of the concentrated occurrence of reflection signals is coupled with the change of the motion beat.

[0039] Therefore, the error feature table not only describes the distribution characteristics of reflected interference in the energy, spatial, and temporal dimensions, but also records the correspondence between the concentrated periods of reflected signals and the phase of the motion beat, enabling a unified expression of the reflection path data in the delay list and the motion beat data of the chip handling process within the error feature table. Through this data integration method, the results of time-coupled calculations are fully reflected in the structure of the error feature table, making the correlation between reflection angle changes and beat phase an integral part of the error feature table, rather than merely existing as an analysis process.

[0040] Before performing real-time image tracking based on the image error feature table, the current motion beat phase of the real-time detected image is determined by first establishing the correspondence between the motion beat phase intervals and reflection concentration periods recorded in the image error feature table. Specifically, the time scale of the real-time detected image is matched with the start and end time ranges in the image error feature table to determine the corresponding motion beat phase interval and motion direction state of the image.

[0041] After completing the motion beat phase localization, the trend of the current bright spot intensity changing over time is phase-marked based on the correlation between the reflection angle change and the beat phase in the error feature table. That is, when calculating the bright spot intensity change curve, not only are the intensity values ​​of the bright spot recorded in consecutive frames, but the motion beat phase of that frame is also marked, so that the bright spot intensity change is synchronously expressed with the motion state. In this way, bright spot change patterns generated under different motion phases can be distinguished, thereby determining whether the bright spot change belongs to a periodic reflection phenomenon triggered by the motion beat.

[0042] Furthermore, when identifying areas overlapping with the previous frame, the distribution of reflection paths within the same beat phase interval in the image error feature table is used as a reference. When the position of the bright spot in the current image matches the spatial displacement trend under the corresponding beat phase recorded in the image error feature table, the area is marked as a beat-related repeated reflection area; when the bright spot change does not appear within the corresponding beat phase interval, it is marked as a non-beat-related illumination change. This distinction method ensures that the determination of repeated reflection areas is no longer based solely on bright spot intensity changes, but rather combines motion beat phase information for filtering.

[0043] When calculating the time-occupancy index, the duration of the beat phase interval in the image error feature table is used as a weighting factor. Specifically, the time proportion of consecutive bright spot overlap frames within the same beat phase interval is normalized according to the duration of that phase interval, so that the time-occupancy index reflects the "actual proportion of the exposure cycle occupied by the beat-related reflection signal". In this way, the time-occupancy index not only includes bright spot intensity variation data, but also beat phase coupling relationship, thus making the time-occupancy index a result of the combined effect of reflection angle change and motion beat.

[0044] Therefore, in the subsequent adjustment of the light field rhythm, the adjustment of the non-synchronous shutter speed rhythm is based not only on the changes in the intensity of the bright spots, but also on the phase misalignment control based on the distribution of the beat phase intervals recorded in the error feature table, so that the exposure rhythm avoids the motion phase in which the reflection signal is concentrated. At the same time, the trigger point of the reverse half-shot illumination is also set according to the position of the beat transition node in the error feature table, so that the illumination rhythm and the motion beat form a phase-reverse relationship.

[0045] Based on the false image feature table, the intensity of bright spots changes over time in the real-time detection image, the area overlapping with the previous frame image is identified, the repeated reflection area is extracted and the time occupancy index is calculated to reflect the proportion of the reflection signal occupying the normal exposure time, and to serve as the control basis for light field rhythm adjustment. To continuously track light field interference over time and identify recurring bright spot areas caused by delayed reflections in real-time monitoring footage, the intensity of bright spots over time is continuously monitored based on an image error feature table. Areas overlapping with the previous frame are located in consecutive frames, and recurring reflection areas are further extracted and time occupancy indexes are calculated. This quantifies the proportion of reflection signals on the exposure time axis, providing a control basis for light field rhythm adjustment. The specific implementation steps are as follows: Based on the bright spot energy distribution and spatial location distribution information in the false image feature table, the intensity change of bright spots over time is tracked frame by frame in the real-time detection image. The false image feature table contains the energy peak, duration (number of frames), and spatial coordinate distribution of each reflection event. By mapping these data to the illumination intensity distribution of the real-time detection image, each bright spot is mapped to its historical features in a time series. In consecutive frames of the detection image, the illumination intensity change curve of the bright spot region is extracted for each time scale, and its increasing or decreasing trend between consecutive frames is observed. Through this continuous tracking method, a time chain of bright spot intensity changes can be constructed in the time dimension, giving each bright spot a clear temporal location throughout its appearance, enhancement, and decay. In this process, the false image feature table provides an initial energy reference, making the tracking results continuous and comparative, ensuring that the bright spot intensity change accurately reflects the true evolution of the reflected signal. This step, through temporally continuous brightness comparison, extends the temporal behavior of reflected light in the real-time image, providing a continuous illumination trajectory for subsequent spatial overlap identification.

