Municipal road construction quality supervision system based on artificial intelligence

Abstract

The application discloses a municipal road construction quality supervision system based on artificial intelligence and relates to the technical field of intelligent supervision of municipal road engineering, comprising a loop account building module, a seal identification module, a facing review module and a release control module. The loop account building module divides the inspection well ring belt into continuous sectors, extracts the first paste edge, the first ramming and the return events and gives them sequence numbers, and generates a loop event account book; the seal identification module compares the corresponding sequence numbers, determines the earliest formed seal sector in combination with the optimization judgment; the facing review module constructs a facing weakened review belt according to the seal sector according to the circumferential facing logic, and updates the time sequence lagging sector to a later-formed limited state; the release control module determines the conventional repair condition based on the limited state or the placeholder, and outputs the final supervision instruction in combination with the communication timeout degradation mechanism. The application can accurately identify the internal ring compaction uneven defects caused by local first sealing, and realize the quantitative closed-loop control of hidden diseases.

Description

Technical Field

[0001] This invention relates to the field of intelligent supervision technology for municipal road engineering, and specifically to an artificial intelligence-based municipal road construction quality supervision system. Background Technology

[0002] In modern urban road construction and upgrading projects, anti-settlement manholes are widely used. An anti-settlement manhole and its surrounding area constitute a composite rigid-flexible transition structure consisting of a manhole cover, manhole base, adjusting ring, local asphalt fill layer, and the original road surface. Due to the narrow, closed-loop space around the manhole and strict slope elevation requirements, large compaction equipment cannot directly intervene in the initial forming stage of localized areas. The construction of the manhole-surround fill layer relies heavily on manual localized filling, layered compaction with handheld small tools, and interface bonding treatment. Because of the narrow working space and discontinuous construction height, the junction between the manhole cover and the surrounding road surface is extremely prone to early defects such as localized subsidence, cracking, and loosening.

[0003] Traditional acceptance mechanisms or conventional visual inspection systems for manhole construction quality often suffer from a technical blind spot, prioritizing superficial results while neglecting the internal forming sequence. Existing technologies primarily rely on static measurements of surface elevation differences and flatness after construction, or image recognition technology to locate external damage, failing to detect hidden structural defects beneath a smooth surface. When performing segmented compaction within a closed loop, there is a clear sequence of procedures. A specific sector within the loop typically completes aggregate embedding first, forming a strong, rigid constraint boundary. This first-formed sector then compresses the lateral flow space of the remaining filling material, causing the sector opposite its circumference to become spatially confined during subsequent material replenishment and compaction. This naturally leads to a hidden weakening zone with insufficient compaction and interlayer bonding faults. This internally biased structure, with one side tightly packed and the other loose, is easily masked by surface flatness after initial overall leveling and compaction. This results in manholes with internal structural defects being incorrectly deemed acceptable and proceeding to subsequent paving processes, ultimately leading to rapid damage under vehicle dynamic load impacts. How to overcome the limitations of existing static appearance detection technology, accurately identify internal circumferential force bias and uneven compaction defects caused by localized initial sealing by analyzing the sequential action logic in the formation process of the well perimeter ring, and transform hidden defects that are difficult to observe directly into quantifiable and clear business control instructions, is a technical problem that needs to be solved in this field. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes an artificial intelligence-based municipal road construction quality supervision system. This system solves the problems of existing static appearance detection which focuses on appearance but neglects the internal forming sequence, cannot penetrate the smooth appearance to accurately identify internal circumferential stress imbalance and uneven compaction defects caused by the premature sealing of the perimeter of the anti-settlement inspection well, and is difficult to transform these hidden defects into quantifiable and clear business control instructions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an artificial intelligence-based municipal road construction quality supervision system, comprising: The ring-forming ledger module is used to obtain the original construction event dataset, divide the perimeter of the anti-settlement inspection well into continuous sectors, extract the first edge contact, first compaction and return events from the original construction event dataset, assign natural construction sequence numbers according to the order of events, mark the pre-set placeholders for events that have not occurred, and generate a ring-forming event ledger containing the first edge contact sequence number, the first compaction sequence number and the return sequence number. The sealing recognition module is used to compare the first edge number with the first tamping number, take the maximum value to construct the sealing position number, traverse the continuous sector, combine the return number and the sealing position number to perform priority optimization and determine the earliest formed sealing sector. The opposing verification module is used to construct the opposing weakening verification band according to the circumferential opposing logic based on the sealing sector, and generate a verification task data structure containing the business identifier field. When the first edge sequence number or the first ramming sequence number of the continuous sector within the opposing weakening verification band is greater than the sealing position number corresponding to the sealing sector, the business identifier field is updated to the post-forming restricted state. The release control module is used to identify preset placeholders or post-forming restricted states based on the review task data structure to determine the conditions for triggering regular rework, output the final regulatory instruction, and execute the closed-loop action of the regulatory ledger.

[0006] Preferably, the ring-based accounting module obtains the original construction event dataset, including: A local plane coordinate system is established based on the reference point for the geometric center of the anti-settlement manhole cover. Extract the local perspective coordinate system data output by the vision device and the geodetic coordinate system data output by the high-precision positioning device; By using an affine transformation matrix, local perspective coordinate system data and geodetic coordinate system data are uniformly mapped to a local planar coordinate system to construct the original construction event dataset.

