1. A system (100) for detection of suspicious objects under uncertain climatic conditions in a border surveillance environment, the system (100) comprising: a surveillance camera (102) configured to capture visual data from a monitored border region; a processing unit (104) operatively connected to the surveillance camera (102); and a memory (106) storing instructions that, when executed by the processing unit (104), cause the following operations, characterized in that the processing unit (104) is configured to: receive the visual data captured by the surveillance camera (102); perform weather-aware preprocessing on the visual data to enhance image quality according to environmental conditions including rain, fog, sandstorm, and snowfall; extract spatial features from the pre-processed visual data through a convolutional neural network architecture; extract contextual features from the pre-processed visual data through a Swin Transformer architecture configured to capture global relationships within a visual scene; combine the spatial features and the contextual features through a dynamic weather-gating mechanism that adjusts feature contribution according to the detected environmental condition; perform localization and classification of objects within the visual data through an object detection framework based on a YOLOv8 backbone using fused features obtained from the dynamic weather-gating mechanism; estimate prediction uncertainty associated with detected objects through an uncertainty estimation process; generate detection results comprising object class information, bounding box coordinates, confidence score, and uncertainty level for identification of suspicious objects in the monitored border region; and transmit the generated detection results to a computing device (200) through a communication unit (108) operatively coupled to the processing unit (104) for further analysis and monitoring.

2. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to display the generated detection results onto a user interface (202) installed on the computing device (200).

3. The system (100) as claimed in claim 1, wherein the detected objects are selected from unmanned aerial vehicles, weapons, unauthorized human activity, or a combination thereof present in the monitored border region.

4. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to perform rain removal, dehazing, contrast enhancement, noise suppression, and snow artifact reduction during the weather-aware preprocessing according to the detected climatic condition.

5. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to analyse long-range spatial relationships within the visual scene through the Swin Transformer architecture to support contextual understanding of complex environments.

6. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to determine feature weighting between spatial features and contextual features through the dynamic weather-gating mechanism based on environmental cues present in the visual data.

7. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to perform anchor-free object detection through the YOLOv8 backbone for identification of objects with different spatial scales.

8. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to quantify prediction confidence associated with detected objects under degraded visual conditions through probabilistic inference.

9. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to initiate a secondary verification process when the uncertainty level associated with detected objects exceeds a predefined threshold.

10. A method (400) for detection of suspicious objects under uncertain climatic conditions in a border surveillance environment, the method (400) is characterized by steps of: receiving visual data captured by a surveillance camera (102); performing weather-aware preprocessing on the visual data to enhance image quality according to environmental conditions including rain, fog, sandstorm, and snowfall; extracting spatial features from the pre-processed visual data through a convolutional neural network architecture; extracting contextual features from the pre-processed visual data through a Swin Transformer architecture configured to capture global relationships within a visual scene; combining the spatial features and the contextual features through a dynamic weather-gating mechanism that adjusts feature contribution according to the detected environmental condition; performing localization and classification of objects within the visual data through an object detection framework based on a YOLOv8 backbone using fused features obtained from the dynamic weather-gating mechanism; estimating prediction uncertainty associated with detected objects through an uncertainty estimation process; generating detection results comprising object class information, bounding box coordinates, confidence score, and uncertainty level for identification of suspicious objects in the monitored border region; and transmitting the generated detection results to a computing device (200) through a communication unit (108) operatively coupled to the processing unit (104) for further analysis and monitoring. Date: April 08, 2026 Place: Noida Nainsi Rastogi Patent Agent (IN/PA-2372) Agent for the Applicant

Description:BACKGROUND
Field of Invention
[001] Embodiments of the present invention generally relate to computer vision-based surveillance system and particularly to a system and method for detection of suspicious objects.