[0046] After completing the time tracking of bright spot intensity, the spatial distribution information of the previous frame is combined to identify the region overlapping with the previous frame in the real-time detection image. By analyzing the trend of bright spot position changes between adjacent frames, the continuous spatial movement path of the bright spot is compared to determine whether there is an illumination region overlapping with the previous frame in the current frame. When the bright spot occupies a similar spatial range in two consecutive frames and the change in illumination intensity remains relatively consistent, the region can be identified as a repeated superposition region of reflected light. At this time, the spatial boundary of this region in consecutive frames is uniformly marked, forming a continuous illumination occupancy segment on the time axis. In this way, the reflected light that enters the imaging area late can be distinguished from the residual illumination information of the previous frame, thereby identifying repeated bright spots caused by multiple reflections. Based on this, a spatial overlap mapping between consecutive frames is formed, so that the reflected signal is clearly labeled in the spatial dimension, providing an accurate spatial range for subsequent extraction of repeated reflection regions.

[0047] After identifying regions overlapping with the previous frame, continuous temporal analysis is performed on these spatially overlapping regions to extract repeating reflection regions and determine their duration along the time axis. By statistically analyzing the existence time of bright spot regions in consecutive frames, bright spot regions that appear continuously and maintain a stable spatial position are categorized as repeating reflection regions. Within repeating reflection regions, the light intensity typically exhibits periodic variations, meaning it recurs in adjacent exposure cycles but with varying intensities. By tracking these variations, the temporal distribution density and duration of reflected light can be obtained, thus determining the proportion of reflected signal occupancy within the exposure time. Each repeating reflection region corresponds to a time interval and spatial range, and its start and end times can be marked in the time series. In this way, a continuation chain of reflected signals in the temporal dimension is formed, giving each repeating reflection event a clear temporal boundary definition. This step transforms spatial overlap into temporal occupancy characteristics, laying the foundation for calculating the impact of reflected signals on normal exposure time.

[0048] Based on the extracted repeating reflection regions, a time occupancy index is calculated to reflect the proportion of normal exposure time occupied by the reflected signal. The time occupancy index is calculated based on the ratio of the duration of the reflected signal on the time axis to the normal exposure cycle. This is achieved by accumulating the duration of all repeating reflection regions to obtain the proportion of the reflected signal within one exposure cycle. The time occupancy index not only reflects the sustained characteristics of the delayed entry of reflected light into the imaging region but also characterizes its interference with the effective exposure time. A larger time occupancy index indicates that the reflected signal occupies more time within the exposure cycle, potentially leading to image superimposition and overexposure. Conversely, a smaller time occupancy index effectively suppresses reflection interference, resulting in higher utilization of exposure time. During the calculation, the time occupancy proportions of each reflection event are merged based on the reflection signal time period information, bright spot energy, and spatial distribution data recorded in the image error feature table. This ensures that the final time occupancy index accurately reflects the light field interference situation throughout the entire detection cycle. This index serves as the control basis for light field rhythm adjustment, guiding the synchronous adjustment of subsequent exposure rhythm and illumination phase to ensure that light field interference is coordinated and compensated in time.

