Method for structure completion and comparison of incomplete fingerprint traces

Abstract

This invention relates to a method for structural completion and comparison of incomplete fingerprint traces. The method involves acquiring fingerprint trace data; dividing the structure into valid and missing areas according to preset criteria; determining structural anchor points within the valid area and limiting the completion from these anchor points to the missing area; detecting break points around the anchor points and screening for true break points; using the true break points and anchor points to complete the missing area with ridge structures to obtain completed structural segments, and labeling each completed structural segment with a confidence level and a revocable flag; performing an initial comparison based on the valid area and anchor points to obtain a candidate set; using the completed structural segments according to their confidence level during verification comparison of the candidate set; triggering a backtracking mechanism when a structural conflict criterion is met, revoking the conflicting completed structural segment based on the revocable flag, and either adjusting the dependent break points and then recompleting the conflicting locality; and outputting the comparison results and evidence composition information, wherein the evidence composition information includes at least the original structural basis and the completion verification basis.

Description

Technical Field

[0001] This invention relates to the field of biometric identification technology, specifically a method for structural completion and comparison of incomplete fingerprint traces. Background Technology

[0002] Existing matching technologies for incomplete fingerprints or small fingerprint areas typically increase the amount of matching information by expanding or restoring the original fingerprint image. For example, the technical solution disclosed in patent document CN116092132A involves acquiring a first fingerprint image to be matched and expanding it using a pre-trained feature extraction network and image generation network to generate an expanded second fingerprint image. The original and expanded images are then used together for matching to improve the recognition rate of small fingerprints. However, this type of technology relies heavily on image generation or feature restoration models to expand the fingerprint image as a whole. The expansion process depends heavily on the latent distribution patterns learned from the training data and uses a generation network to restore image information in missing areas. While this approach can increase the area of ​​the image available for matching to some extent, it is essentially a probabilistic restoration at the image level, lacking explicit constraints on the actual ridge structure. For example, the expansion process does not progressively verify the continuity of ridges, the spacing between ridges, or the true continuation of fracture structures. It also lacks a structural inference mechanism based on structural anchor points or actual fracture points, which may lead to directional deviations, abnormal spacing, or structural inconsistencies in the generated expanded ridges in local areas. If the generative network generates ridge patterns in missing regions that do not conform to the true fingerprint structure, it may introduce erroneous structural information into subsequent matching processes, interfering with the matching results. Furthermore, these methods typically participate directly in feature similarity calculations after image expansion, without hierarchical reliability control of the expanded structure, and lack mechanisms for dynamic verification or backtracking of the expanded structure during matching. Therefore, it is difficult to correct deviations in the expansion results in a timely manner, thus affecting the reliability and stability of the matching results to some extent.

[0003] Furthermore, in actual criminal investigation or latent fingerprint recognition scenarios, incomplete fingerprints often not only suffer from insufficient area but are also frequently accompanied by complex interfering structures such as drag marks, repeated ridges, local smears, or contamination. These abnormal structures significantly affect the direction and continuity of the ridges. However, existing technologies typically focus on completing or expanding the overall image using deep learning models, without specifically identifying and isolating these structural interferences, and lack spatial constraints on the completion process. For example, the expansion process in existing technologies mainly involves obtaining feature vectors through feature extraction networks, reconstructing image regions through generative networks, and then completing matching calculations through similarity fusion. This process does not distinguish between valid structural regions and structural interference regions, nor does it establish spatial boundary constraints for ridge completion. Therefore, when drag marks or smears are present in the image, the generative network may mistakenly identify these abnormal structures as normal ridges and expand them, resulting in inconsistencies between the completed ridges and the actual ridge structure. Meanwhile, existing technologies typically lack a cross-verification mechanism based on the spacing between adjacent lines when identifying fracture lines. The identification of fracture locations relies heavily on local features or model predictions, making it difficult to guarantee the accuracy of fracture point determination. Furthermore, during the matching stage, matching results are usually determined solely by a similarity threshold. When erroneous structures are generated in the expanded region, the system lacks a mechanism for backtracking and re-inferring erroneous structures, and cannot re-infer or correct locally completed structures. Therefore, in complex incomplete fingerprint scenarios, existing technologies still have certain shortcomings in terms of structural authenticity, controllability of the completion process, and traceability of matching results, making it difficult to achieve stable and reliable incomplete fingerprint completion and comparison while ensuring structural consistency. Summary of the Invention

[0004] The purpose of this invention is to provide a method for structural completion and comparison of incomplete fingerprint traces, thereby solving some of the drawbacks and deficiencies pointed out in the background art.

[0005] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows: a method for structural completion and comparison of incomplete fingerprint traces, comprising: acquiring fingerprint trace data; dividing the effective structural area and the missing structural area according to a preset criterion; determining structural anchor points in the effective structural area and limiting the completion from the anchor points to the missing area;

[0006] Breakpoints are detected around the anchor points and true breakpoints are selected; the missing area is filled with a textured structure using the true breakpoints and anchor points to obtain a complete structure segment, and each complete structure segment is marked with a confidence level and a revocable mark; an initial comparison is performed between the effective area of ​​the structure and the anchor points to obtain a candidate set; when performing a verification comparison on the candidate set, the complete structure segment is used according to the confidence level.

[0007] When the structural conflict criterion is met, a backtracking of the evidence is triggered. The conflict-filled structural segment is canceled according to the revocable mark, and the conflict-filled segment is re-filled after adjusting the dependent break point. The comparison results and evidence composition information are output, and the evidence composition information includes at least the original structural basis and the completion verification basis.

[0008] Furthermore, the division also includes identifying structural interference areas: when local lines show signs of repetition, dragging trails, or smearing / covering, and contradict the direction of line continuation, they are marked as structural interference areas; and the structural anchor points are limited to be selected only from the effective structural area, and the expansion path of structural completion does not cross the structural interference areas.