[0007] Preferably, the ring-based accounting module extracts the first edge contact, first compaction, and return events from the original construction event dataset, including performing the following processing by combining spatial, frequency domain, and trajectory features: The continuous contact length in the original construction event dataset is extracted based on the perspective transformation algorithm and the calculus algorithm. When the continuous contact length reaches the preset contact ratio threshold of the total length of the inner boundary of the continuous sector, the first contact event is extracted. Based on the Fast Fourier Transform, the peak energy of the excitation frequency band in the elevation direction in the original construction event dataset is extracted. When the peak energy of the excitation frequency band exceeds the effective excitation work threshold, the first ramming event is extracted. Extract continuous spatial movement trajectories from the original construction event dataset. When a spatial movement trajectory continuously crosses at least two new continuous sectors and then undergoes a turning-back action, extract the return event.

[0008] Preferably, when the ring-building module extracts the first post edge, first entry, and return events, it also includes: The first edge contact event, the first insertion event, and the return event are used as target events. The target device outer frame corresponding to the target event is mapped to a two-dimensional plane to generate a polygonal contact surface region. If the polygonal contact surface region falls into the boundary line of adjacent consecutive sectors, the boundary disambiguation rule is executed; Boundary disambiguation rules include calculating the Boolean intersection area between the polygonal contact surface region and adjacent continuous sectors, and assigning the target event to the continuous sector with the larger Boolean intersection area; The instantaneous forward direction vector is extracted based on high-precision positioning coordinates or inertial measurement unit data under continuous timestamps. If the Boolean intersection area is consistent and the instantaneous forward direction vector cannot be effectively extracted, the target event is assigned according to the cyclic increasing traversal order of continuous sector segments.

[0009] Preferably, when the sealing recognition module traverses consecutive sectors, it executes component identifier normalization rules based on cyclic congruence logic, including: Combine the total number of sectors to execute cyclic congruence logic, and locate the two adjacent sectors on both sides of the consecutive sector; Continuous sectors that meet the spatiotemporal constraints are identified as candidate sectors for closure. The spatiotemporal constraints are as follows: (1) The first edge number and the first ramming number of the adjacent sector segments on both sides have been recorded, and the first edge number and the first ramming number of the adjacent sector segments on both sides have not been marked as preset placeholders. (2) The sequence number of consecutive sector segments has already been recorded; (3) The first edge number and the first ramming number are already recorded in the continuous sector segment itself; Among the candidate sealing segments, priority optimization is performed by combining the return sequence number and the sealing position number to determine the earliest formed sealing segment.

[0010] Preferably, the sealing recognition module performs priority optimization determination, including: The candidate sector that returns to the earliest sequence number is determined as the sealing sector; If there are multiple candidate sectors for sealing with the earliest return sequence number, the candidate sector with the smallest sealing position number among the candidate sectors for sealing with the earliest return sequence number shall be determined as the sealing sector. If there are multiple candidate sectors with the smallest sealing position number, perform one-way polling on the candidate sectors with the smallest sealing position number in the cyclic ascending order of consecutive sectors, and determine the first candidate sector hit by the one-way polling as the sealing sector.

[0011] Preferably, the opposing verification module constructs an opposing weakening verification band according to the circumferential opposing logic, including: The total number of consecutive sector segments is limited to an even number. The total number of sector segments and the specific index of the closing sector segment are obtained. After offsetting the specific number index of the sealed sector by half of the total number of sectors, the corresponding sector number is calculated by performing a cyclic congruence logic based on the total number of sectors, and the corresponding center sector number is located. The opposing central sector segment, together with the adjacent left and right consecutive sector segments along the circumferential direction, is combined to construct the opposing weakened core band.

[0012] Preferably, after the counterpart review module generates the review task data structure containing the business identifier field, it also includes: Determine whether the sequence number corresponding to the consecutive sector within the counter-weakening verification band is marked as a preset placeholder; If the sequence number corresponding to consecutive sectors within the counter-weakened verification band is marked as a preset placeholder, the safety fallback mechanism is triggered. The safety fallback mechanism includes updating the business identifier field to force the continuous sector with the corresponding sequence number marked as a preset placeholder to be locked as a mandatory review target in the review task data structure.

[0013] Preferably, after the release control module identifies the restricted status based on the data structure of the review task, it also includes: Receive supplementary data for consecutive sectors that are in a post-forming restricted state; Perform alignment verification to determine whether the characteristics of the supplementary data match the missing process type corresponding to the continuous sector in the post-forming restricted state. The missing process type includes the first edge-attaching event missing type or the first ramming event missing type. When matching supplementary data is received, and the supplementary data once again meets the judgment criteria of the first edge event or the first insertion event, the business identifier field corresponding to the continuous sector in the review task data structure that is in the post-forming restricted state is dynamically reset, and the post-forming restricted state is lifted.

