Description:FORM 2
THE PATENTS ACT, 1970
(39 of 1970)
&
The Patent Rules, 2003
COMPLETE SPECIFICATION
(See sections 10 & rule 13)
1. TITLE OF THE INVENTION
A REAL-TIME CAMERA-BASED SYSTEM AND METHOD FOR POTHOLE DETECTION, MAPPING AND QUANTIFICATION
2. APPLICANT (S)
NAME NATIONALITY ADDRESS
DIVYASAMPARK IHUB ROORKEE FOR DEVICES MATERIALS AND TECHNOLOGY FOUNDATION IN Indian Institute of Technology Roorkee, Roorkee-247667, Uttarakhand, India.
3. PREAMBLE TO THE DESCRIPTION
COMPLETE SPECIFICATION

The following specification particularly describes the invention and the manner in which it is to be performed.

FIELD OF INVENTION:
[001] The present invention relates to the field of of intelligent transportation systems, computer vision, and road infrastructure monitoring. The present invention in particular relates to a real-time camera-based system and method for pothole detection, mapping and quantification.
DESCRIPTION OF THE RELATED ART:
[002] The detection and quantification of road surface anomalies, particularly potholes, are critical for maintaining road safety, optimizing maintenance schedules, and reducing vehicle damage. Traditionally, the identification and measurement of such defects have relied on manual surveys or sensor-rich platforms involving LiDAR, stereo cameras, ultrasonic sensors, or inertial measurement units (IMUs). These methods are often cost-prohibitive, require specialized hardware, or lack scalability across large networks of roads. Recent advancements in computer vision have enabled camera-based detection of road damage. However, many existing systems either rely on multiple sensors for dimensional analysis or are limited to classification tasks without generating accurate physical measurements. Furthermore, some prior approaches estimate only the presence of damage without providing real-world area or volume information, which is essential for prioritizing repairs and estimating material requirements.
[003] Reference may be made to the following:
[004] Publication No. IN202541057903 relates to an innovative approach to road safety and maintenance. This system leverages the Sharp GP2YOA02YKOF infrared sensor and the Spark Fun AOXL345 accelerometer to detect road anomalies, specifically potholes. Sensor data is processed through an Arduino UNO ADC unit, which triggers two key actions upon detecting a pothole: the vehicle's speed is reduced, and GPS coordinates are recorded.
[005] Publication No. IN202531021554 relates to a portable system for real-time pothole detection and position tracking, with the goal of improving road safety and infrastructure maintenance. The system integrates various sensors, including a depth-sensing module and a camera module with a local processing unit put in the car.
[006] Publication No. IN202541020352 relates to the design and fabrication of the pothole using pothole-induced vehicle damage and accidents are on the rise, necessitating effective detection and repair strategies. This paper introduces an algorithm optimized for low-cost LiDAR sensors to detect and quantify potholes with high accuracy. The algorithm uses curvature-based analysis and voxelization to assess pothole size, with tests in Edmonton, Alberta, showing reliable detection across varying sizes and shapes.
[007] Publication No. IN202541015005 relates to road infrastructure plays a crucial role in ensuring safe and efficient transportation, yet potholes remain a persistent issue, causing vehicle damage, traffic disruptions, and accidents.
[008] Publication No. IN202541010838 relates to an EDGE-AI powered pothole detection system and method designed to enhance road safety through real-time detection and mapping of potholes. The system integrates deep learning models, including YOLOv8 and SSD-MobileNetv2, optimized for deployment on edge computing devices such as NVIDIA Jetson and Google Coral. Utilizing a dual-sensor setup comprising LiDAR and RGB cameras, the invention captures 3D depth information to ensure accurate detection even in low-visibility conditions. GPS functionality further enables real-time pothole localization, providing actionable insights for road maintenance teams.
[009] Publication No. IN202521006585 relates to the increasing frequency of road accidents caused by deteriorating road conditions, particularly potholes, has raised significant concerns regarding road safety. Potholes are a leading cause of accidents, vehicle damage, and inefficiencies in road travel. To address this growing issue, an - intelligent pothole detection and notification system integrated with the Internet of Things (loT) has been proposed to enhance road safety and reduce risks for drivers. The system utilizes a Camera Module fixed on the vehicle, capturing continuous live video streams from the vehicles perspective. This video feed is then processed by a custom-trained YOLOv8 model, which detects potholes in real time by analyzing the frames of the video sequence. YOLOv8, known for its speed and accuracy, enables quick and precise pothole detection, marking the potholes in the video feed with bounding boxes. Once potholes are detected, the system not only marks their locations on a digital platform but also provides an immediate voice alert to the driver, notifying them with a message such as, be cautious, there is a pothole.
[010] Publication No. IN202441088123 relates to maintaining road infrastructure is essential for ensuring public safety and vehicle longevity. Potholes, which are depressions in the road surface caused by wear and tear, pose significant hazards and can lead to accidents and vehicle damage. Traditional pothole detection methods rely on manual inspection, which is time-consuming, labor-intensive, and prone to errors. This project aims to develop an automated pothole detection system using computer vision and deep learning techniques. By leveraging OpenCV for image processing and Tensor Flow Keras or PyTorch for deep learning model development, the system will provide accurate and timely detection of potholes, thereby enhancing road maintenance efficiency.