Description of Related Art
[002] Border surveillance at sensitive regions such as the Line of Control presents significant technical challenges due to extreme and unpredictable environmental conditions. Adverse weather conditions such as heavy rain, dense fog, sandstorms, and snowfall reduce visibility and degrade the quality of captured visual data. Such degradation leads to difficulty in identifying suspicious objects, including unauthorized personnel, weapons, and aerial threats. The presence of noise, low contrast, occlusion, and distortion in visual scenes further complicates accurate object detection, thereby affecting the reliability of surveillance systems deployed in such environments.
[003] Existing surveillance systems rely on conventional computer vision techniques and deep learning-based object detection models such as Convolutional Neural Networks and You Only Look Once frameworks. These systems process visual data obtained from cameras and perform object detection based on pre-trained models developed using standard datasets. Some systems incorporate image enhancement techniques or fixed preprocessing methods to improve visual clarity under certain environmental conditions. In addition, recent approaches employ transformer-based architectures to capture contextual information within visual scenes for improved detection performance.
[004] However, existing solutions exhibit several limitations when deployed in adverse and uncertain climatic conditions. Conventional models fail to generalize effectively across varying weather scenarios due to dependence on clear-weather training data. Fixed preprocessing techniques lack adaptability to dynamic environmental changes and do not address diverse distortion patterns. Furthermore, existing systems provide deterministic outputs without any indication of prediction reliability, which may result in false detections or missed threats. The absence of mechanisms to adapt feature extraction based on environmental conditions further reduces detection accuracy and overall system robustness in real-world border surveillance scenarios.
[005] There is thus a need for an improved and advanced system and method for detection of suspicious objects that can administer the aforementioned limitations in a more efficient manner.
SUMMARY
[006] Embodiments in accordance with the present invention provide system for detection of suspicious objects. The system comprising a surveillance camera configured to capture visual data from a monitored border region. The system further comprising a processing unit operatively connected to the surveillance camera, characterized in that the processing unit is configured to receive the visual data captured by the surveillance camera, perform weather-aware preprocessing on the visual data to enhance image quality according to environmental conditions including rain, fog, sandstorm, and snowfall, extract spatial features from the pre-processed visual data through a convolutional neural network architecture, extract contextual features from the pre-processed visual data through a Swin Transformer architecture configured to capture global relationships within a visual scene, combine the spatial features and the contextual features through a dynamic weather-gating mechanism that adjusts feature contribution according to the detected environmental condition, perform localization and classification of objects within the visual data through an object detection framework based on a YOLOv8 backbone using fused features obtained from the dynamic weather-gating mechanism, estimate prediction uncertainty associated with detected objects through an uncertainty estimation process, generate detection results comprising object class information, bounding box coordinates, confidence score, and uncertainty level for identification of suspicious objects in the monitored border region; and transmit the generated detection results to a computing device through a communication unit operatively coupled to the processing unit.
[007] Embodiments in accordance with the present invention further provide a method for detection of suspicious objects. The method comprising steps of: receiving the visual data captured by the surveillance camera, performing weather-aware preprocessing on the visual data to enhance image quality according to environmental conditions including rain, fog, sandstorm, and snowfall, extracting spatial features from the pre-processed visual data through a convolutional neural network architecture, extracting contextual features from the pre-processed visual data through a Swin Transformer architecture configured to capture global relationships within a visual scene, combining the spatial features and the contextual features through a dynamic weather-gating mechanism that adjusts feature contribution according to the detected environmental condition, performing localization and classification of objects within the visual data through an object detection framework based on a YOLOv8 backbone using fused features obtained from the dynamic weather-gating mechanism, estimating prediction uncertainty associated with detected objects through an uncertainty estimation process, generating detection results comprising object class information, bounding box coordinates, confidence score, and uncertainty level for identification of suspicious objects in the monitored border region; and transmitting the generated detection results to a computing device through a communication unit operatively coupled to the processing unit for further analysis and monitoring.
[008] Embodiments of the present invention may provide a number of advantages depending on their particular configuration. First, embodiments of the present application may provide system for detection of suspicious objects.
[009] Next, embodiments of the present application may provide system for detection of suspicious objects that improves detection accuracy of suspicious objects under adverse environmental conditions including rain, fog, sandstorm, and snowfall.