[0009] Furthermore, the screening of true fracture points includes trial-and-error mutual verification: the candidate fracture point is extended along its continuity direction within the structural missing area with a preset step length, and adjacent lines are simultaneously probed; when the two form a mutual verification relationship with the same direction and continuous spacing at the boundary of the missing area or at the remaining structure on the opposite side, it is determined to be a true fracture point; otherwise, it is determined to be a non-true fracture point or a fracture point to be verified.

[0010] Furthermore, the identification of the structural interference area adopts a two-way mutual verification: the consistency is determined based on the extension direction of the fracture endpoint and the main continuation direction of the local lines; only when the two contradict each other and there are repeated lines, drag trails or signs of smearing and covering at the same time, it is marked as the structural interference area.

[0011] Furthermore, after the structural interference area is marked, a buffer zone of a preset width is set on the outside; the structural anchor point and the completion starting position are limited to not falling into the buffer zone, and when the completion is advanced to the boundary of the buffer zone, it is only allowed to extend around the boundary without crossing into the structural interference area.

[0012] Furthermore, the buffer zone is configured with segmented width: along the boundary of the structural interference area, the boundary segment orthogonal to the direction of ridge continuation is set as the first width, and the boundary segment oblique to the direction of ridge continuation or with a trail pointing is set as the second width, which is greater than the first width.

[0013] Furthermore, when the completion progresses to the boundary of the buffer zone, boundary attachment extension is performed: the completion direction is locked to extend unidirectionally along the current boundary tangentially until a connectable texture line outside the structural missing area is detected, and then the completion is continued after leaving the boundary; if no connectable texture line is detected within a continuous preset length, the completion is terminated.

[0014] Furthermore, the preset step size is a hierarchical step size sequence, including a first step size and a second step size; the first step size is used to locate the mutual verification position on the opposite side, and the second step size is used to confirm the same direction of continuity and the relationship of spacing in the neighborhood of that position, and the second step size is smaller than the first step size; when the second step size probe fails to pass the preset mutual verification criterion, the candidate break point is determined as the break point to be verified.

[0015] Furthermore, the adjacent lines of the synchronous trial are the first adjacent line and the second adjacent line located on both sides of the candidate break point, and their distance relationship with the candidate break point is used as the mutual verification constraint; when only the first adjacent line or only the second adjacent line satisfies the distance maintenance criterion, the candidate break point is determined as the break point to be verified and the mutual verification missing item is recorded. The break point to be verified is only allowed to participate in the trial-type mutual verification again in the local re-completion after the proof of contradiction.

[0016] Furthermore, when the fracture point to be proven participates again in the trial-and-error mutual verification during the local re-completion after the reversal of the proof, a missing item compensation determination is performed: the compensation side is determined based on the recorded missing items in the mutual verification, and the second-level adjacent ridge line on the same side as the candidate fracture point is selected on the compensation side to participate in the synchronous trial; when the synchronous trial satisfies the spacing maintenance criterion and forms a continuous relationship in the same direction with the adjacent ridge line on the other side, the fracture point to be proven is adjusted to a true fracture point; otherwise, it remains a fracture point to be proven.

[0017] The beneficial effects of this invention are as follows: The method for structural completion and comparison of incomplete fingerprint traces proposed in this invention divides fingerprint trace data into effective and missing structural areas. Within the effective area, structural anchor points are determined as the structural basis for the continuation of ridge lines. Combined with breakpoint detection and true breakpoint screening, this achieves orderly completion of the ridge line structure in the missing area. The completion process is constrained by the ridge line direction and spacing at the anchor points, ensuring that the completed ridge lines maintain consistency with the original structure. This allows for the recovery of structurally continuous ridge line information even under incomplete fingerprint conditions. Furthermore, by setting a confidence level and a revocable flag for the completed structural segments, the completed structures can be used in subsequent comparisons at different levels, improving the reliability and flexibility of the completed structures in comparisons, thereby enhancing the accuracy and matching efficiency of incomplete fingerprint comparisons.

[0018] Furthermore, this invention effectively avoids misleading the completion process due to abnormal structures such as repeated lines, dragging trails, or smearing / covering by setting up a structural interference area identification mechanism and a buffer zone restriction strategy. It also improves the accuracy of true break point identification through trial-and-error mutual verification, hierarchical step-size verification, and adjacent line spacing constraints. In addition, a rebuttal backoff mechanism is introduced during the comparison process. When a structural conflict occurs between the completed structure and the candidate fingerprint, the conflicting completed structure segment can be revoked based on the revocable mark, and the local area can be re-completed, thereby reducing the impact of erroneous completion comparisons on the results. By simultaneously outputting the original structural basis and the completion verification basis, the comparison process is made traceable and the evidence is complete, which helps to improve the credibility and practical value of the incomplete fingerprint comparison results. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the core logic of the incomplete fingerprint structure completion and comparison method of the present invention.

[0020] Figure 2 This is a schematic diagram illustrating the identification of structural partitions, anchor points, and structural interference areas in Embodiment 1 of the present invention.

[0021] Figure 3 This is a geometric schematic diagram of the segmented variable-width buffer band in Embodiment 1 of the present invention.

[0022] Figure 4 This is a schematic diagram of the boundary attachment extension path selection and counter-evidence rollback in Embodiment 1 of the present invention.

[0023] Figure 5 This is a schematic diagram showing the structural partitions, anchor points, and candidate fracture points in Embodiment 2 of the present invention.

[0024] Figure 6 This is a determination diagram of mutual verification scoring, graded step size scoring and compensation scoring in Embodiment 2 of the present invention.

[0025] Figure 7 This is a schematic diagram of the backtracking, missing item compensation, and reconnection in Embodiment 2 of the present invention. Detailed Implementation

[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] Combined with appendix Figure 1 This invention discloses a method for structural completion and comparison of incomplete fingerprint traces. First, fingerprint trace data is acquired. This data can be obtained from on-site acquisition equipment or an image database, and includes structural information such as the shape, direction, and spacing of the fingerprint lines. Then, the fingerprint trace data is divided into regions based on preset criteria. These criteria are based on the continuity, clarity, and completeness of the local structure. A region is considered structurally valid when the fingerprint lines are continuous and the spacing remains stable. A region is considered structurally missing when the fingerprint lines are clearly interrupted, obscured, or missing, and the direction cannot be directly identified. This division creates valid regions with structural basis and missing regions requiring structural completion.