[011] Publication No. IN202341060083 relates to a solution for pothole detection utilizing convolutional neural network (CNN) technology. Potholes on roads pose significant safety risks and contribute to vehicle damage. This innovation employs Tensor Flow and Keras to develop a CNN-based system for classifying road potholes based on images.
[012] Publication No. IN202321054264 relates to a system for pothole detection. The system comprises a set of sensors a control unit a display device a GPS a cloud and a power supply unit. The set of sensors comprises at least one ultrasonic sensor and at least one proximity sensor.
[013] Publication No. IN202341071902 relates to a device for detecting a pothole in a vehicle by a rearview camera. The vehicle comprises an image capturing unit positioned at a rear side of the vehicle and at least one sensing element for detecting at least one vehicle parameter. The device comprises a processor for detecting a variation in the at least one vehicle parameter. The processor activates the image capturing unit 14 for capturing a cluster of images of the road when the variation is detected.
[014] Publication No. IN202311065153 relates to a pothole detection system designed for vehicles. The system comprises an ultrasonic depth sensor capable of accurately detecting the depth of potholes on the road surface. A signal transmitter is strategically installed on the road infrastructure, while a controller communicates with both the ultrasonic depth sensor and the signal transmitter.
[015] Publication No. IN202421047756 relates to potholes on road surfaces pose an enduring challenge to road safety and infrastructure maintenance. This paper introduces an innovative approach to address this issue by integrating geospatial data, machine learning, and cloud computing. The proposed system aims to automatically detect and assess potholes in real-time, providing timely alerts to relevant stakeholders for efficient repair and maintenance. Geospatial information, including GPS coordinates, is incorporated to precisely locate detected potholes, enabling the prioritization of maintenance efforts and the optimization of repair resources.
[016] Publication No. IN202311073616 relates to the IOT-enabled pothole detection and road safety system is an inventive solution to enhance road safety and road infrastructure maintenance. Utilizing advanced sensors and AI algorithms, the system detects and classifies potholes in real-time, providing immediate alerts to drivers. It offers precise geolocation data for efficient road maintenance and shares critical information with external authorities through a cloud-based platform.
[017] Publication No. IN202441085298 relates to a pothole detection system using modified YOLOv8 Architecture on Zynq ZCU104 UltraScale+ SoC for real-time road monitoring. An object of the invention is to provide a real-time pothole detection system based on a modified YOLOv8 architecture, implemented on the Zynq ZCU104 UltraScale+ SoC platform. The system uses an optimized deep learning model to detect potholes under diverse conditions, leveraging FPGA parallel processing for low-latency inference. Key architectural elements include customized backbone, neck, and head layers, enabling multi-scale detection of potholes.
[018] Publication No. IN202411086529 relates to pothole detection system that overcomes the limitations of prior art by combining multi-sensor fusion, AI-driven analysis, real-time crowd-sourced data, and edge computing. The system utilizes various sensors to collect comprehensive road condition data, which is then processed by AI algorithms for accurate pothole identification and assessment.
[019] Publication No. IN202441083855 relates to road maintenance is one of the relevant issues in urban infrastructure. Potholes and cracks have much to do with driving safety and automobile performance. Traditionally, these defects have been detected through costlier equipment and labor-intensive processes. Smart road paint technology for pothole detection, significantly improving the safety and efficiency of autonomous vehicles.
[020] Publication No. IN202441059170 relates to an innovative Internet of Things (IoT) enabled road surface monitoring system that utilizes machine learning techniques for automated pothole detection and analysis. This system provides a comprehensive solution for efficient road maintenance and enhances road safety for motorists.
[021] Publication No. CN116311173 relates to a multi-sensor fusion unmanned vehicle road surface pothole detection method, which comprises the steps that an unmanned vehicle scans a front road surface through sound waves, a camera shoots when a pothole is detected, and the shot image is preprocessed according to an optical sensor; according to image information shot by the camera, calculating the width sum between the edge of the pothole and the edges of the roads on the two sides; carrying out threshold segmentation on the preprocessed image to obtain a binary image, and counting the number of white pixel points in the binary image to judge whether water exists in the potholes or not; according to the method, the unmanned vehicle runs more stably, unknown road conditions are explored in advance, whether the road bumpy condition meets the advancing requirement of the electric vehicle or not is automatically judged, the driving safety of the electric vehicle is improved, and the driving safety of the electric vehicle is improved.
[022] Publication No. CN101441769 relates to a method for real-time vision positioning for a monocular camera, which belongs to the field of computer vision. The method comprises the following steps: firstly, acquiring an object image characteristic point set to establish an object image database and perform real-time trainings; secondly, modeling the camera to acquire model parameters of the camera; and thirdly, extracting a real-time image characteristic point set by the camera, matching real-time image characteristic points with the characteristic point set in the object database, and removing error matching and performing an affine inspection to acquire characteristic point pairs and object type information.