[0010] Next, embodiments of the present application may provide system for detection of suspicious objects that enhances robustness and reliability of surveillance systems by adapting feature extraction based on dynamic weather conditions.
[0011] Next, embodiments of the present application may provide system for detection of suspicious objects that reduces false positives and missed detections through incorporation of prediction uncertainty estimation.
[0012] Next, embodiments of the present application may provide system for detection of suspicious objects that enables real-time or near real-time object detection and monitoring suitable for deployment in resource-constrained border regions.
[0013] Next, embodiments of the present application may provide system for detection of suspicious objects that provides adaptive fusion of spatial and contextual features for improved identification of partially occluded, distant, or low-visibility objects.
[0014] Next, embodiments of the present application may provide method for detection of suspicious objects that enables adaptive preprocessing of visual data based on detected environmental conditions to maintain consistent input quality.
[0015] Next, embodiments of the present application may provide method for detection of suspicious objects that facilitates effective integration of multi-scale feature representations through coordinated extraction of local and global information.
[0016] Next, embodiments of the present application may provide method for detection of suspicious objects that allows dynamic adjustment of feature contribution using a weather-dependent gating strategy for optimized decision-making.
[0017] Next, embodiments of the present application may provide method for detection of suspicious objects that supports reliable interpretation of detection outputs by incorporating uncertainty-aware evaluation of predictions.
[0018] Next, embodiments of the present application may provide method for detection of suspicious objects that enables scalable and flexible implementation across different surveillance platforms and deployment scenarios.
[0019] These and other advantages will be apparent from the present application of the embodiments described herein.
[0020] The preceding is a simplified summary to provide an understanding of some embodiments of the present invention. This summary is neither an extensive nor exhaustive overview of the present invention and its various embodiments. The summary presents selected concepts of the embodiments of the present invention in a simplified form as an introduction to the more detailed description presented below. As will be appreciated, other embodiments of the present invention are possible utilizing, alone or in combination, one or more of the features set forth above or described in detail below.
BRIEF DESCRIPTION OF THE DRAWINGS
[0021] The above and still further features and advantages of embodiments of the present invention will become apparent upon consideration of the following detailed description of embodiments thereof, especially when taken in conjunction with the accompanying drawings, and wherein:
[0022] FIG. 1 illustrates a block diagram of system for detection of suspicious objects, according to an embodiment of the present invention;
[0023] FIG. 2 illustrates a connectivity diagram of the system with a computing device, according to an embodiment of the present invention;
[0024] FIG. 3 illustrates components of processing unit of the system for detection of suspicious objects, according to an embodiment of the present invention; and
[0025] FIG. 4 depicts a flowchart of a method for detection of suspicious objects, according to an embodiment of the present invention.
[0026] The headings used herein are for organizational purposes only and are not meant to be used to limit the scope of the description or the claims. As used throughout this application, the word "may" is used in a permissive sense (i.e., meaning having the potential to), rather than the mandatory sense (i.e., meaning must). Similarly, the words “include”, “including”, and “includes” mean including but not limited to. To facilitate understanding, like reference numerals have been used, where possible, to designate like elements common to the figures. Optional portions of the figures may be illustrated using dashed or dotted lines, unless the context of usage indicates otherwise.
DETAILED DESCRIPTION
[0027] The following description includes the preferred best mode of one embodiment of the present invention. It will be clear from this description of the invention that the invention is not limited to these illustrated embodiments but that the invention also includes a variety of modifications and embodiments thereto. Therefore, the present description should be seen as illustrative and not limiting. While the invention is susceptible to various modifications and alternative constructions, it should be understood, that there is no intention to limit the invention to the specific form disclosed, but, on the contrary, the invention is to cover all modifications, alternative constructions, and equivalents falling within the scope of the invention as defined in the claims.
[0028] In any embodiment described herein, the open-ended terms "comprising", "comprises”, and the like (which are synonymous with "including", "having” and "characterized by") may be replaced by the respective partially closed phrases "consisting essentially of", “consists essentially of", and the like or the respective closed phrases "consisting of", "consists of”, the like.