[0028] After the area is divided, structural anchor points are determined within the effective structural area. Structural anchor points are locations where the ridge structure is stable and has clear directional characteristics. These anchor points can be formed by ridge endpoints, bifurcation points, or stable nodes of continuous ridge segments. Selecting structural anchor points within the effective structural area provides a reliable structural basis for subsequent structural completion. The completion process is then limited to extending from the structural anchor points towards the structurally missing areas. During this extension, the ridge direction and spacing at the anchor points are used as continuation criteria to infer the ridge continuation trend within the missing areas, ensuring that the completed structure maintains consistency with the ridge direction and continuous ridge spacing within the effective area. This allows the ridge structure in the missing areas to be reasonably continued based on existing structural information, providing a stable structural foundation for subsequent structural completion and fingerprint comparison.

[0029] After determining the structural anchor points, fracture points of the ridge lines are detected in the surrounding area. Specifically, the continuity of the ridge lines at the boundary of the effective structural area is analyzed. When a ridge line abruptly terminates in the direction extending towards the structural missing area and there is no naturally closed structure at the termination point, this location is marked as a candidate fracture point. The candidate fracture points are then screened to identify true fracture points. The screening process is based on the direction of ridge line continuation, the spacing between ridge lines, and the structural consistency of adjacent ridge lines. When the direction of continuation of a candidate fracture point is consistent with the overall direction of the surrounding ridge lines, and adjacent ridge lines have the same continuation trend in the corresponding direction, the candidate fracture point is determined to be a true fracture point. Candidate fracture points that do not meet the above conditions are not included in the subsequent structural completion process.

[0030] After identifying the true fracture point, the true fracture point and structural anchor points are used as the structural basis to complete the texture structure in the missing area. The completion process extends the textures within the missing area segment by segment according to the direction of texture continuity and the relationship between texture spacing, ensuring that the completed textures maintain directional continuity and consistent spacing with the textures in the effective structural area, thus obtaining a completed structural segment. Each generated completed structural segment is marked with a confidence level and a revocable flag. The confidence level is determined based on the degree of consistency between the completed structure and the original texture structure, reflecting the reliability of the completed structure. The revocable flag is used to revert the corresponding completed structural segment in case of structural conflicts during subsequent comparisons.

[0031] After structural completion, an initial comparison is first performed based on the original ridge structure within the effective structural region and the determined structural anchor points. A candidate set matching the structural features is then selected from the fingerprint database. Subsequently, a confirmatory comparison is performed on the candidate set. During the confirmatory comparison, the completed structural structures are selectively used according to their confidence level. Completed structures with higher confidence levels are given priority in the comparison calculation, while those with lower confidence levels are only used as auxiliary references during the comparison process.

[0032] After completing the verification comparison, a comprehensive judgment is made on the structural consistency issues that arose during the comparison process. When a significant structural contradiction exists between the completed structural segment and the candidate fingerprint structure, it is determined that the structural conflict criterion is met. The structural conflict criterion is based on the consistency of the line direction, the relationship between line spacing, and the continuity of line connections. When the completed structural segment cannot form a reasonable continuity with the corresponding line in the candidate fingerprint at the comparison position, or when there is a significant directional deviation and spacing imbalance between the completed line and adjacent lines, the completed structural segment is considered to have a structural conflict. In this case, a rebuttal backoff mechanism is triggered to avoid misleading comparison results due to erroneous completions.

[0033] After triggering the reversal rollback, conflicting complete structural segments are located based on the reversible markers set for each complete structural segment during the completion phase. For complete structural segments identified as sources of conflict, a reversal process is first performed, removing the complete structural segment from the current structural model. Simultaneously, the breakpoints upon which the complete structural segment was generated are re-evaluated. If it is determined that a breakpoint might cause a deviation in the structural continuity direction, the breakpoint is adjusted or reselected. Subsequently, based on the updated breakpoints and existing structural anchor points, structural completion is re-performed on the conflicting local missing areas, ensuring that the new complete structure maintains a reasonable continuity with the surrounding lines.

[0034] After completing the reconstruction and passing structural consistency verification, the final comparison result is output. Simultaneously, corresponding evidence composition information is generated. This evidence composition information explains the basis for the comparison conclusion, including at least the original structural basis and the reconstruction verification basis. The original structural basis refers to the directly observable ridge structure information within the effective structural area and its corresponding anchor point positions. The reconstruction verification basis refers to the reconstruction structural segment formed within the missing area through breakpoint continuation and structural constraints, as well as the results of verifying the reconstruction structure during the comparison process.

[0035] Based on the division of fingerprint trace data into structurally valid and structurally deficient areas, structural interference areas are further identified. Structural interference areas are regions that interfere with the judgment of the true structure of the fingerprint lines. Specifically, structural feature analysis is performed on local fingerprint morphology. When repetitive fingerprint patterns appear in local areas (i.e., multiple lines are highly similar and densely arranged over a short distance), or when dragging trails appear (i.e., the lines are stretched in a certain direction to form a long, thin trailing structure), or when there are signs of smearing or covering (i.e., the boundaries of the lines are blurred and some lines are covered or overlapped), these areas are considered potential structural anomalies. Consistency judgment is then made based on the overall direction of the fingerprint lines' continuation. When the above-mentioned anomalies clearly contradict the normal continuation direction of the surrounding lines, the area is marked as a structural interference area.

[0036] After marking the structural interference area, spatial constraints are applied to the subsequent structural completion process. Structural anchor points are selected only from the valid structural area to ensure that the anchor point positions have a stable and reliable basis for the ridge structure, thereby avoiding the interference area from misleading the completion direction. Simultaneously, during ridge structure completion, the completion path is restricted, ensuring that the completion process can only extend within the valid structural area and the structural missing area. If the completion path approaches or points towards the structural interference area, the completion extension in that direction is stopped and a new continuation path is selected, thus preventing the completion process from crossing the structural interference area. Through these methods, the impact of repeated ridges, drag trails, or smear marks on the ridge continuation judgment can be reduced.