[0014] Preferably, the release control module outputs the final regulatory instruction, which also includes performing the following processing in conjunction with the communication timeout degradation mechanism: If the time elapsed since the first edge contact event exceeds the time threshold determined based on the asphalt mixture temperature drop window, and the review task data structure fails to be fully transmitted, the communication timeout degradation mechanism will be activated, the final regulatory instruction will be forcibly set to a repair pending state, and a retransmission request will be issued. During the communication timeout degradation mechanism, if the coordinates of the positioning terminal corresponding to the external paving equipment are found to fall within the electronic fence bound to the perimeter of the anti-settlement inspection well, a spatial physical interruption signal is triggered, interrupting the retransmission request and forcibly setting the final regulatory instruction to the illegal coverage alarm state.

[0015] Compared with existing technologies, it has the following advantages: This proposed AI-based municipal road construction quality supervision system constructs a sequential, ring-shaped event ledger covering the entire perimeter of anti-settlement inspection wells. By capturing the temporal sequence of the first edge-attaching and first compaction actions in a local area, it accurately locates the earliest formed sealing segment. Based on this, the system introduces circumferential opposition and cyclic congruence algorithms to transform the lateral spatial compression and circumferential force offset caused by the initial sealing on one side into precise coordinate addressing, directly locating the hidden weakened core zone relative to its position. This mechanism breaks through the limitations of traditional visual inspection, which relies solely on the surface flatness and height difference after construction. It successfully transforms the difficult-to-observe internal looseness and uneven compaction into calculable geometric topological relationships, achieving effective prediction and precise targeting of deep defects in rigid-flexible transition structures.

[0016] For the identified weakened verification zones, this solution establishes a dynamic state machine flow and release control system with mandatory interception capabilities. The system marks pre-set placeholders for processes that have not yet occurred and automatically assigns a restricted forming state to the corresponding sector based on time lag, thereby intercepting routine business release instructions. Combined with a communication timeout degradation mechanism based on the asphalt mixture temperature drop window and a spatial intrusion interruption signal linked to external paving equipment, the system can trigger the highest level alarm under extreme communication disruptions or forced compaction conditions. This process, from the underlying algorithmic level, prevents the potential for defective areas to flow into the next paving process, transforming abstract internal hazard warnings into quantifiable and clear final regulatory instructions, effectively ensuring the overall forming quality of local hidden works in municipal roads. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system framework of the present invention.

[0018] Figure 2 This is a schematic diagram of the system flow of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figures 1 to 2 This application provides an artificial intelligence-based municipal road construction quality supervision system, including a ring-based accounting module, a sealing recognition module, a counter-verification module, and a release control module; The ring-based ledger module is used to structure the discrete and complex on-site construction actions within the perimeter ring of the anti-settlement manhole into a ring-based event ledger that can be directly calculated and addressed later. The module first establishes a local planar coordinate system with the geometric center of the anti-settlement manhole cover as the origin and defines the perimeter ring. The inner boundary of the ring is the outer edge of the manhole seat, and the outer boundary is the contact line between the infill layer and the old pavement. The module then obtains the original construction event dataset D for this manhole location. Dataset D originates from the discrete physical coordinate set collected by the underlying sensors and the accompanying local timestamps. Subsequently, the perimeter ring is uniformly divided into N continuous sectors along the circumference, denoted as the i-th sector. Here, i is the sector traversal index, which ranges from 1 to N.

[0021] It should be noted that, for the spatial registration of multi-source heterogeneous sensors, the ring-building module arranges the position reference point based on the geometric center of the anti-settlement inspection manhole cover during the initialization stage. It performs joint spatial calibration of the local perspective coordinate system of the vision device and the geodetic coordinate system of the high-precision positioning device (such as RTK or UWB terminal) carried by the compaction machine. Through the affine transformation matrix, the multi-modal data is uniformly mapped to the same local plane coordinate system, thereby eliminating the spatial alignment blind spot of multi-source data at the bottom layer.

[0022] In this embodiment, parameter N is preferably 12. The local planar coordinate system provides a unified projection reference plane for chaotic field actions, including the well perimeter ring and continuous sector. This design incorporates physical spatial boundaries to address inherent structural contradictions within the rigid-flexible transition zone. The parameter N is set to an even number, 12, to align with the traditional on-site mapping conventions for dial orientation. Furthermore, within the standard reinforcement zone of a 1.5m x 1.5m anti-settlement inspection well, the outer arc length corresponding to each of the 12 divisions is approximately 30 to 40 centimeters. This dimension precisely encompasses the effective compaction width of a single pass using mainstream small handheld vibratory compactors, thus providing the discrete computational unit with strict representativeness of the physical construction entity and effectively reducing the spatial addressing complexity of the system.

[0023] For dataset D, the ring-building module sets triple feature determination conditions of spatial, frequency domain, and trajectory, for each sector. Only three types of core events with a necessary causal relationship to circumferential forming are extracted. The first type is the first edge contact event, which is recorded when the ring forming and accounting module determines that the filling material first approaches the outer edge of the well base and forms boundary contact. The set judgment threshold is that the continuous contact length formed between the filling material and the outer edge of the well base is not less than the target sector. The preset contact ratio threshold for the total length of the inner boundary is preferably one-third in this embodiment. The second type is the first compaction event, where the ring-forming and accounting module determines that the local compaction tool has pressed the material into the target sector for the first time. The corresponding narrow loop is recorded, and combined with the high-frequency oscillation characteristics of the equipment trajectory data in the Z-axis elevation direction, the energy peak value in the excitation frequency band (preferably in the 30 Hz to 60 Hz range) that matches the rated working frequency of the handheld vibratory compactor is extracted by fast Fourier transform. If and only if the amplitude of this energy peak value exceeds the preset effective excitation work threshold, the low-frequency background noise generated by the mechanical movement is filtered out, thus confirming that the action is in the excitation working state with actual compaction energy output. The third type is the return event, where the loop-forming accounting module records the continuous spatial movement trajectory of the working equipment or personnel, which advances along the circumferential direction and then returns to the sector where effective construction has already occurred. For actions of adjacent sectors, the set spatial span mandatory condition is that a continuous spatial movement trajectory must cross at least two new sectors before turning back to be considered valid.