[023] Publication No. CN111274939 relates to a monocular camera-based road pavement pit damage automatic extraction method, which comprises the following steps of: determining the fluctuation degree of a pavement pit according to a pixel gray value in an image so as to obtain a pit position; aiming at the pavement cross section represented by each row of pixels at the pit position, fitting a pavement cross section curve by taking the position of the pixel in the row as x and the pixel gray value as y; dividing the road surface cross section curve into a left part, a middle part and a right part, and respectively solving the maximum difference values of the highest points B, A and C of the three parts, straight lines AB and AC and the road surface cross section curve on the y axis as the depth data of the road surface cross section potholes; and solving depth data for all pavement cross sections at the same pit position, wherein the maximum depth is the pit depth.
[024] Publication No. WO2016207749 relates to a device and method of locating potholes. The method includes capturing, using a camera, an image including a road surface having at least one pothole therein. Extracting from the image a portion of the image including the road surface having at least one pothole therein.
[025] Publication No. CN111325079 relates to a pavement pit detection method applied to a vehicle-mounted visual system. The pavement pit detection method comprises the following steps: S1, image acquisition: acquiring a pavement pit image in front of a driving road as an input image; s2, inverse perspective transformation is performed on the input image, image coordinates are transformed to a world coordinate plane, and a perspective image effect is transformed to an overlooking effect; s3, performing edge extraction on the inverse perspective transformation image; s4, performing morphological processing on the edge detection result image; s5, closed edge contour detection is carried out, and the area is extracted; and S6, judging the pits in the area, and if the judgment result is yes, outputting the position of the area.
[026] Publication No. KR20200007165 relates to a pothole detection system using a vision sensor mounted on an unmanned aerial vehicle and a method thereof. A position and a size of a pothole scattered across a wide range of areas can be quickly identified by using an unmanned aerial vehicle (drone) to quickly repair the pothole with the highest risk of accidents. In addition, an order of priority of the pothole requiring repair can be determined by realization of measuring the position and the size of the pothole by using an image processing algorithm and a deep learning algorithm after a photogrammetry technique is applied and a porthole image collected from the unmanned aerial vehicle is processed as a normal image.
[027] Patent No. US9416499 provides a system and method for sensing and managing pothole locations and pothole characteristics. An additional aspect of the present invention is to provide a system that may acquire, fuse, and analyze pothole sensing data from several sources to identify potholes in need of maintenance or repair. Further, the system may be configured to create and distribute recurring reports of pothole repair data for use by roadway officials.
[028] Publication No. US2013155061 relates to an autonomous pavement assessment system may receive depth data indicative of the depth of pixels that collectively comprise multiple defective areas of pavement. For each defective area, the system may fit a plane to it; generate a histogram that indicates the frequency of its pixels at different depths; dynamically determine a depth noise threshold for it; generate a binary image of it based on its noise threshold; and generate a depth map of it containing only the pixels that have a depth that meets or exceeds its depth noise threshold.
[029] Publication No. CN104864909 provides a road surface pothole detection device based on vehicle-mounted binocular vision. The road surface pothole detection device comprises a GPS module, a binocular camera module, a vibration sensor module, an outer trigger module, a power supply module, a main controller module, an external storage module, a computer, a high-speed memory and a router.
[030] Publication No. CN112017170 discloses a road pavement pit slot identification method, device and equipment based on a three-dimensional light and shadow model.
[031] Publication No. US2017342669 relates to a vehicle system for automatic repairing of road potholes includes: a laser camera that measures distance from a vehicle to the pothole and calculates surface area based on the image and distance information, a first support shaft that moves back and forth, a multi operation device that cuts and crushes asphalt and flattens asphalt concrete, a heating device that melts asphalt, an asphalt vacuum suction device that sucks in asphalt and stores fragments in a residue storage tank, an asphalt concrete storage tank that stores asphalt concrete and supplies asphalt concrete around the road pothole, a residue storage tank that stores crushed asphalt sucked in, an oil supply nozzle that supplies oil, an air supply pump that supplies strong air for cleaning, and a vehicle device that operates power switch, cutting device motor, asphalt vacuum suction device, air supply pump, asphalt concrete volume calculation part, and roller part.
[032] Publication No. US2014355839 discloses an exemplary apparatus and associated method for analyzing surface degradation. The apparatus can include a sensor configured to acquire images of a surface; and a processing device configured to correlate the acquired images to a geo-coordinate, to extract at least one property of a surface abnormality identified in at least one of the acquired images, and to generate trend data based on changes over time in the at least one property of the surface abnormality identified in the images, which are correlated to a common geo-coordinate.