[0029] As used herein, the singular forms “a”, “an”, and “the” designate both the singular and the plural, unless expressly stated to designate the singular only.
[0030] FIG. 1 illustrates a block diagram of a system 100 for detection of suspicious objects under uncertain climatic conditions in a border surveillance environment, according to an embodiment of the present invention. The system 100 may provide a robust, adaptive, and reliable surveillance system that may intelligently process visual data, dynamically adapt to environmental conditions, and provide accurate detection results suitable for monitoring, alert generation, and decision-making in real-time.
[0031] According to the embodiments of the present invention, the system 100 may incorporate non-limiting hardware components to enhance the processing speed and efficiency such as the system 100 may comprise a surveillance camera 102, a processing unit 104, a memory 106, and a communication unit 108 operatively coupled to a computing device 200 (as shown in FIG. 2). In an embodiment of the present invention, the hardware components of the system 100 may be integrated with computer-executable instructions for overcoming the challenges and the limitations of the existing systems.
[0032] In an embodiment of the present invention, the surveillance camera 102 may be configured to capture visual data from a monitored border region. The visual data may comprise image frames or video streams indicative of surrounding environmental conditions and objects present within the monitored region.
[0033] In an embodiment of the present invention, the monitored border region may be indicative of a geographical area requiring continuous surveillance for detection of suspicious activities or objects. The monitored border region may include, but not limited to, a Line of Control region, a national border, a restricted security zone, a remote terrain, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the monitored border region, including known, related art, and/or later developed technologies.
[0034] The surveillance camera 102 may be, but not limited to, a visible spectrum camera, an infrared camera, a thermal imaging camera, imaging device, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the surveillance camera 102, including known, related art, and/or later developed technologies.
[0035] In an embodiment of the present invention, the processing unit 104 may be operatively connected to the surveillance camera 102 and configured to receive the visual data captured by the surveillance camera 102. In an embodiment of the present invention, the processing unit 104 may be configured to perform weather-aware preprocessing on the received visual data to enhance image quality based on environmental conditions such as rain, fog, sandstorm, and snowfall. In an embodiment of the present invention, the detected objects may include unmanned aerial vehicles, weapons, and unauthorized human activity within the monitored border region.
[0036] The processing unit 104 may be configured to classify detected objects into predefined categories based on feature representations extracted from the visual data. In an embodiment of the present invention, the processing unit 104 may be configured to perform rain removal using streak suppression filters, dehazing using atmospheric scattering models, contrast enhancement using histogram equalization, noise suppression using spatial filtering techniques, and snow artifact reduction using glare removal algorithms. Selection of preprocessing technique may depend on detected climatic condition associated with the visual data.
[0037] The preprocessing may include, but not limited to, noise reduction, dehazing, contrast enhancement, deblurring, or removal of weather-induced artifacts. Embodiments of the present invention are intended to include or otherwise cover any type of preprocessing techniques, including known, related art, and/or later developed technologies. The pre-processed visual data may be further processed by the processing unit 104 to extract spatial features using a convolutional neural network architecture. The spatial features may represent localized characteristics such as edges, textures, and object boundaries within the visual data. In addition, the processing unit 104 may extract contextual features from the pre-processed visual data using a Swin Transformer architecture configured to capture global relationships and dependencies within a visual scene.
[0038] In an embodiment of the present invention, the processing unit 104 may be configured to analyse long-range spatial relationships within the visual scene through the Swin Transformer architecture by segmenting the visual data into hierarchical windows and computing self-attention across shifted windows. The contextual feature representation may capture global dependencies for improved understanding of complex environments.
[0039] In an embodiment of the present invention, the processing unit 104 may be configured to selectively apply preprocessing techniques based on the detected environmental condition, wherein rain conditions may correspond to application of rain streak removal and noise suppression techniques, fog or sandstorm conditions may correspond to dehazing and contrast enhancement techniques, and snowfall conditions may correspond to snow artifact removal and glare reduction techniques. The processing unit 104 may dynamically adjust preprocessing parameters such as filter strength, kernel size, and enhancement coefficients based on severity of the detected environmental condition.