[0049] Non-uniform shutter rhythm control is executed based on the time occupancy index, so that the interval between adjacent exposure cycles is adaptively adjusted according to the intensity of reflection interference. At the same time, reverse half-photograph brightness rhythm adjustment is executed, so that the illumination source illuminates the detection area with opposite phase during the reflection interference period, and a sliding dark window is dynamically inserted within the exposure window to block the late reflection signal, thereby achieving dynamic balance of light field rhythm and completing high-precision detection of chip mounting defects. To achieve coordinated control of exposure and illumination rhythms in the time dimension of the light field, a non-uniform shutter rhythm adjustment is performed based on a time-occupancy index. This adaptively adjusts the interval between adjacent exposure cycles according to the intensity of reflection interference. Simultaneously, a reverse half-shot illumination rhythm adjustment ensures that the illumination source illuminates the detection area with opposite phase during periods of concentrated reflection interference. Furthermore, a sliding dark window is dynamically inserted within the exposure window to block late-arriving reflection signals. This achieves a dynamic balance of light field rhythms on the time axis, ensuring that illumination intensity and exposure sequence remain coordinated and stable during image acquisition. The specific implementation steps are as follows: After obtaining the time occupancy index, non-synchronous time control is applied to the exposure rhythm to establish a dynamic correspondence between the exposure cycle and the intensity of reflection interference. The time occupancy index reflects the proportion of time occupied by the reflected signal within the exposure cycle; therefore, it is used as a key parameter for adjusting the rhythm during exposure control. Based on the changing trend of the time occupancy index, the time interval between consecutive exposure cycles is adjusted non-synchronously, shortening the exposure interval during periods of strong reflection interference and lengthening it during periods of weak reflection interference. This non-synchronous control method makes the exposure time distribution no longer fixed but adaptively changes according to the distribution characteristics of the interference signal. By adjusting the exposure interval frame by frame, the delayed signal of reflection interference can be made to avoid the main exposure window, thereby reducing the probability of late-arriving light entering the imaging area. This step, through the dynamic coupling between the time occupancy index and the exposure rhythm, enables the exposure time allocation process to have temporal adaptability, providing a basis for the coordination of subsequent illumination phases.

[0050] After completing the non-uniform shutter rhythm control, a reverse half-beat illumination rhythm adjustment is performed based on the adjusted exposure cycle. This ensures that the illumination source illuminates the detection area with the opposite phase during periods of reflection interference. The reverse half-beat illumination rhythm refers to reversing the phase of the illumination light and the temporal distribution of the reflection interference signal on the exposure time axis, causing the illumination peak and the occurrence time of the interference signal to be staggered by half a beat cycle. This phase-reversed illumination control method reduces the illumination intensity during periods of concentrated reflection interference, thereby weakening the incident energy of the reflected light, while increasing the illumination intensity during non-interference periods to maintain balanced image brightness. This step creates a temporal anti-phase match between the illumination cycle of the light source and the distribution rhythm of the reflection signal, ensuring that the reflected light enters the imaging area during the illumination energy decay phase, thus reducing its superimposed impact on the main exposure image. Through the reverse half-beat rhythm adjustment, the energy peak of the illumination field and the energy peak of the reflection signal are misaligned in time, thereby actively canceling out the interference energy and maintaining a balanced incident state of light within the exposure window.

[0051] Building upon the reverse half-shot exposure rhythm adjustment, a sliding dark window is dynamically inserted into the exposure window to further reduce the impact of late-arriving reflection signals on imaging. The sliding dark window refers to setting a movable shading period within the exposure cycle. This period dynamically slides according to the occurrence time of the reflection interference signal, temporarily closing the light-receiving channel during the concentrated interference period within the exposure window, thereby blocking late-arriving reflection signals from entering the imaging area. Through the dynamic insertion of the sliding dark window, illumination shielding in the temporal dimension can be achieved without changing the overall exposure time, preventing the delayed portion of reflected light from entering the imaging process. The time position and width of this dark window are adjusted in real-time according to the time occupancy index. When the reflection signal lasts a long time in the exposure cycle, the shading duration of the sliding dark window is correspondingly extended; when the reflection interference is sparsely distributed, the shading time of the dark window is correspondingly shortened. In this way, the effective time of the exposure window and the distribution of illumination interference complement each other, thereby achieving a coordinated unity between temporal shading and exposure. This step isolates the reflection signal in the temporal domain through dynamic shading, enabling the exposure window to achieve a self-balancing state of light field energy on the time axis.

[0052] After completing non-synchronous shutter speed control, reverse half-photograph brightness adjustment, and sliding dark window insertion, the light field rhythm of the entire exposure cycle is comprehensively adjusted to achieve a dynamic balance between illumination and exposure in the time dimension. By coordinating the light intensity distribution, exposure time allocation, and shading periods within each exposure cycle, the light field energy can exhibit a periodic and stable distribution on a continuous time axis. At this point, the occurrence of reflection interference signals and the peak of light source illumination mutually avoid each other, late reflection signals are effectively shielded by the sliding dark window, the effective light intensity of the exposure window remains uniform, and the illumination contrast during the imaging process is balanced. Through this process, the exposure rhythm and illumination rhythm are synchronously controlled in the time domain, ensuring that the light field maintains rhythmic consistency throughout the entire detection cycle. Ultimately, the incident light, reflection propagation, and image exposure form a coordinated closed loop in the time dimension, eliminating the delay effect of light field interference, and maintaining stable edge sharpness and contrast of the image, thereby achieving high-precision detection of chip mounting defects.