[033] Patent No. US11810364 relates to a vehicle safety system communicably coupled to a vehicle, comprises processing circuitry coupled to a camera unit and a LiDAR sensor, the processing circuitry to execute logic operative to analyze an image captured by one or more cameras to identify predicted regions of road damage, correlate LiDAR sensor data with the predicted regions of road damage, analyze the LiDAR sensor data correlated with the predicted regions of road damage to identify regions of road damage in three-dimensional space; and output one or more indications of the identified regions of road damage, wherein the processing circuitry is coupled to an interface to the vehicle, the processing circuitry to output an identification of road damage.
[034] Patent No. US10967862 relates to a computing system can receive sensor log data from one or more first autonomous vehicles (AVs) operating throughout a transport service region, and analyze the sensor log data against a set of road anomaly signatures. Based on analyzing the sensor log data against the set of road anomaly signatures, the computing system may identify road anomalies within the transport service region, and generate a resolution response for each road anomaly. The computing system can then transmit the resolution response to one or more second AVs to cause the one or more second AVs to respond to the road anomaly.
[035] Publication No. US2024247928 relates to a system for estimating the depth of an at least partially water-filled pothole by a driver assistance system includes an optical camera sensor configured to detect the pothole based on camera data, a thermal camera sensor configured to determine a road surface temperature including the surface of the pothole based on temperature determination of pixels in a thermal pixel image, a processor configured to determine the surface temperature of the pothole detected by the optical camera sensor, based on the road surface temperature detected by the thermal camera sensor, and estimate the depth of the pothole based on its surface temperature.
[036] Publication No. EP4254361 relates to a system in a vehicle for detecting road anomalies. The system includes one or more sensors and a processor. The one or more sensors are configured to generate one or more images of a view in front of the vehicle. The processor in communication with the one or more sensors. The processor is configured to receive the one or more images from the one or more sensors; detect a front moving vehicle in the one or more images; analyse motion of the front moving vehicle in comparison with an expected path of the front moving vehicle; and determine a form of the road anomaly based on the analysed motion of the front moving vehicle.
[037] Publication No. CN119251807 provides a road surface pothole recognition method, device and equipment, a medium and a vehicle, and relates to the technical field of road surface pothole recognition. The method comprises the following steps: acquiring behavior monitoring data of people in a vehicle in real time; and under the condition of determining that the personnel behavior monitoring data meets the preset behavior condition, obtaining a road surface pothole judgment result that the current road surface has a pothole.
[038] Publication No. CN120219384 relates to a road pothole detection method and system, comprising the following steps: S1: acquiring an original image stream of a road surface through a vehicle-mounted multispectral camera, and performing motion artifact elimination to generate a vibration-compensated image; S2: generating a shadow-suppressed image through asymmetric gamma correction; S3: extracting candidate pothole areas using a dual-threshold connected domain analysis method; S4: determining a surface damaged area when the contrast difference exceeds a texture mutation threshold; S5: extracting the statistical area ratio of continuous pixel clusters, and determining a valid pothole feature when the ratio is greater than 60%; S6: calculating the eccentricity of the minimum enclosing ellipse and the ellipse area ratio to generate a final pothole detection report.
[039] Publication No. ID2025/05155 relates to a method for detecting potholes and estimating their width using deep learning and a depth camera. Autonomous vehicles require the ability to detect obstacles such as potholes and analyze their width. This invention introduces a method for detecting, tracking, and estimating the width of potholes simultaneously and autonomously using only vision sensors that apply artificial intelligence in the form of deep learning.
[040] Publication No. IN202541021974 relates to a system for detecting potholes and providing adaptive notifications to users in real-time. The system includes processors; memory coupled to the processors. The processors receive real-time visual data captured by autonomous vehicles. The processors can process real-time visual data using a pre-stored object detection model to identify surface irregularities indicative of potholes.
[041] Publication No. IN202541049381 relates to a new way of pothole detection automation using image processing and machine learning techniques. Using computer vision, I suggest to detect the potholes in the road surface using the irregularities seen in the road surface. Then, key image processing methods are used to preprocess and enhance the images; these include edge detection, contour analysis, texture segmentation and the like.
[042] Publication No. IN202541058737 relates to a system for real-time pothole detection, that comprises an autonomous pothole detection and reporting robot configured to detect potholes in real-time while autonomously navigating through road environments based on real-time sensor inputs and onboard processing, a communication network configured to enable data transmission between the robot and external monitoring systems, a supervision unit connected to the robot via the communication network and configured to accept operational data from the robot, supervise robot operations, and maintain communication links with external control centers, further comprising a manual intervention module and a robot control module configured to monitor and control the robot remotely. The system also includes a user device operatively connected to the supervision unit and configured to display real-time robot status, pothole detection data, and maintenance alerts.
[043] Publication No. KR20250100571 relates to a pothole detection device and method using a linear laser and a camera, and the device according to the present invention includes a linear laser generator that irradiates a line laser beam at a predetermined angle to a road surface in front of a vehicle to form a linear laser pattern, a camera that photographs the road surface on which the linear laser pattern is formed to obtain a two-dimensional image, and a control unit that extracts a laser profile corresponding to an expected trajectory area of the left and right wheels of the vehicle from the two-dimensional image, calculates a pothole depth based on the laser profile, and outputs a pothole occurrence warning signal when the pothole depth exceeds a predetermined standard.