[0040] In an embodiment of the present invention, the processing unit 104 may be configured to process the pre-processed visual data through a parallel multi-branch architecture comprising a convolutional neural network branch and a Swin Transformer branch. The convolutional neural network branch may be configured to extract localized spatial features including edges, textures, and object boundaries, while the Swin Transformer branch may be configured to extract global contextual features including long-range dependencies and spatial relationships across the visual scene. The parallel processing may enable simultaneous extraction of complementary feature representations for subsequent fusion.
[0041] In an embodiment of the present invention, the processing unit 104 may be configured to update preprocessing parameters and feature weighting parameters based on feedback obtained during operation. The processing unit 104 may be configured to implement a self-adaptive learning mechanism that refines preprocessing strategies and feature fusion behaviour based on historical detection performance and environmental variations. The self-adaptive learning mechanism may enable the system 100 to adjust to new weather patterns and unseen environmental conditions without requiring complete retraining.
[0042] In an embodiment of the present invention, the processing unit 104 may be configured to estimate prediction uncertainty associated with detected objects using probabilistic inference techniques. The uncertainty estimation process may comprise estimation of epistemic uncertainty representing model uncertainty due to limited training data and aleatoric uncertainty representing data uncertainty due to noise and environmental distortions. The processing unit 104 may be configured to implement Bayesian inference mechanisms, Monte Carlo dropout techniques, or probabilistic output layers to quantify uncertainty associated with object detection predictions under degraded visual conditions.
[0043] In an embodiment of the present invention, the processing unit 104 may be configured to train the convolutional neural network architecture, the Swin Transformer architecture, and the object detection framework based on the YOLOv8 backbone using a combination of publicly available adverse weather datasets and domain-specific datasets. The publicly available datasets may include, but not limited to, Adverse Conditions Dataset with Correspondences (ACDC), Driving in Dense Fog Dataset (DENSE), and Detection in Adverse Weather Nature Dataset (DAWN). Further, the processing unit 104 may be configured to utilize a custom-generated dataset representing Line of Control scenarios, wherein the custom-generated dataset may comprise simulated variations in visibility, illumination, occlusion, and object appearance corresponding to rain, fog, sandstorm, and snowfall conditions. The processing unit 104 may be further configured to learn correlations between environmental conditions and optimal feature weighting parameters associated with the dynamic weather-gating mechanism during a training phase, thereby enabling adaptive performance during deployment without retraining.
[0044] In an embodiment of the present invention, the processing unit 104 may be configured to determine feature weighting between spatial features and contextual features through the dynamic weather-gating mechanism using environmental cues extracted from the visual data. The dynamic weather-gating mechanism may assign adaptive weights to feature maps generated from the convolutional neural network architecture and the Swin Transformer architecture for optimized feature fusion.
[0045] In an embodiment of the present invention, the processing unit 104 may be configured to perform tracking of detected objects across consecutive frames of the visual data. The tracking may be performed using temporal association techniques including Kalman filtering, optical flow estimation, or deep learning-based tracking algorithms. The tracking may enable continuous monitoring of movement patterns associated with suspicious objects within the monitored border region.
[0046] In an embodiment of the present invention, the processing unit 104 may be further configured to determine environmental conditions associated with the visual data by analysing environmental cues present within the visual data. The environmental cues may include, but not limited to, fog density, rain streak patterns, snow particle distribution, sand particle concentration, illumination intensity, and contrast degradation. The processing unit 104 may be configured to implement a weather classification model or feature-based inference mechanism to identify a current environmental condition from a predefined set of climatic conditions including rain, fog, sandstorm, and snowfall. The detected environmental condition may be utilized as an input parameter for controlling the weather-aware preprocessing and the dynamic weather-gating mechanism.