[0037] When identifying structural interference zones, a two-way verification mechanism is used to assess the local ridge structure. The fracture endpoints of ridges are detected at the boundaries of the effective structural zone or the edges of the structurally missing zone, and their extension direction is determined based on the direction of the ridges at the fracture endpoints. The extension direction indicates the direction in which the ridge may continue to extend within the missing area. Simultaneously, the overall structure of the ridges within a certain range around the fracture endpoints is analyzed to determine the main continuation direction of the ridges in that area. The main continuation direction is determined based on the overall direction and arrangement trend of multiple continuous ridges, reflecting the main directional characteristics of the local ridge structure.

[0038] Subsequently, the consistency between the extension direction of the fracture endpoint and the main continuation direction of the local ridges is determined. When the two directions are basically consistent, it indicates that the continuation trend of the fracture endpoint is consistent with the local ridge structure, and such areas are not considered structural interference areas. When there is a significant deviation or opposite trend between the extension direction of the fracture endpoint and the main continuation direction of the local ridges, further abnormal feature detection is performed on the area. If repeated ridge patterns, dragging trails, or signs of smearing are detected in the area at the same time, it is considered that there is significant structural interference in the area. Only when the extension direction contradicts the main continuation direction and the above-mentioned abnormal patterns appear simultaneously is the area marked as a structural interference area.

[0039] After marking the structural interference zone, a buffer zone of preset width is set outside the interference zone. This buffer zone forms a continuous restricted area around the boundary of the interference zone, isolating the interference lines from affecting the structural completion process. During subsequent structural completion, the positions of structural anchor points and the starting position of completion are constrained, ensuring that neither falls within the buffer zone, thus guaranteeing that the completion starting point is located in a structurally stable region. When the completion path advances towards the structural missing area and approaches the buffer zone boundary, directional control is applied to the path. At this point, completion is only allowed to extend along the buffer zone boundary, and is not permitted to cross the buffer zone into the structural interference zone, thereby preventing abnormal lines in the interference zone from affecting the line continuation judgment.

[0040] The buffer zone employs a segmented, variable-width design. Directional analysis is performed on different boundary segments along the boundary of the structural interference zone. When a boundary segment is orthogonal to the overall direction of the ridge line, it is set as a buffer zone of the first width to ensure a basic safety distance in that direction during the completion process. When a boundary segment is oblique to the direction of the ridge line or when a drag trail is present in that area, the boundary segment is set to a second width greater than the first width, thereby expanding the isolation range and further reducing the impact of abnormal ridge lines on structural completion.

[0041] When the completion process reaches the buffer zone boundary, a boundary attachment extension strategy is executed. Specifically, the current completion direction is adjusted to extend unidirectionally along the tangential direction of the buffer zone boundary, allowing the completion process to slide along the boundary outside the boundary. During the extension process, the system continuously checks whether there are any connectable texture structures outside the structurally missing area. When a texture structure consistent with the current completion direction and with a reasonable spacing is detected, the completion process detaches from the boundary and continues to complete the structurally missing area. If no connectable texture is detected within a continuous preset length range, it is determined that there is no reasonable continuous structure in that direction, thus terminating the completion process to avoid generating unreliable completed structural segments.

[0042] After detecting candidate breakpoints, they are screened to determine the true breakpoints. This implementation uses a trial-and-error mutual verification mechanism for determination. First, the direction of continuation of the lines at the candidate breakpoint is determined based on their local orientation. Then, a trial extension with a preset step length is performed along this direction within the structural gap. The trial extension simulates the possible continuation paths of the lines in the gap. Simultaneously, adjacent lines located on either side of the candidate breakpoint are tested synchronously. The synchronous tests are performed with the same direction and step length to observe the overall continuation trend of multiple lines in the gap.

[0043] During the probing process, the relationship between the extended path and the boundary of the structural missing area or the remaining structure on the opposite side is continuously monitored. When the probing path formed by the extension of the candidate fracture point and the synchronous probing path of the adjacent ridge line form a continuous relationship in the same direction at the boundary of the missing area or the remaining ridge line structure on the opposite side, and the spacing between each ridge line is consistent with the original spacing in the effective area of ​​the structure, the ridge line continuity relationship corresponding to the fracture point is considered to have structural mutual verification. At this time, the candidate fracture point is determined to be a true fracture point and is allowed to participate in the subsequent ridge line structure completion.

[0044] When the proposed extension fails to establish a consistent continuity with adjacent ridges, or when the extension path cannot form a reasonable connection with the remaining structure on the opposite side, the candidate fracture point is considered to lack sufficient structural basis. Candidate fracture points that clearly cannot form a continuity relationship are judged as non-true fracture points and do not participate in the subsequent completion process. Candidate fracture points whose continuity trend is not yet clear but have a certain structural possibility are marked as fracture points to be verified, so as to be further verified in subsequent structural analysis or completion processes.

[0045] When probing for mutual verification of candidate fracture points, a hierarchical step size sequence is used. This sequence includes a first step size and a second step size. The first step size extends along the direction of the candidate fracture point's ridge line to quickly locate potential points of mutual verification within the structurally missing area. The first step size determines whether the extension path can approach the boundary of the missing area or the remaining ridge line structure on the opposite side, thus establishing a preliminary mutual verification relationship. Once the first step size locates a potential mutual verification location, a second step size, smaller than the first step size, is used to further probe within the neighborhood of that location. The second step size is used to precisely confirm whether the extended ridge line and adjacent ridge lines form a continuous relationship in the same direction, and to determine whether the spacing between each ridge line remains consistent with the original spacing in the effective structural area. If the second step size probe fails to meet the preset mutual verification criteria, it indicates that the candidate fracture point lacks sufficient structural evidence; in this case, it is determined as a fracture point to be verified and not immediately included in the completion process.