[0024] It should be noted that the specific implementation of calculating the continuous contact length of the first edge contact event in this embodiment is to combine the camera intrinsic parameter matrix and the calibrated field extrinsic parameter matrix, use a perspective transformation algorithm to transform the pixel edge into an actual physical projection curve in a local planar coordinate system, and then use a calculus algorithm to calculate the curve length of the line segment that coincides with the inner rigid boundary of the well seat. If the pixel curve is briefly interrupted due to personnel or equipment obstruction in the field scene, the system uses a linear multi-frame temporal interpolation algorithm based on the kinematic trajectory of the equipment to continuously complete the interrupted edge, which serves as the scientific basis for determining the one-third threshold. The first edge contact event characterizes the material locking the rigid boundary of the well seat. The first ramming event, combined with frequency domain filtering and amplitude verification, scientifically eliminates the possibility of the equipment crossing the travel distance under no-load conditions. The return event forcibly shields the in-situ reciprocating adjustment jitter at the algorithm logic level. The above core events accurately extract the key state transition nodes that determine the quality of internal layering.

[0025] The extracted valid events are mapped to the corresponding sectors. If the physical location of an event falls on the boundary line between two adjacent sectors, the boundary disambiguation rule is applied to assign the event to the sector with the larger area of ​​influence.

[0026] It should be noted that, regarding the boundary disambiguation rule, the physical location specifically refers to the polygonal contact surface region mapped to the two-dimensional plane through the equipment's outer frame, eliminating the ambiguity of a single dimensionless geometric point. By calculating the Boolean intersection area of ​​the polygonal contact surface region and the surface regions of two adjacent sectors, the sector with the larger area of ​​action is scientifically determined. If the intersection areas are consistent and the instantaneous forward direction vector of the target equipment (configured by time differentiation of high-precision positioning coordinates under continuous timestamps, or calculated by fusing data from the inertial measurement unit at the end of the equipment) cannot be effectively extracted, the system uses the increasing direction of the sector traversal index i as a fallback principle to ensure that the underlying data mapping has a high degree of uniqueness under complex engineering boundaries.

[0027] For each sector Based on the chronological order of the actions, the three types of events are assigned and recorded as natural construction sequence numbers, ultimately generating Well Circumference Event Ledger B. Ledger B includes the first edge sequence number. and the first serial number and return to serial number If a certain sector If an event has not occurred by the end of the current process, the ledger field is forcibly marked as a preset placeholder, distinct from all positive integer natural sequence numbers, and the use of zero as a filler is prohibited. All three sequence numbers are naturally increasing sequential numbers, and the preset placeholder is used to identify missing processes that have not yet occurred, avoiding the underlying numerical sorting algorithm from mistakenly identifying zero values ​​as the earliest order of occurrence, thus preventing the business analysis results from being reversed.

[0028] It should be noted that the aforementioned natural construction sequence numbers have a strict local lifecycle relativity. The system uses the local timestamps uploaded by each sensor, based on the construction site gateway's reference clock, and performs unified offset calibration within the local area network using a network time protocol or precise time protocol. Then, it triggers an incrementing integer counter specific to the current single anti-settlement manhole construction task according to the real time sequence. The parameter design of the local topology counting greatly reduces the out-of-order interference caused by on-site network latency and clock asynchrony of various IoT devices. Simultaneously, when the edge gateway becomes unavailable for an extended period due to power outages or communication interruptions, leading to the risk of nonlinear timestamp drift, the ring-based ledger module will trigger a degradation protection mechanism, forcibly freezing and invalidating the ledger data of the current manhole construction unit and generating an on-site manual review warning instruction. This aligns the disordered spatial coordinate flow into a spatiotemporal topology sequence reflecting the construction organization process, providing highly robust data support for subsequent offset edge locking and tracing.

[0029] The sealing recognition module, based on the well-circumference ring-forming event ledger B output by the ring-forming ledger module, identifies where the well perimeter first forms a closure, thereby accurately locating potential local locking segments. First post side number First, enter the serial number. And return to the serial number For each sector The sealing identification module constructs the sealing position number. The computational logic used is as follows: .