[0053] This invention introduces dual-band synchronous acquisition and light-shadow rhythm control during the imaging process, enabling a continuous and adjustable exposure rhythm of light intensity over time. This allows for pre-adjustment and rhythm correction of the light field before reflection interference occurs. By establishing a light-shadow rhythm template and a delay list, changes in the light field and the reflection path are synchronously correlated in the time dimension, effectively suppressing multiple reflection signals in high-reflection areas. This prevents delayed light rays from superimposing on the main image, ensuring the sharpness and brightness balance of the imaging edges. This method fundamentally suppresses optical artifacts such as bright spots and ghosting in the detection image, improves the image signal-to-noise ratio, and provides a stable and reliable optical foundation for subsequent defect identification.

[0054] This invention guides non-synchronous shutter speed control with time-squeezing indexing and reverse half-shot illumination coordination, achieving a dynamic balance between exposure time and illumination phase. By inserting a sliding dark window within the exposure window, delayed reflection signals are effectively blocked, and the light field energy distribution becomes more uniform, thereby reducing false positives and improving defect identification accuracy. This light field control method can adaptively adjust the exposure speed according to the intensity of reflection interference, achieving temporal illumination optimization during the detection process, enabling the system to maintain stable image quality and consistent judgment during continuous detection.

[0055] 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. A chip mounting defect detection method integrating dual-band vision and light field modulation, characterized in that, Includes the following steps: Collect visible light image sequences and near-infrared image sequences and synchronize them in time. Generate a light and shadow rhythm draft based on a unified time scale to form a benchmark for changes in light intensity. Based on the light and shadow rhythm draft, detect the intensity change of bright spots, determine the time point and spatial location of late bright spots, perform inter-frame difference calculation on the dual-band image sequence, obtain the reflection path and generate a delay list; By time coupling the reflection path data in the delay list with the chip handling cycle data, the concentrated time period of the reflection signal is extracted and an error feature table is generated. Based on the error feature table, track the brightness intensity changes, identify overlapping areas in consecutive frames, extract repeated reflection areas, and calculate the time squeeze index; Based on the time-occupancy index, non-synchronous shutter speed and reverse half-shot brightness adjustment are performed, and a sliding dark window is inserted in the exposure window to block late reflection signals.

2. The chip mounting defect detection method integrating dual-band vision and light field modulation according to claim 1, characterized in that, The steps to generate a light and shadow rhythm draft are as follows: Acquire visible light image sequences and near-infrared image sequences within a unified time scale range, ensuring that the start time and duration of the visible light image acquisition channel and the near-infrared image acquisition channel are consistent in time. Based on a unified time scale, the visible light image sequence and the near-infrared image sequence are synchronized at the frame level, and each pair of images is formed into a corresponding frame under the same time scale and uniformly numbered. Illumination intensity information of consecutive frames is extracted from the time-synchronized joint image sequence. The brightness changes of visible light images and near-infrared images are compared frame by frame to form a light and shadow rhythm draft. Based on the light and shadow rhythm draft, normalization processing is performed to form a continuous light intensity change curve on the time scale, establishing the periodic fluctuation characteristics and phase change relationship of light in the time dimension.

3. The chip mounting defect detection method integrating dual-band vision and light field modulation according to claim 2, characterized in that, During the generation of the light and shadow rhythm draft, the brightness changes of visible light images and near-infrared images are synchronously compared according to the time scale. The light and shadow rhythm curve is formed by the changes in light intensity under the continuous time scale, and the change pattern of brightness rise, decay and transition stages is recorded on the time axis, so that the light intensity forms a continuous distribution in the time dimension.

4. The chip mounting defect detection method integrating dual-band vision and light field modulation according to claim 2, characterized in that, The steps for creating the delay list are as follows: Within the time scale of the light and shadow rhythm draft, the intensity of bright spot areas is detected for the visible light image sequence and near-infrared image sequence corresponding to each moment, the brightness distribution information in the continuous image frames is extracted and the light intensity mapping relationship is established. Based on the brightness intensity detection results, the trajectory of the brightness spot changes in a continuous time scale is analyzed to determine the time point and spatial location of the late-arriving brightness spot, and the directionality of the reflection path is determined by the spatial difference of the dual-band brightness spots. Given the time point and spatial location of the late bright spot, inter-frame difference calculation is performed on the visible light image sequence and the near-infrared image sequence to identify the incident angle and return trajectory of the reflection path and establish the spatial distribution relationship. Reflection path data is generated based on the incident angle and return trajectory of the reflection path, and the data is summarized in time and space to form a delay list.