[0046] During the synchronous probing process, the adjacent lines participating in the mutual verification are the first and second adjacent lines located on either side of the candidate break point, with their distance relative to the candidate break point serving as the mutual verification constraint. During the probing extension, the continuity trends of the first and second adjacent lines are simultaneously detected, and the changes in their distance from the candidate break point's extension path are compared. When both adjacent lines on both sides satisfy the distance preservation criterion, it indicates that the structural continuity relationship is stable. When only the first or only the second adjacent line satisfies the distance preservation criterion, it indicates that there is a missing term in the mutual verification structure. In this case, the candidate break point is determined as the break point to be verified, and the corresponding mutual verification missing term is recorded. The break point to be verified does not directly participate in the current completion process; it is only allowed to participate in the probing mutual verification again in the subsequent local re-completion after the proof-of-contrast regression.

[0047] When a breakpoint to be proven participates again in trial-and-error cross-verification during local re-completion after a reversal of proof, a missing item compensation determination is performed. Specifically, the side requiring compensation is determined based on the previously recorded missing cross-verification items. Then, on that side, a second-level adjacent ridge line on the same side as the candidate breakpoint is selected to participate in a new synchronous trial. By adding ridge structure information to the compensation side, the extension path is re-verified. When the new synchronous trial result satisfies the spacing maintenance criterion and forms a continuous relationship in the same direction with the adjacent ridge line on the other side, the breakpoint to be proven is adjusted to a true breakpoint. If a stable cross-verification relationship still cannot be formed, it remains a breakpoint to be proven, thereby avoiding unreliable structures from participating in ridge line completion.

[0048] Example 1:

[0049] In this embodiment, a fragmented handprint from a nighttime intrusion into a jewelry store is used as an example to illustrate the process of structural completion and comparison of fragmented handprint traces. The scene was the outside of a glass display case. Technicians used a multispectral handprint acquisition instrument for criminal investigation to acquire the handprint image. The acquisition resolution was 1000 dpi, and the original image size was 1536×1536 pixels. After preprocessing, 62 continuous identifiable line segments were obtained, along with 19 suspected missing line segments and 41 bifurcation points and endpoint features. The middle of the handprint showed a dragging trace formed after the security film was lifted, along with smearing and covering left by wiping, making it impossible to directly connect and complete the middle section. Simply crossing it with the shortest path would misidentify the dragging trace as a real line.

[0050] After performing orientation field and ridge continuity analysis on the image, the region was first divided into a structurally valid area and a structurally missing area based on ridge clarity, grayscale stability, and ridge-valley periodicity consistency. Then, structural interference areas were identified by combining local anomaly morphology. The structurally valid area retained most of the stable ridges on the upper left, lower left, and outer right edges. The structurally missing area was mainly located in the center of the palm print, with an area of ​​approximately 214×168 pixels. Technicians noted that near the left edge of the missing area, the extension directions of several broken endpoints were inconsistent with the main continuation direction of the remaining ridges on the right. Additionally, there were localized repetitive ridge shadows, trails extending along the tear direction of the security film, and masking bands caused by grayscale saturation. Therefore, this area could not be directly used as a completion channel. See the attached diagram corresponding to the above partitioning process. Figure 2 , Figure 2 The diagram shows the spatial distribution of the effective structural area, missing structural area, interference structural area, and seven structural anchor points on the palm print base map. The blue rectangles correspond to the missing structural area, the orange-red areas correspond to the interference structural area, and the yellow dots correspond to the structural anchor points selected from the effective structural area. This visually reflects the constraint that subsequent completion must not cross the interference structural area.

[0051] To avoid misidentifying drag trails as real structures, this embodiment employs a two-way verification method to identify structural interference regions. A structural interference region determination score is calculated for candidate anomaly regions:

[0052]

[0053] in, For repeating ridge response, To adjust the trail intensity, To apply coverage, Extend the pointing angle to the fracture endpoint. The angle of the main continuation direction of the local texture line. , , , These are weighting coefficients. They are taken within this anomaly region. , , , , and set , , , After substituting, there is

[0054]

[0055]

[0056]

[0057]

[0058] In this embodiment, the threshold for the structural interference region is set to 0.58. Since 0.6422 > 0.58, this abnormal region is marked as a structural interference region. The results indicate that the anomaly is not simply a missing value, but rather a result of directional inconsistencies and the combined effects of trails and occlusions. Subsequent completion paths must not traverse this region. Figure 2 The diagram also shows a schematic arrow pointing from the fracture endpoint to 128° and from the local main continuation direction to 86°. The difference in direction between the two further supports the determination that the area is a structural interference zone, rather than a normal crossable missing zone.

[0059] After structural partitioning is completed, structural anchor points are selected only from the effective structural area. In this embodiment, seven structural anchor points are selected from the upper left stable ridge group, the lower left complete rotary ridge group, and the right outer edge continuous ridge group. All anchor points are located in areas with stable grayscale and ridge-valley periodic errors of less than 8%. Both the anchor points and the completion start position are limited to outside the buffer zone to avoid the start position falling in a high-risk area affected by the wake. When expanding from the anchor point to the missing area, if the direction of advancement points to the structural interference area, the straight advancement is immediately stopped, and the direction is changed to detour along the boundary of the buffer zone. It is not allowed to cross the boundary and enter the interior of the structural interference area. Figure 2The seven anchor points shown are located in the upper left stable ridge group, the lower left rotating ridge group, and the right outer edge continuous ridge group, respectively, which intuitively reflects the selection rule that the anchor points only come from the effective area of ​​the structure.