[044] Publication No. KR20250114968 relates to a pothole detection system applying deep learning to image data collected by an image collection module according to the present invention includes: an image collection module that collects image data together with location information and collection time information; a pothole detection server that receives image data collected by a plurality of the image collection modules, improves the resolution of low-resolution image data using deep learning, detects a pothole from the image data having improved resolution, and calculates size information of the detected pothole; a maintenance company server that receives size information and location information of the detected pothole from the pothole detection server; and a maintenance manager terminal that receives size information and location information of the pothole from the maintenance company server so that the maintenance manager can perform maintenance; thereby having the effect of improving road safety by identifying the point of occurrence of a pothole and taking action.
[045] Publication No. US2021226382 relates to a connector assembly device includes a connector element and a connector position assurance (CPA) device mounted to move relative to the connector element between a delivery position and a locking position.
[046] Patent No. US10843945 relates to a pilot filter system for monitoring water quality in a water treatment system. The pilot filter system includes: a filter vessel containing a downward-moving bed of filtration media, the filter vessel being fluidly connected to the water treatment system such that sample influent water from the water treatment system flows into the filter vessel and over the bed, wherein the filter vessel has a filtrate outlet port through which some filtrate is discharged; a media recycle conduit fluidly connected to the filter vessel; and a media recycler for circulating a mixture of water and used filtration media through the media recycle conduit to a point upstream of the bed in the filter vessel.
[047] Publication No. EP3300000 relates to a method, for a verification process that performs neighbor discovery for one or more feature points projected to an m-dimensional space (m is a natural number equal to or greater than 2), includes: acquiring a feature point group including one or more feature points projected to coordinate values of the m-dimensional space ordered in a coordinate value order on each of two or more coordinate axes that define the m-dimensional space (m is a natural number equal to or greater than 2); selecting a datum axis on which a comparison time number in neighbor discovery is small, the comparison time number being obtained by performing simulation of neighbor discovery.
[048] Reference may be made to an article entitled “Robust video-based pothole detection and area estimation for intelligent vehicles with depth map and kalman smoothing” by Dehao Wang, Haohang Zhu, Yiwen Xu, and Kaiqi Liu; arxiv.org; 27 May 2025 talks about a robust pothole area estimation framework that integrates object detection and monocular depth estimation in a video stream is proposed in this paper. First, to enhance pothole feature extraction and improve the detection of small potholes, ACSH-YOLOv8 is proposed with ACmix module and the small object detection head. Then, the BoT-SORT algorithm is utilized for pothole tracking, while DepthAnything V2 generates depth maps for each frame. With the obtained depth maps and potholes labels, a novel Minimum Bounding Triangulated Pixel (MBTP) method is proposed for pothole area estimation. Finally, Kalman Filter based on Confidence and Distance (CDKF) is developed to maintain consistency of estimation results across consecutive frames. The results show that ACSH-YOLOv8 model achieves an AP (50) of 76.6%, representing a 7.6% improvement over YOLOv8. Through CDKF optimization across consecutive frames, pothole predictions become more robust, thereby enhancing the method’s practical applicability.
[049] Traditionally, the detection and measurement of potholes and similar road surface anomalies have relied on either manual inspection, sensor fusion platforms (using LiDAR, stereo cameras, radar, or inertial measurement units), or basic image-based classification methods. Existing systems that use cameras often focus only on detecting the presence of potholes without providing precise geometric quantification such as area or volume. Others that provide volume estimation require complex and expensive setups involving active depth sensors or stereo vision.
[050] There remains a need for a low-cost, scalable, and accurate system that can not only detect potholes but also quantify their physical dimensions, specifically area and volume, in real time, using minimal hardware.
[051] In order to overcome above listed prior art, the present invention aims to provide a monocular vision-based system and method capable of detecting potholes, estimating their area and volume using image segmentation and depth prediction, and mapping them in real-world coordinates for large-scale infrastructure management. This invention offers a significant advancement by enabling real-time, low-cost, and accurate estimation of both area and volume of road surface anomalies using only a monocular camera. By integrating visual segmentation, monocular depth inference, and pixel-to-world transformations through calibration and homography, the system achieves dimensional accuracy comparable to multi-sensor platforms while maintaining deployment simplicity and affordability. In summary, this invention transforms standard imaging hardware into a complete road anomaly quantification tool, thereby enhancing existing road inspection methods and enabling scalable integration into conventional vehicles, fleets, and mobile platforms.
OBJECTS OF THE INVENTION:
[052] The principal object of the present invention is to provide a real-time camera-based system and method for pothole detection, mapping and quantification.
[053] Another object of the present invention is to provide a system that accurately estimates the area and volume of potholes using a single monocular camera.
[054] Yet another object of the present invention is to enable pixel-to-real-world conversion of segmented pothole regions through camera calibration and homography transformation, ensuring dimensional accuracy in physical units.