[0030] It should be noted that the maximum value function is used to construct the sealing position number. This accurately characterizes the physical evolution of the rigid-flexible transition structure during construction. The local locking edge of the perimeter ring of the anti-settlement inspection well possesses complex technological attributes. A sector only possesses the basic physical conditions for forming a rigid boundary seal after both the first edge-applying event and the first compaction event have actually occurred. If only edge-applying occurs without effective compaction, the aggregate inside the asphalt material remains in a loose state. If only compaction occurs without sufficient edge-applying, no interface constraint on the well seat is formed. The later of the two time topological sequences is taken as the earliest reference position for the sector to complete the closure preparation. This calculation logic effectively avoids system misjudgment caused by premature triggering of a single action and accurately restores the spatiotemporal nodes of the mechanical interlocking of the rigid-flexible transition structure.

[0031] Subsequently, the sealing recognition module traverses all continuous sectors, filtering out candidate sectors that meet multiple spatiotemporal constraints. The filtering criteria require the sector... Three hard constraints must be met simultaneously. Firstly, the sector... Both adjacent sector segments on both sides have completed the initial edge application and initial compaction, meaning the corresponding sequence numbers are not marked as preset placeholders. Secondly, sector segments It already has a valid return sequence number. Thirdly, sector Its own first post edge number With the first rammed serial number All have been effectively recorded.

[0032] It should be noted that, because sectors are arranged in a closed circular pattern in physical space, adjacent sectors on both sides use modulo-N cyclic congruential logic to execute component identifier normalization rules for spatial addressing at the underlying data index. For sectors... The specific index numbers of its adjacent two sector segments are: and By using component identifier normalization rules, the system achieves seamless spatial mapping of physical closed loops at the algorithm level, effectively eliminating program exceptions caused by linear traversal addressing out-of-bounds errors. Furthermore, the above three indicators constitute multiple spatiotemporal constraints for sealing and positioning, significantly improving the system's anti-interference capability. It requires that adjacent sector segments on both sides possess basic actions, indicating that the current sector segment is indeed within a continuously closed force-bearing area of ​​the circumferential path, effectively eliminating isolated work surfaces formed by workers arbitrarily selecting filling points. It also requires that it possess a return sequence number. This means that the construction equipment has completed at least one circumferential crossing and then physically turned back and closed. The integration of multiple spatial and order constraints significantly improves the engineering robustness of candidate decisions under complex construction interference.

[0033] For all candidate sealing segments that meet the constraints, the sealing recognition module performs a priority optimization decision to determine the sealing segment that is formed earliest in the entire ring. Here, k is the specific index number of the sealed sector. The specific optimization rules are as follows: prioritize selecting the return sequence number. The earliest candidate sector. If there are multiple candidate sectors, return the sequence number. If they are the same, select the sealing position number. The smallest candidate sector is the one that appears first in the time topology. When the candidate sector... and When all are completely identical, the sealing recognition module performs a one-way polling according to the cyclically increasing order of sector traversal index i, and selects the first matched candidate sector as the uniquely determined sealing sector. .

[0034] It should be noted that the complex situation of multiple machines operating concurrently at the engineering site can easily lead to multiple sectors having identical extreme parameters. The deterministic one-way polling degradation rule aims to eliminate the randomness of the system output under extreme data overlap conditions, ensuring the determinism of algorithm execution and the stable reproduction of the output state. Simultaneously, if no sector in ledger B meets the candidate conditions, the sealing identification module will trigger the data discontinuity degradation strategy. The system will mark the current well location as incomplete and automatically suspend the process, triggering the front-end mobile device to generate a manual on-site review instruction. The combination of the underlying anti-ambiguity rules and degradation strategy ensures that the system's spatial addressing maintains high uniqueness and closed-loop business control even under severe boundary conditions.

[0035] The sealing recognition module ultimately outputs a uniquely identified sealing sector. Sealing section In physical essence, it represents the initial locking point that first assumes the closure constraint in the wellbore loop path. Precisely capturing the sealing sector. This solidifies the abstract construction event flow into specific spatial anchor points, establishing a reliable geometric alignment benchmark for subsequent derivation of the offset weakening risk area based on the circumferential offset mechanism.

[0036] Among them, the offset verification module is used to shift the focus of supervision from the initial sealing point to the later-forming restricted area, which is highly susceptible to hidden quality defects, based on the physical bias mechanism. The offset verification module outputs the uniquely determined sealing sector from the previous module. Using the baseline input, extract the specific index k of the sealed sector and the total number N of the evenly divided sectors around the well. The offset verification module calculates the offset sector number r according to the circumferential offset logic. Assuming N is even, the calculation logic used is as follows: In terms of computational logic, this means that after offsetting the specific number index of the sealed sector by half of the total number of sectors, it is mapped by combining cyclic congruence logic.

[0037] It should be noted that the aforementioned circumferential opposition logic accurately maps the structural mechanical offset forming law of the circumferential ring of the anti-settlement inspection well. The circumferential ring is a closed circular structure, and when the sealing sector... After the initial formation of strong rigid boundary constraints and aggregate interlocking, the space for lateral material flow is greatly compressed. The sector segments geometrically opposite to these segments on the circumference face physical dilemmas of restricted filling and restricted compaction during subsequent material replenishment and forming, naturally creating a physical correspondence between first-locked edges and subsequent constraints. By using computational logic to precisely address the corresponding sector segment number r, the abstract, high-incidence area of ​​hidden defects is visualized as a definite spatial coordinate index.