[0060] To further control the risk of detours, a buffer zone is set up outside the structural interference zone, and the width is varied in segments at different boundary sections. The width of the buffer zone is calculated using the following formula:

[0061]

[0062] in, The width of the buffer band corresponding to the boundary of the j-th segment. Based on the width, The tangential direction angle of the boundary. The direction angle for the continuation of the ridge line. The risk factor is the direction of the trail. , This is the adjustment coefficient. In this embodiment, we take... Pixels , , For approximately orthogonal boundary segments, take... , ,but

[0063]

[0064]

[0065]

[0066]

[0067] Rounded to 10 pixels. For oblique segments with trails pointing towards the boundary, take... , ,but

[0068]

[0069]

[0070]

[0071]

[0072] The result is 12 pixels after rounding. This shows that the buffer zone at the obliquely intersecting boundary segment with a trail pointing outwards is wider, allowing for further extrapolation of high-risk detour areas. See the attached diagram corresponding to this process. Figure 3 , Figure 3The diagram shows the buffer zone settings for two typical boundary segments outside the boundary of the structural interference zone. The approximately orthogonal boundary segment uses a buffer zone of the first width, while the oblique boundary segment with the trail pointing in the direction of the trail uses a buffer zone of the second width. Figure 3 The geometric relationship expressed illustrates that The results are not simply numerical differences, but are directly used to expand the extrapolation range of high-risk areas, thereby constraining the completion path to detour in safer areas.

[0073] When the completion reaches the buffer zone boundary, boundary attachment extension is performed. The completion direction is locked to unidirectional extension along the current boundary tangentially. Completion continues only after a connectable ridge is detected outside the structurally missing area; if no connectable ridge is detected within a continuous 40-pixel range, the completion is terminated. To select the more reliable solution from the two candidate detour paths, the cost of the boundary attachment extension path is calculated:

[0074]

[0075] in, For the first The change in direction of each propulsion step To complete the distance from the path to the boundary of the structural interference zone, The tangential direction angle of the boundary. This is the actual propulsion direction angle. , , As weight, It is a small constant. In this embodiment, it is taken as... , , , See the attached diagram corresponding to the above path selection. Figure 4 , Figure 4 The diagram shows two candidate detour paths, the structural interference zone, the buffer zone boundary, the subsequent conflicting completion segment C5, and the reconnected trajectory after backtracking. This can be used to compare and understand the relationship between path cost and backtracking.

[0076] For candidate paths Substitute the four propulsion steps into the equation. The first step has... , , Then the cost of this step is

[0077]

[0078]

[0079]

[0080] The cost of the second step is

[0081]

[0082]

[0083]

[0084] The cost of the third step is

[0085]

[0086]

[0087]

[0088] The cost of the fourth step is

[0089]

[0090]

[0091]

[0092] Therefore The total cost is

[0093]

[0094] For candidate paths Perform the same calculation. The costs for the four steps are as follows:

[0095]

[0096]

[0097]

[0098]

[0099] therefore

[0100]

[0101] because Therefore, the chosen path It follows the boundary of the buffer zone. This path has less directional fluctuation and is generally farther from the boundary of the interference zone, making it suitable for boundary attachment extension. Figure 4 The solid blue line corresponds to the final selected path. The orange dashed lines represent paths that were not used. .Depend on Figure 4 It can be further seen that, It maintains a smoother tangential transition near the structural disturbance region, while It is closer to the boundary of the interference zone, thus having a higher distance penalty and greater overall risk.

[0102] After completing the local completion according to the above rules, a total of 11 completed structural segments were formed. Among them, 7 segments achieved continuous connection with the remaining ridge lines on the opposite side after leaving the buffer zone, with directional deviations all less than 9°, and were marked as high confidence level. The other 4 segments, due to their long attachment boundary distance, although forming local connectivity, relied on boundary detours for judgment during the journey, and were therefore assigned medium confidence level and marked as revocable. For 2 attempts where no connectable ridge lines were found on the outside within 40 consecutive pixels, the system directly terminated and did not include them in the completed structural segments, thus avoiding the introduction of edge noise errors into subsequent comparisons. Figure 4 The connection position at the end of the middle path also reflects the process of the completion breaking away from the buffer zone boundary and reconnecting to the outer continuous line, which can be used as a local illustration of the formation logic of the above 11 completion structure segments.

[0103] In the comparison phase, an initial comparison was conducted based on the effective structural region and 7 structural anchor points, yielding 27 candidate palm prints from a sample database of jewelry store employees and personnel who recently visited the display case. The initial comparison primarily utilized 62 effective ridge segments, 41 original bifurcation and endpoint features, and the geometric relationships between anchor points, without directly relying on revocable completion segments. Subsequently, a verification comparison was performed, in which all 7 high-confidence completion segments were included, while the 4 revocable completion segments were used only as supplementary verification evidence. After verification, the candidate set converged from 27 to 2. Candidate A and Candidate B scored similarly on the original structural basis, but Candidate A showed a significant conflict with a medium-confidence completion segment obtained by circling the boundary. Figure 4 The location of the conflict has been marked, with the original reliable completion segment C5 indicated by a red dashed line, which facilitates comparison with the subsequent completed trajectory after rollback.

[0104] The conflict manifests as follows: the completed segment C5 is inferred to be a smooth continuation in the right middle of the palm print, while the corresponding position of candidate A inherently exhibits a stable bifurcation, and the distance between the two sides of this bifurcation deviates from the similar distance within the effective area of ​​this palm print by 14.6%, exceeding the 10% tolerance set in this embodiment. Based on this, the system triggers a backtracking mechanism, canceling the completed segment C5 marked as revocable, and backtracking to the boundary attachment extension starting point it relies on. A new, more outward-facing exit is selected, causing the completed segment to delay by 8 pixels after leaving the buffer zone before connecting to the opposite side ridge. After re-completed, the original smooth connection is no longer formed at this point; instead, it remains a discontinuous termination structure, consistent with the palm print characteristics of candidate B, thus excluding candidate A and ultimately determining candidate B as the unique matching target.

[0105] Example 2:

[0106] In this embodiment, a fragmented handprint from a theft case at a logistics warehouse was found at the scene in the high-value electronic component storage area on the east side of the warehouse. The handprint remained on the outer surface of a metal sealing buckle. Investigators first used magnetic powder to reveal the fingerprint, then used a high-resolution metal surface handprint acquisition device to obtain an image at a resolution of 1200 dpi, with an image size of 1680×1320 pixels. The handprint had a narrow, elongated missing section in the center, formed by rust and peeling, with a measured width of approximately 4.8 mm, causing multiple lines to be interrupted in the middle. After preprocessing, the system extracted 58 valid structural line segments. The structural missing area was concentrated in the slightly right-central position, and 14 candidate break points were detected around the anchor points.