[055] Still another object of the present invention is to provide a system and method for depth estimation based on monocular imagery for reconstructing the 3D shape of a pothole, enabling volumetric analysis.
[056] Yet another object of the present invention is to ensure real-time or near-real-time processing of captured road images for on-device execution and scalable deployment in moving vehicles.
[057] Still another object of the present invention is to provide low-cost and easily integrable solution for road agencies, municipalities, and vehicle operators for automated road condition assessment and preventive maintenance planning.
SUMMARY OF THE INVENTION:
[058] The present invention relates to a real-time camera-based system and method for pothole detection, mapping and quantification System. This is a monocular imaging system with a modular pipeline enables real-time estimation of both the area and volume of road surface anomalies using only a single camera input, without the need for stereo vision, LiDAR, or depth sensors.
[059] The system comprises: at least one image capture device configured to obtain images of a road surface; a processing unit operatively coupled to the image capture device and configured to execute a pothole detection algorithm employing a trained object detection model; a depth estimation module coupled to the processing unit and configured to determine one or more dimensional parameters of a detected pothole; a geolocation module configured to associate a geographic coordinate with the detected pothole; and a communication module configured to transmit pothole data comprising said coordinate and said dimensional parameters to a remote server.
[060] In another embodiment, the system is installed on a moving vehicle, wherein the processing unit performs real-time analysis to enable immediate mapping updates.
[061] In yet another embodiment, a plurality of such systems are deployed across multiple vehicles, wherein each system contributes pothole data to a centralized, continuously updated pothole database.
[062] In yet another embodiment, system performs pixel-to-world transformation using camera calibration and road-plane homography, enabling metric quantification of segmented regions and 3D projection of depth values within those regions to generate a point cloud and compute volume, all using monocular input. The invention operates real-time on edge devices and transmit results to a centralized mapping platform enhances its novelty and industrial applicability.
BREIF DESCRIPTION OF THE INVENTION
[063] It is to be noted, however, that the appended drawings illustrate only typical embodiments of this invention and are therefore not to be considered for limiting of its scope, for the invention may admit to other equally effective embodiments.
[064] FIG. 1 illustrates the system setup comprising a monocular camera mounted on a vehicle for capturing road surface images used in pothole detection and quantification.
[065] FIG. 2 shows the high-level block diagram of the system architecture including the image capture module, processing modules for detection, segmentation, depth estimation, and volume calculation, and communication interface for map-based reporting.
[066] FIG. 3 depicts the workflow for pothole area estimation, where object detection is followed by segmentation, boundary extraction, and area computation using geometric principles.
[067] FIG. 4 illustrates the workflow for pothole volume estimation, comprising depth map generation, 3D coordinate conversion using camera parameters, point cloud construction, and convex hull-based volume computation.
[068] FIG. 5 shows the camera calibration setup using a checkerboard pattern for estimating intrinsic and extrinsic parameters essential for accurate pixel-to-world coordinate transformation.
[069] FIG. 6 demonstrates the data transmission and mapping architecture, where computed area and volume data are sent to a cloud server for aggregation, deduplication, and visualization.
[070] FIG. 7 provides an example of a geospatial map displaying detected potholes along with estimated area and volume data for infrastructure maintenance planning.
DETAILED DESCRIPTION OF THE INVENTION:
[071] The present invention provides a system and method for the automated detection and quantification of road surface anomalies, particularly potholes, using a monocular camera. The system enables real-time estimation of physical dimensions, including area and volume, without reliance on stereo vision, active depth sensors, or inertial measurement units. The invention may be embodied in a vehicle-mounted configuration or any suitable mobile imaging platform.
[072] Referring to FIG. 1, the system comprises a vehicle equipped with a monocular imaging device (100) rigidly mounted to face the road surface (300). The imaging device is operably connected to a processing unit (200), which may be integrated onboard the vehicle. The processing unit is powered by a power supply and may optionally interface with a positioning module for determining geographic coordinates and a communication module for transmitting processed results. A local display or indicator module may be used to convey detection feedback to the user.
[073] FIG. 2 illustrates the functional blocks of the processing pipeline. An image capture module receives input from the monocular camera (100), GPS Module (400), power setup (500) and forwards it to a detection module (600), which identifies candidate road anomalies based on visual features within the image. The identified regions are passed to a segmentation module, which generates pixel-wise masks delineating the extent of the surface depression. A depth estimation module concurrently processes the image data to generate pixel-level depth values for the observed scene. Outputs from the segmentation and depth estimation modules are supplied to an area estimation module and a volume estimation module (700). The area estimation module (700) includes a submodule for extracting boundary coordinates from the segmented region and applies a geometric computation algorithm to estimate the enclosed surface area. A pixel-to-world transformation module utilizes (900) previously calibrated camera parameters to convert pixel dimensions into physical units. The volume estimation module (700) includes a 3D projection module that transforms masked pixel and depth information into three-dimensional spatial coordinates. A point cloud generation module (1000) then constructs a 3D representation of the pothole geometry. The volume estimation module computes the enclosed volume using a surface-based reconstruction technique. The resulting data, including area and volume, may be geotagged and transmitted to a server or mapped locally.