[0038] After calculating the opposing sector number r, the opposing verification module locates the corresponding opposing center sector. With the center sector as the opposite side. Centered on the left, including its adjacent sector along the circumference. Adjacent sector to the right Together, they construct the opposite weakened complex kernel band Z. The constructed set is represented as follows: .

[0039] It should be noted that when constructing the opposite weakened core band Z, the left adjacent sector Adjacent sector to the right The index number also uses the modulo N cyclic congruential logic execution component identifier normalization rule for wraparound processing.

[0040] Specifically, the precise index number of the left adjacent sector is: The precise index number of the right adjacent sector is By using component identifier normalization rules, the system ensures addressing continuity across start and end boundaries 1 and N at the algorithm level. The physical significance of constructing a verification zone using three consecutive sectors lies in the spatially continuous expansion characteristics of internal structural relaxation and interlayer bonding weakening caused by well-circumferential offset locking. This typically forms a continuous stress influence zone in the opposing central sector and its adjacent area. The three-sector configuration not only completely encloses the internal weakening influence zone but also avoids the waste of computing power and on-site manpower caused by blindly verifying the entire ring, significantly enhancing the accuracy of focusing regulatory resources.

[0041] For the constructed opposing weakened review band Z, the opposing review module generates a matching review task data structure A. Review task data structure A includes the sealing sector number k, the opposing center sector number r, and the independent number of each sector within the opposing weakened review band Z. It also initializes the review completion flag, supplementary recording edge-fitting flag, supplementary recording insertion flag, and a business identifier field indicating whether each sector is in a post-forming restricted state.

[0042] The opposing verification module performs data type validation and timing determination. If the event sequence number corresponding to any sector within the opposing weakened verification band Z is a valid natural integer value, the preset placeholder is excluded, and its first edge sequence number or first insertion sequence number is numerically greater than the closing sector. When the corresponding event sequence number indicates a timing lag, the corresponding verification module determines and updates the business identifier field of the sector to a post-forming restricted state based on this timing lag characteristic.

[0043] It should be noted that the post-forming restricted status is a composite business judgment indicator. For sectors in the post-forming restricted status, the system will automatically intercept regular release instructions, forcing the front-end business flow to complete on-site inspection and return supplementary or duplicate records before the process release requirements can be met. Simultaneously, if any sector within the counter-weakened review band Z is recorded as a preset placeholder distinct from the positive integer natural sequence number during the loop-building stage, the counter-review module executes a safety fallback mechanism, forcibly locking that sector as a mandatory review target in the review task data structure A, prohibiting on-site personnel from skipping it. This safety fallback mechanism, from a system mechanism perspective, blocks the loophole of concealing and releasing missing processes, ensuring the rigor of the quality supervision closed loop.

[0044] The final output of the counter-verification module is the counter-weakening verification zone Z, which is spatially precisely located, and the verification task data structure A. The counter-weakening verification zone Z transforms the internal circumferential compaction unevenness that is difficult to detect with the naked eye on the engineering site into a spatial task area that can be directly issued and quantitatively verified by the supervision system, providing a unique and compliant business adjudication object for the generation of subsequent process release and rework instructions.

[0045] The release control module is used to output a unique and clear final regulatory instruction for a specific anti-settlement inspection well construction unit based on the verification results data of the opposing weakened verification zone Z, thus completing the closed loop of construction quality supervision. The release control module uses the opposing weakened verification zone Z output by the previous module and the verification task data structure A collected and transmitted for this verification zone as its underlying input data. Verification task data structure A includes verification completion markers for each sector, supplementary edge-attaching markers, supplementary compaction markers, post-forming restricted status indicators, and rework operation closed-loop confirmation records.

[0046] The release control module assesses whether the business rework conditions are triggered based on the review task data structure A. To avoid relying on vague subjective experience, the system sets up mandatory business rules based on deterministic states. The release control module determines that a regular rework condition is triggered when any of the following state-type conditions are met: 1. A sector within the weakened review zone Z has not completed review; 2. A sector within the weakened review zone Z has not completed edge-fitting; 3. A sector within the weakened review zone Z has not completed compaction; 4. A sector within the weakened review zone Z is still recorded as a post-forming restricted state; 5. The rework operation within the weakened review zone Z has not completed closed-loop confirmation.

[0047] It should be noted that prioritizing explicit record-based or status-based business rules effectively avoids the disturbance of on-site environmental noise to the identification results, significantly enhancing the robustness and engineering feasibility of the judgment logic. The hidden risk of internal loop sequence bias is transformed into specific process supplementation and status verification actions, fundamentally preventing the risk of defective structures flowing into the next process. For sectors in a post-forming restricted state, when the system receives supplementary data from the field equipment, the system aligns and verifies whether the characteristics of the supplementary data match the missing process type (edge ​​bonding or ramming) that triggered the restricted state of the sector. Only after receiving valid supplementary data or rework verification that matches and re-meets the one-third boundary contact length or peak excitation energy judgment criteria can the release control module dynamically reset the business identifier field of the corresponding sector in the verification task data structure A. This action removes the post-forming restricted state, breaks the judgment deadlock, and drives the state machine to the next node.