[0107] After performing orientation field estimation and ridge-valley period analysis on the image, clear, continuous, and gray-level stable areas are classified as effective structural regions, while banded areas where corrosion and peeling render the ridge lines unreadable are classified as structurally missing regions. Completion is only allowed to extend from structural anchor points selected within the effective structural regions to the structurally missing regions. In this embodiment, six structural anchor points are selected in the stable areas of the upper left, middle left, and upper right of the palm print, with the relative distance error between anchor points controlled within 7%. Centered on these six anchor points, 14 candidate fracture points are detected within an 18-pixel radius. Of these, eight are subsequently confirmed as true fracture points, four are identified as fracture points to be verified, and two are identified as non-true fracture points. The corresponding figures for the above structural partitioning, anchor point layout, and candidate fracture point distribution are shown in the attached figures. Figure 5 , Figure 5 It also shows the coarse positioning direction with a first step length of 0.42 mm and the fine confirmation direction with a second step length of 0.14 mm.

[0108] The determination of candidate fracture points employs a trial-and-error mutual verification method. For each candidate fracture point, it is first roughly located along its continuity direction by traversing the structural gap area with a first-step length. Then, it is refined and confirmed in the adjacent area on the opposite side with a second-step length smaller than the first-step length. Simultaneously, adjacent lines on both sides are probed. The first-step length is used to locate the mutual verification position on the opposite side; in this embodiment, it is 0.42 mm. The second-step length is used to confirm the continuity in the same direction and the maintenance of spacing; in this embodiment, it is 0.14 mm. If the candidate fracture point and the first and second adjacent lines simultaneously satisfy the conditions of continuity in the same direction and maintenance of spacing at the boundary of the gap area or at the remaining structure on the opposite side, it is determined to be a true fracture point. If only one side of the adjacent lines satisfies the condition of maintaining spacing, it is determined to be a fracture point to be verified, and the missing mutual verification item is recorded. If the directional difference is too large and the spacing on both sides is unstable, it is determined to be a non-true fracture point. Figure 5 The text highlights D7, D11, and D3, with D7 corresponding to subsequent true fracture point determination samples, D11 corresponding to fracture points to be verified and backtracking compensation samples, and D3 corresponding to non-true fracture point samples.

[0109] Calculate a trial-and-error cross-verification comprehensive score for candidate breakpoint D7:

[0110]

[0111] in, The angle difference between the direction of the candidate fracture point probe and the direction of the remaining ridge line on the opposite side. This represents the deviation between the actual ridge spacing and the reference spacing. The probing distance to cross the missing region, , , This is the scale parameter. For D7, take... , , , , , Substituting into

[0112]

[0113]

[0114]

[0115] In this embodiment, the mutual verification scoring threshold for true breakpoints is set to 0.22, because... Furthermore, since the left and right adjacent ridges of D7 maintain a reference spacing within the second step neighborhood, D7 is determined to be a true fracture point. In contrast, the candidate fracture point D3 has a directional difference of 11.8°, a spacing deviation of 0.17 mm, and a trial distance of 4.9 mm, calculated using the same formula...

[0116]

[0117] because The minimum acceptable value was not met, and no stable mutual verification was formed between adjacent lines on both sides; therefore, D3 was determined to be a non-true break point. The corresponding attached figure for the comparison of the above mutual verification scores and thresholds is shown below. Figure 6 .

[0118] For candidate fracture point D11, the first step of length probing has found the remaining striations on the opposite side. In the second step of fine confirmation, the adjacent striations on the left side meet the spacing requirement, while the adjacent striations on the right side cannot form a complete constraint due to the gaps at the corrosion edges. Therefore, further calculation of the graded step confirmation score is required:

[0119]

[0120] in, The coarse positioning mutual verification score obtained with the first step length is... To obtain a detailed mutual verification score using the second step length, The integrity of the constraint between adjacent ridges on both sides is set to 1 if both sides are satisfied, 0.5 if only one side is satisfied, and 0 if neither side is satisfied. For D11, the value is... , , and set , , ,but

[0121]

[0122]

[0123] This embodiment will As a criterion for true break points, As a criterion for determining the break point to be verified. Because... Therefore, D11 was identified as a break point to be verified, and the missing mutual verification item was recorded as a missing item on the right adjacent line. Similar to D11, there are also D6, D9, and D13 as break points to be verified. Thus, the 14 candidate break points resulted in a classification of 8 true break points, 4 break points to be verified, and 2 non-true break points.

[0124] Based on the above classification results, the system generates high-confidence complete structural segments with 8 true fracture points as the core and low-confidence complete structural segments with 4 fracture points to be verified, and uniformly adds a revocable mark. In the initial comparison stage, only the effective structural area, 6 structural anchor points, and high-confidence complete segments derived from true fracture points are used, resulting in 31 candidate samples from 186 handprint templates of warehouse employees, outsourced loading and unloading personnel, and recent equipment maintenance personnel. After entering the verification comparison, the system uses low-confidence complete segments in addition to the high-confidence complete segments for differentiation, reducing the candidate set from 31 to 3, denoted as Candidate A, Candidate B, and Candidate C.

[0125] During the verification comparison, a structural conflict arose between the low-confidence completion segment S11 formed by the break point D11 to be verified and candidate B. Specifically, S11 is connected into a continuous upward ridge line in the middle right, while candidate B has a stable short fork at the corresponding position, and the distance from this fork to the adjacent upper ridge line is 0.41mm, significantly greater than the similar reference distance of 0.27mm in the valid area of ​​this palm print, with a deviation of 51.9%. Based on this, the system triggered a backtracking of the rebuttal verification, first canceling the low-confidence completion segment S11 derived from D11, and then restoring D11 to the state to be verified. Since the original missing item record of D11 is a missing item of the adjacent line on the right, it is only allowed to participate in the exploratory mutual verification again in the local re-completion after the backtracking, and the compensation side is limited to the right side. The attached figure corresponding to the above conflict position, the original completion segment S11, and the missing item record is shown below. Figure 7 .