[074] FIG. 3 depicts the process for estimating pothole area. The image capture module acquires a frame that is processed by the detection module to produce a region of interest. The segmentation module (700) generates a binary mask corresponding to the detected anomaly. A boundary extraction submodule identifies the perimeter of the masked region, and the area estimation module determines the physical area of the surface depression by applying geometric computations on the transformed boundary coordinates. The transformation from image space to world coordinates is facilitated by a calibration-based module, which uses intrinsic and extrinsic camera parameters to derive scale and perspective information for accurate area estimation.
[075] FIG. 4 illustrates the steps involved in volume estimation (800). The depth estimation module produces a depth representation of the image, which, in combination with the segmented region from the segmentation module, enables isolation of the target anomaly in three dimensions.
[076] Using intrinsic camera parameters and road-plane assumptions, the 3D projection module maps pixel locations and associated depth values to spatial coordinates. The point cloud generation module constructs a three-dimensional point set for the anomaly region. The volume estimation module computes the volume enclosed by this point set using a geometric reconstruction approach, which may include, but is not limited to, convex or surface-meshing techniques. The resulting volume is expressed in physical units and optionally associated with confidence metrics based on reconstruction density.
[077] FIG. 5 shows the camera calibration procedure (900). A reference calibration pattern of known dimensions is imaged under controlled conditions to derive intrinsic parameters, including focal lengths and principal point, as well as distortion coefficients and extrinsic orientation with respect to the road plane. The calibration parameters are used to establish a planar transformation between the image and road surface, enabling accurate projection of segmented and depth data into metric space. These transformations are fundamental to both area and volume estimation modules and ensure repeatable and scalable field deployment.
[078] FIG. 6 presents a schematic representation of the data transmission and mapping workflow. Output data generated by the processing unit (200), including geolocation, surface area, volume, and optional severity indicators, is transmitted to a server via the communication interface.
[079] The server (1000) comprises a data storage unit, a data validation and merging engine, and a mapping service. These components manage the aggregation, deduplication, and visualization of anomaly records across multiple observations and vehicles. An operator interface and application programming interface provide access to the processed data for further analysis, prioritization, and integration with road asset management systems.
[080] FIG. 7 illustrates an example output interface. Pothole (300) detections are visualized on a map with overlays indicating dimensions, confidence, severity, and status (e.g., newly detected, observed multiple times, marked as repaired). Filtering and sorting tools enable stakeholders to query and prioritize entries based on pre-defined thresholds or operational constraints.
[081] The invention will be more fully understood from the following examples. These examples are to be constructed as illustrative of the invention and not limitative thereof:
[082] Example 1: A vehicle-mounted monocular camera connected to an processing unit. The processing unit executes a convolutional neural network trained on road defect imagery to identify potholes. Upon detection, the system computes the pothole’s approximate width and depth using monocular depth estimation. GPS coordinates are recorded and transmitted via a cellular LTE module to a municipal maintenance server.
[083] Example 2: Introduces a stereo camera setup and uses stereo disparity calculations for depth estimation, which contradicts the core inventive step of this invention that relies solely on a single monocular camera with monocular depth inference. Including stereo vision may cause confusion during examination and weaken the novelty claims.
[084] Example 3: A stationary roadside unit equipped with a high-resolution camera and solar-powered processing node. The system continuously monitors a section of a highway, detecting pothole formation over time. Dimensional changes are recorded, and trend analysis is performed to forecast deterioration rates, enabling proactive maintenance scheduling.
[085] Example 4: A hybrid deployment combining vehicle-based and roadside systems. Data from both sources is aggregated in the central server, with duplicate detections reconciled to improve location accuracy and severity scoring.
[086] Example 5: Integration with an augmented reality maintenance interface. When maintenance crews approach a reported pothole location, the system overlays measured pothole dimensions on a mobile device screen to assist in repair preparation.
[087] The present disclosure provides several technical and practical advantages over existing technologies and methods for road surface anomaly detection and quantification. These include, but are not limited to, the following:
[088] The system operates solely with a monocular imaging device, eliminating the need for stereo cameras, LiDAR, radar, or depth sensors. This significantly reduces hardware costs, simplifies installation, and enables integration into low-cost platforms, including standard vehicles and mobile inspection units.
[089] Unlike conventional image-based systems that only detect or classify potholes, the present invention enables precise area and volume quantification in physical units through pixel-to-world transformations and 3D geometric processing.
[090] The use of camera calibration and homography-based transformations ensures that pixel-level segmentation and depth data are reliably converted into accurate real-world measurements without requiring GNSS, IMU, or external positioning sensors.
[091] The architecture supports real-time or near-real-time processing on lightweight edge computing devices, eliminating the need for cloud-based inference and enabling offline operation in bandwidth-constrained environments.