[0048] Based on the above judgment results and system communication status, the release control module outputs the final regulatory instruction U. The final regulatory instruction U includes four system final states, and its specific mathematical calculation logic is as follows: In the formula, This indicates that the final regulatory order is to release the application, and the triggering condition is that the full-element review within the counter-weakened review band Z is passed and there are no regular rework triggering conditions. This indicates that the final regulatory instruction is a rework, and the triggering condition is any regular rework triggering condition triggered by the opposite weakened review with Z. This indicates that the final regulatory instruction is pending repair, triggered by the system triggering a communication timeout degradation mechanism. This indicates that the final regulatory instruction is an illegal coverage alarm, triggered by a system-triggered spatial physical interruption signal.

[0049] After the final regulatory instruction U is generated, the release control module executes the closed-loop action of the regulatory ledger. When the final regulatory instruction U is for repair, the system automatically intercepts the regular business flow and freezes the current manhole location's process status. The system simultaneously generates a repair work order containing the coordinates of the abnormal sector, prohibiting equipment and personnel from advancing to the subsequent pavement paving stage. When the final regulatory instruction U is for release, the system removes the process restriction flag for the current manhole location, allowing the node to flow.

[0050] It should be noted that in extreme situations such as on-site communication interruption or equipment power failure, if the preset time threshold is exceeded from the date the review task is issued... If the data structure A for the review task still fails to be completely transmitted back to the cloud, the release control module will initiate a timeout degradation and suspension mechanism. Time threshold Based on the time window for the temperature of the asphalt mixture around the well to drop to the effective compaction lower limit temperature, it is preferably set within 30 to 45 minutes after the first edge-attaching event. At this time, the system will forcibly set the final monitoring instruction U to the repair pending state and continuously send retransmission requests to the site gateway. If, during this period, the system identifies that the paving equipment has intruded into the electronic fence of the current manhole construction through the coordinate data of the on-board positioning terminal, thereby triggering a spatial physical interruption signal, the release control module will immediately interrupt the retransmission request because the concealed works have been physically damaged. At this time, the system will forcibly set the final monitoring instruction U to the illegal coverage alarm state, triggering the highest level of on-site work stoppage defense. The degradation and alarm mechanism transforms the fuzzy time concept into an objective Boolean condition that can be precisely executed by the computer, effectively preventing silent illegal release caused by communication failures, and effectively ensuring the safety and reliability of the construction quality of the underlying rigid-flexible transition structure.

[0051] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An artificial intelligence-based municipal road construction quality supervision system, characterized in that, include: The ring-forming ledger module is used to obtain the original construction event dataset, divide the perimeter of the anti-settlement inspection well into continuous sectors, extract the first edge contact, first compaction and return events from the original construction event dataset, assign natural construction sequence numbers according to the order of events, mark the pre-set placeholders for events that have not occurred, and generate a ring-forming event ledger containing the first edge contact sequence number, the first compaction sequence number and the return sequence number. Among them, the inner boundary of the manhole ring is the outer edge of the manhole seat, and the outer boundary is the contact line between the filling layer and the old road surface. The manhole ring is evenly divided into N continuous fan segments along the circumferential direction. The ring-based construction accounting module extracts the first edge contact, first compaction, and return events from the original construction event dataset, and performs the following processing by combining spatial, frequency domain, and trajectory features: The continuous contact length in the original construction event dataset is extracted based on the perspective transformation algorithm and the calculus algorithm. When the continuous contact length reaches the preset contact ratio threshold of the total length of the inner boundary of the continuous sector, the first contact event is extracted. Based on the Fast Fourier Transform, the peak energy of the excitation frequency band in the elevation direction in the original construction event dataset is extracted. When the peak energy of the excitation frequency band exceeds the effective excitation work threshold, the first ramming event is extracted. Extract continuous spatial movement trajectories from the original construction event dataset, and extract the return event when a spatial movement trajectory continuously crosses at least two new continuous sectors and then produces a turning action. The sealing recognition module is used to compare the first edge number with the first tamping number, take the maximum value to construct the sealing position number, traverse the continuous sector, combine the return number and the sealing position number to perform priority optimization and determine the earliest formed sealing sector. Specifically, for each continuous sector segment, the maximum value between the first edge-attaching sequence number and the first ramming sequence number of the continuous sector segment is determined as the sealing position number of the continuous sector segment. When the sealing recognition module traverses consecutive sectors, it executes component identifier normalization rules based on cyclic congruence logic, specifically including: Combine the total number of sectors to execute cyclic congruence logic, and locate the two adjacent sectors on both sides of the consecutive sector; Continuous sectors that meet the spatiotemporal constraints are identified as candidate sectors for closure. The spatiotemporal constraints are as follows: (1) The first edge number and the first ramming number of the adjacent sector segments on both sides have been recorded, and the first edge number and the first ramming number of the adjacent sector segments on both sides have not been marked as preset placeholders. (2) The sequence number of consecutive sector segments has already been recorded; (3) The first edge number and the first ramming number are already recorded in the continuous sector segment itself; Among the candidate sealing segments, priority optimization is performed by combining the return sequence number and the sealing position number to determine the earliest formed sealing segment; The sealing recognition module performs priority optimization, specifically including: The candidate sector that returns to the earliest sequence number is determined as the sealing sector; If there are multiple candidate sectors for sealing with the earliest return sequence number, the candidate sector with the smallest sealing position number among the candidate sectors for sealing with the earliest return sequence number shall be determined as the sealing sector. If there are multiple candidate sectors with the smallest sealing position number, perform one-way polling on the candidate sectors with the smallest sealing position number in the cyclic ascending order of consecutive sectors, and determine the first candidate sector hit by the one-way polling as the sealing sector. The opposing verification module is used to construct the opposing weakening verification band according to the circumferential opposing logic based on the sealing sector, and generate a verification task data structure containing the business identifier field. When the first edge sequence number or the first ramming sequence number of the continuous sector within the opposing weakening verification band is greater than the sealing position number corresponding to the sealing sector, the business identifier field is updated to the post-forming restricted state. The opposition verification module constructs an opposition weakening verification band according to the circumferential opposition logic, specifically including: The total number of consecutive sector segments is limited to an even number. The total number of sector segments and the specific index of the closing sector segment are obtained. After offsetting the specific number index of the sealed sector by half of the total number of sectors, the corresponding sector number is calculated by performing a cyclic congruence logic based on the total number of sectors, and the corresponding center sector number is located. The opposing center sector is combined with the left and right adjacent continuous sector segments along the circumferential direction to construct the opposing weakened core band. The review task data structure includes the sealing sector number, the opposing center sector number, the independent number of each continuous sector within the opposing weakened review zone, and initializes the review completion mark, supplementary recording edge mark, supplementary recording insertion mark, and business identifier field for each continuous sector as well as whether it is in a post-forming restricted state. The release control module is used to identify preset placeholders or post-forming restricted states based on the review task data structure to determine the conditions for triggering regular rework, output the final regulatory instruction, and execute the closed-loop action of the regulatory ledger.