5. The chip mounting defect detection method integrating dual-band vision and light field modulation according to claim 4, characterized in that, During the bright spot intensity detection process, the change in light intensity in the bright spot area is continuously monitored according to the time scale of the light and shadow rhythm draft. By comparing the brightness change trend under adjacent time scales, the reflected light that enters the imaging area late is identified, and the time starting point and spatial coordinates of the reflection path are determined based on the sudden change in bright spot intensity.

6. The chip mounting defect detection method integrating dual-band vision and light field modulation according to claim 4, characterized in that, The steps for generating the artifact feature table are as follows: After obtaining the delay list, the reflection path data and the motion beat data of the chip handling process are aligned with the time reference, and each reflection path event and the corresponding time period motion beat information are arranged on a unified time scale to form a time sequence correspondence. Based on the time alignment results, analyze the time correspondence between the change in reflection angle in the reflection path data and the displacement phase in the motion beat, and determine the synchronization relationship between the change in reflection path direction and the chip motion phase; Based on the temporal correlation between the change in reflection angle and the beat phase, the coupled data is extracted to determine the time interval in which the reflected signal is concentrated and to correspond it with the corresponding motion phase. A false image feature table is generated based on the time interval in which the reflected signals occur, and the bright spot energy distribution, spatial location distribution and duration frame number information are recorded.

7. The chip mounting defect detection method integrating dual-band vision and light field modulation according to claim 6, characterized in that, The determination of the time interval of concentrated reflection signals is achieved by tracking the continuity of reflection angle changes and the energy accumulation trend of bright spots. The energy enhancement period, reflection angle shift period and bright spot dissipation period of reflected light are classified so that each group of concentrated reflection events in the false image feature table can simultaneously record the start and end time range, light intensity change curve and spatial displacement trend.

8. The chip mounting defect detection method integrating dual-band vision and light field modulation according to claim 6, characterized in that, The time-constrained index calculation steps are as follows: Based on the bright spot energy distribution information and spatial location distribution information in the false image feature table, the change of bright spot intensity over time is tracked frame by frame in the real-time detection image to construct a time chain of bright spot intensity change; Based on the time tracking results of the bright spot intensity, combined with the spatial distribution information of the previous frame image, the region overlapping with the previous frame image in the real-time detection image is identified and the trend of light intensity change is determined. Based on the identified overlapping areas, the existence time of bright spot areas in consecutive frames is statistically analyzed, repeated reflection areas are extracted, and the duration on the time axis is determined. Based on the temporal distribution of the repeated reflection areas, a time occupancy index is calculated to reflect the proportion of normal exposure time occupied by the reflected signal, and the time occupancy index is used as the control basis for light field rhythm adjustment.

9. The chip mounting defect detection method integrating dual-band vision and light field modulation according to claim 8, characterized in that, The calculation of the time occupancy index is based on the ratio of the duration of the reflected signal on the time axis to the normal exposure cycle. By accumulating the duration of the repeated reflection area, the occupancy ratio of the reflected signal within the exposure cycle is obtained, and the time occupancy index is used as a parameter for adjusting the light field rhythm.

10. The chip mounting defect detection method integrating dual-band vision and light field modulation according to claim 8, characterized in that, The steps are as follows: Non-synchronous shutter control is executed based on the time-occupancy index; the exposure interval is adaptively adjusted according to the intensity of reflection interference; and reverse half-shot brightness adjustment is performed. A sliding dark window is inserted during the interference period to block late-arriving reflection signals. After obtaining the time occupancy index, non-synchronous time control is performed on the exposure rhythm. The time interval between adjacent exposure cycles is adjusted according to the changing trend of the time occupancy index, so that the exposure interval is adaptively adjusted according to the intensity of reflection interference. Based on the adjusted exposure cycle, reverse half-photograph brightness rhythm adjustment is performed, and the phase of the illumination light and the time distribution of the reflected interference signal are reversed, so that the illumination light illuminates the detection area with the opposite phase during the reflected interference period. Based on the results of reverse half-photograph brightness rhythm adjustment, a sliding dark window is dynamically inserted in the exposure window to block reflected interference signals from entering the imaging area during the interference period; Based on the results of non-synchronous shutter rhythm control, reverse half-shot bright rhythm adjustment, and sliding dark window insertion, the light intensity distribution and exposure time allocation within the exposure cycle are coordinated and adjusted to form a dynamic balance of light field rhythm.

Images