[092] Processed results (area, volume, location, severity) can be efficiently transmitted to a central server for mapping, aggregation, and deduplication. The system supports integration into fleet-based maintenance operations and municipal asset management workflows.
[093] The system enables continuous monitoring of pothole lifecycle states (e.g., new, growing, repaired) by associating detections with location, time, and confidence metrics, thereby enhancing maintenance planning and budget prioritization.
[094] The invention allows fully contactless quantification of potholes without requiring physical interaction with the road surface, manual measurements, or invasive ground-based sensors.
[095] The use of segmentation and depth estimation techniques allows the system to operate under varying lighting conditions and partial occlusions, improving reliability in real-world environments.
[096] The modular structure of the system allows it to be integrated into a wide range of platforms, including dashcams, smartphones, vehicle inspection cameras, and UAVs, without dependence on proprietary hardware.
[097] By producing continuous and scalable data on pothole severity and progression, the system supports proactive maintenance interventions, reducing long-term repair costs and improving road safety.
[098] The system significantly reduces the cost and complexity of implementation, particularly for deployment on conventional vehicles, fleet-based systems, or autonomous platforms.
[099] Numerous modifications and adaptations of the system of the present invention will be apparent to those skilled in the art, and thus it is intended by the appended claims to cover all such modifications and adaptations which fall within the true spirit and scope of this invention.
, Claims:WE CLAIM:
1. A real-time camera-based system for pothole detection, mapping and quantification comprises
a) a vehicle equipped with at least one monocular imaging device (100) to obtain images of a road surface rigidly mounted to face the road surface (300), a power supply (500) and global positioning system (400)
b) a processing unit (200), integrated onboard the vehicle operatively coupled to the at least one image capture device, powered by a power supply (500), interfaced with a positioning module for determining geographic coordinates and a communication module for transmitting processed results including local display or indicator module to convey detection feedback to the user characterized in that
o a detection module (600), to identify candidate road anomalies based on visual features within the image
o a geolocation module (400) configured to determine a geographic coordinate associated with each identified pothole; and
o a segmentation module, which generates pixel-wise masks delineating the extent of the surface depression and depth estimation module concurrently processes the image data to generate pixel-level depth values for the observed scene wherein the outputs from the segmentation and depth estimation modules are supplied to an area estimation module and a volume estimation module (700 ) which includes a submodule for extracting boundary coordinates from the segmented region and applies a geometric computation algorithm to estimate the enclosed surface area.
o A pixel-to-world transformation module utilizes (900) previously calibrated camera parameters to convert pixel dimensions into physical units.
o A point cloud generation module (1000) then constructs a 3D representation of the pothole geometry.
2. The real-time camera-based system for pothole detection, as claimed in claim 1, wherein the road surface anomaly is a pothole, crack, or depression.
3. The real-time camera-based system for pothole detection, as claimed in claim 1, wherein the image capture device comprises a monocular camera.
4. The real-time camera-based system for pothole detection, as claimed in claim 1, wherein the conversion of pixel coordinates to world coordinates is performed using a homography matrix derived from camera calibration.
5. The real-time camera-based system for pothole detection, as claimed in claim 1, wherein the estimated volume is computed by generating a convex surface envelope around the projected three-dimensional coordinates
6. The real-time camera-based method for pothole detection, as claimed in claim 1, wherein the process for estimating pothole area includes following steps:
a) acquiring a frame by image capture module that is processed by the detection module to produce a region of interest.
b) generating a binary mask corresponding to the detected anomaly using segmentation module (700).
c) identifying the perimeter of the masked region, and the area estimation module using boundary extraction submodule and determine the physical area of the surface depression by applying geometric computations on the transformed boundary coordinates.
d) transformation from image space to world coordinates facilitated by a calibration-based module, which uses intrinsic and extrinsic camera parameters to derive scale and perspective information for accurate area estimation.
7. The real-time camera-based system and method for pothole detection, as claimed in claim 1, wherein the in volume estimation (800) includes following steps:
a) The depth estimation module produces a depth representation of the image, which, in combination with the segmented region from the segmentation module, enables isolation of the target anomaly in three dimensions.
b) Using intrinsic camera parameters and road-plane assumptions, the 3D projection module maps pixel locations and associated depth values to spatial coordinates.
c) The point cloud generation module constructs a three-dimensional point set for the anomaly region.
d) The volume estimation module computes the volume enclosed by this point set using a geometric reconstruction approach and resulting volume is expressed in physical units and optionally associated with confidence metrics based on reconstruction density.
8. The real-time camera-based system and method for pothole detection, as claimed in claim 1 and 6, wherein the geometric reconstruction approach includes, convex or surface-meshing techniques.
9. The real-time camera-based system and method for pothole detection, as claimed in claim 1 and 6, wherein the segmentation of the region of interest is conditioned using detection prompts generated by a prior object detection module.
10. The real-time camera-based system and method for pothole detection, as claimed in claim 1 and 6, wherein the pixel-to-world coordinate transformation is computed using intrinsic and extrinsic parameters obtained via a camera calibration procedure.