2. The municipal road construction quality supervision system based on artificial intelligence according to claim 1, characterized in that, The ring-based accounting module obtains the original construction event dataset, including: A local plane coordinate system is established based on the reference point for the geometric center of the anti-settlement manhole cover. Extract the local perspective coordinate system data output by the vision device and the geodetic coordinate system data output by the high-precision positioning device; By using an affine transformation matrix, local perspective coordinate system data and geodetic coordinate system data are uniformly mapped to a local planar coordinate system to construct the original construction event dataset.

3. The municipal road construction quality supervision system based on artificial intelligence according to claim 1, characterized in that, When the ring-based account building module extracts the first post edge, first entry, and return events, it also includes: The first edge contact event, the first insertion event, and the return event are used as target events. The target device outer frame corresponding to the target event is mapped to a two-dimensional plane to generate a polygonal contact surface region. If the polygonal contact surface region falls into the boundary line of adjacent consecutive sectors, the boundary disambiguation rule is executed; Boundary disambiguation rules include calculating the Boolean intersection area between the polygonal contact surface region and adjacent continuous sectors, and assigning the target event to the continuous sector with the larger Boolean intersection area; The instantaneous forward direction vector is extracted based on high-precision positioning coordinates or inertial measurement unit data under continuous timestamps. If the Boolean intersection area is consistent and the instantaneous forward direction vector cannot be effectively extracted, the target event is assigned according to the cyclic increasing traversal order of continuous sector segments.

4. The municipal road construction quality supervision system based on artificial intelligence according to claim 1, characterized in that, After the counterpart review module generates the review task data structure containing the business identifier field, it also includes: Determine whether the sequence number corresponding to the consecutive sector within the counter-weakening verification band is marked as a preset placeholder; If the sequence number corresponding to consecutive sectors within the counter-weakened verification band is marked as a preset placeholder, the safety fallback mechanism is triggered. The safety fallback mechanism includes updating the business identifier field to force the continuous sector with the corresponding sequence number marked as a preset placeholder to be locked as a mandatory review target in the review task data structure.

5. The municipal road construction quality supervision system based on artificial intelligence according to claim 1, characterized in that, After identifying the restricted status based on the data structure of the review task, the release control module also includes: Receive supplementary data for consecutive sectors that are in a post-forming restricted state; Perform alignment verification to determine whether the characteristics of the supplementary data match the missing process type corresponding to the continuous sector in the post-forming restricted state. The missing process type includes the first edge-attaching event missing type or the first ramming event missing type. When matching supplementary data is received, and the supplementary data once again meets the judgment criteria of the first edge event or the first insertion event, the business identifier field corresponding to the continuous sector in the review task data structure that is in the post-forming restricted state is dynamically reset, and the post-forming restricted state is lifted.

6. The municipal road construction quality supervision system based on artificial intelligence according to claim 1, characterized in that, The release control module outputs the final regulatory instruction, which also includes the following processing in conjunction with the communication timeout degradation mechanism: If the time elapsed since the first edge contact event exceeds the time threshold determined based on the asphalt mixture temperature drop window, and the review task data structure fails to be fully transmitted, the communication timeout degradation mechanism will be activated, the final regulatory instruction will be forcibly set to a repair pending state, and a retransmission request will be issued. During the communication timeout degradation mechanism, if the coordinates of the positioning terminal corresponding to the external paving equipment are found to fall within the electronic fence bound to the perimeter of the anti-settlement inspection well, a spatial physical interruption signal is triggered, interrupting the retransmission request and forcibly setting the final regulatory instruction to the illegal coverage alarm state.

Images