[0126] Difference, To supplement

[0127] After rollback, the system selects the second-level adjacent ridge line on the same side as D11 to participate in synchronous probing and calculates the missing item compensation judgment score:

[0128]

[0129] in, To compensate for the deviation between the first adjacent ridge line on the side and the reference spacing, To compensate for the deviation between the adjacent second-level ridge lines and the reference spacing, To compensate for the directional difference between the candidate fracture point and the ridge line on the other side, , , As the weight. Take D11 , , , , , , , Substituting into

[0130]

[0131]

[0132]

[0133]

[0134] In this embodiment, the compensation determination threshold is set to 0.63, because Furthermore, the second-level adjacent lines on the right side after compensation form a continuous relationship with the first adjacent lines on the left side in the same direction. Therefore, D11 is adjusted from the fracture point to be proven to the true fracture point.

[0135] After D11 is corrected, the system regenerates the locally completed segment S11. Unlike before the rollback, S11 Instead of connecting to the original lower residual ridge line, it is shifted upwards by 0.19mm to connect to another right-side residual ridge line, ensuring that the local connection relationship is consistent with the corresponding ridge line group in candidate C. Upon re-verification, candidate B, due to its similarity to S11... Candidate A was eliminated because it was inconsistent in both direction and spacing. Candidate A also had an endpoint misalignment in another highly reliable completion segment. In the end, only candidate C was retained as the only matching target.

Claims

1. A method for structure completion and comparison of incomplete handprint traces, characterized in that The method comprises: acquiring fingerprint trace data; dividing a structure effective area and a structure missing area according to preset criteria; determining a structure anchor point in the structure effective area and limiting expansion of the structure from the anchor point to the missing area; detecting a broken point around the anchor point and screening a true broken point; performing structure line completion on the missing area using the true broken point and the anchor point to obtain a completed structure segment, and labeling a confidence level and a reversible mark for each completed structure segment; performing initial comparison based on the structure effective area and the anchor point to obtain a candidate set; performing verification comparison on the candidate set using the completed structure segment according to the confidence level; triggering a counterargument rollback when a structure conflict criterion is met, and removing the conflict completion structure segment and adjusting the dependent broken point according to the reversible mark to re-complete the conflict part; outputting a comparison result and evidence composition information, wherein the evidence composition information at least includes original structure evidence and completion verification evidence.

2. The method of claim 1, wherein The division further comprises identifying a structure interference area: when local lines show signs of repetition, trailing tail, or smearing and covering, and are inconsistent with the direction of line continuation, the area is marked as a structure interference area; and the structure anchor point is selected only from the structure effective area, and the expansion path of the structure completion does not pass through the structure interference area.

3. The method of claim 1, wherein The screening of the true broken point comprises tentative mutual verification: the candidate broken point is tentatively extended in the structure missing area along its continuation direction at a preset step length, and the adjacent lines are synchronously tentatively extended; when the two form a mutual verification relationship of same direction, connectable and distance maintaining at the boundary of the missing area or the opposite side residual structure, the candidate broken point is determined as a true broken point, otherwise it is determined as a non-true broken point or a to-be-verified broken point.

4. The method of claim 2, wherein The identification of the structure interference area adopts bidirectional mutual verification: consistency determination is performed based on the extension direction of the broken end point and the main continuation direction of the local lines; only when the two are inconsistent and at the same time, signs of repeated lines, trailing tail, or smearing and covering appear, the area is marked as the structure interference area.

5. The method of claim 2, wherein After the structure interference area is marked, a preset width buffer zone is set outside; the structure anchor point and the completion starting position are limited not to fall into the buffer zone, and when the completion advances to the buffer zone boundary, it is only allowed to expand along the boundary and not to cross into the structure interference area.

6. The method of claim 5, wherein The buffer zone adopts a segmented variable-width setting: along the boundary of the structure interference area, the boundary segment perpendicular to the continuation direction of the lines is set as a first width, and the boundary segment oblique to the continuation direction of the lines or having a tail direction is set as a second width greater than the first width.

7. The method as claimed in claim 5, wherein the method further comprises: identifying the missing fingerprint trace; and completing the missing fingerprint trace by using the reference fingerprint trace. When the completion advances to the buffer zone boundary, boundary attachment extension is performed: the completion direction is locked to unidirectional extension along the tangent direction of the current boundary, and after the connectable lines outside the structure missing area are detected, the completion continues to detach from the boundary; if no connectable lines are detected within a continuous preset length, the completion is terminated.

8. The method of claim 3, wherein The preset step length is a hierarchical step length sequence, including a first step length and a second step length; the first step length is used for positioning the opposite side mutual verification position, and the second step length is used for confirming the same direction connectable and distance maintaining relationship in the neighborhood of the position, and the second step length is smaller than the first step length; when the second step length tentative extension does not pass the preset mutual verification criterion, the candidate broken point is determined as a to-be-verified broken point.

9. The method of claim 3, wherein The adjacent lines in the synchronous trial are the first adjacent line and the second adjacent line located on both sides of the candidate break point, and their distance relationship relative to the candidate break point is used as the mutual verification constraint. When only the first adjacent line or only the second adjacent line satisfies the distance maintenance criterion, the candidate break point is determined as the break point to be verified and the missing item in the mutual verification is recorded. The break point to be verified is only allowed to participate in the trial-type mutual verification again in the local re-completion after the proof of contradiction.

10. The method of claim 3, wherein When the fracture point to be proven participates in the trial-and-error mutual verification again in the local re-completion after the reversal of the evidence, a missing item compensation judgment is performed: the compensation side is determined based on the recorded missing items in the mutual verification, and the second-level adjacent texture line on the same side as the candidate fracture point is selected on the compensation side to participate in the synchronous trial; when the synchronous trial satisfies the spacing maintenance criterion and forms a continuous relationship in the same direction with the adjacent texture line on the other side, the fracture point to be proven is adjusted to a true fracture point; otherwise, it remains a fracture point to be proven.

Images