[0047] In an embodiment of the present invention, the processing unit 104 may be configured to combine the spatial features and the contextual features through a dynamic weather-gating mechanism. The dynamic weather-gating mechanism may be configured to adjust contribution of the spatial features and the contextual features based on detected environmental conditions. This adaptive feature fusion may enable improved detection performance under varying climatic scenarios. The fused features may be provided to an object detection framework based on a YOLOv8 backbone for performing localization and classification of objects within the visual data.
[0048] In an embodiment of the present invention, the processing unit 104 may be configured to perform anchor-free object detection through the YOLOv8 backbone using grid-based prediction of object centre coordinates, bounding box dimensions, and classification scores. The anchor-free formulation may enable detection of objects across multiple spatial scales within the visual data. The object detection framework may be configured to identify objects of interest including suspicious entities present within the monitored border region. Embodiments of the present invention are intended to include or otherwise cover any type of object detection frameworks, including known, related art, and/or later developed technologies.
[0049] In an embodiment of the present invention, the processing unit 104 may be configured to estimate prediction uncertainty associated with the detected objects through an uncertainty estimation process. The uncertainty estimation process may be configured to evaluate reliability of detection results under uncertain climatic conditions. Based on the detection and uncertainty estimation, the processing unit 104 may generate detection results comprising object class information, bounding box coordinates, confidence score, and uncertainty level for identification of suspicious objects in the monitored border region. In an embodiment of the present invention, the processing unit 104 may be configured to quantify prediction confidence associated with detected objects using probabilistic inference techniques. The probabilistic inference techniques may include Bayesian modelling or stochastic sampling applied to detection outputs to evaluate reliability of predictions under degraded visual conditions.
[0050] In an embodiment of the present invention, the processing unit 104 may be configured to initiate a secondary verification process when the uncertainty level associated with detected objects exceeds a predefined threshold. The secondary verification process may include, but not limited to, activation of an auxiliary sensing modality including thermal imaging camera, infrared camera, or radar-based sensing system, or transmission of detection results to a human operator for manual verification. In an embodiment of the present invention, the processing unit 104 may be configured to initiate a secondary verification process upon detection of uncertainty level exceeding a predefined threshold. The secondary verification process may include activation of auxiliary sensing systems or transmission of detection data to the computing device 200 for further evaluation.
[0051] In an embodiment of the present invention, the processing unit 104 may be configured to reduce false positive detections and missed detections by incorporating uncertainty-aware filtering of detection results. The processing unit 104 may be configured to suppress or flag detection outputs having uncertainty levels exceeding a predefined threshold, thereby improving reliability of detection outcomes under adverse environmental conditions.
[0052] The processing unit 104 may be, but not limited to, a microprocessor, a microcontroller, a graphics processing unit (GPU), an edge computing device, computing hardware, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the processing unit 104, including known, related art, and/or later developed technologies. In an embodiment of the present invention, the system 100 may be configured for deployment on edge computing platforms including embedded graphical processing units and low-power computing devices. The processing unit 104 may be configured to execute optimized versions of the convolutional neural network architecture, the Swin Transformer architecture, and the object detection framework based on the YOLOv8 backbone through model compression, quantization, and pruning techniques. The system 100 may be configured to operate on edge devices including NVIDIA Jetson platforms to enable real-time or near real-time object detection in resource-constrained and remote border environments with reduced latency and energy consumption. The processing unit 104 may further be explained in detail in conjunction with FIG. 3.
[0053] In an embodiment of the present invention, the memory 106 may be operatively coupled to the processing unit 104 and may be configured to store computer-executable instructions. The memory 106 may include volatile memory, non-volatile memory, or a combination thereof. The instructions stored in the memory 106, when executed by the processing unit 104, may enable processing of the visual data for detection of suspicious objects under uncertain climatic conditions. The memory 106 may be, but not limited to, a Random-Access Memory (RAM), a Static Random-access Memory (SRAM), a Dynamic Random-access Memory (DRAM), a Read Only Memory (ROM), an Erasable Programmable Read-only Memory (EPROM), an Electrically Erasable Programmable Read-only Memory (EEPROM), a NAND Flash, a Secure Digital (SD) memory, a cache memory, a Hard Disk Drive (HDD), a Solid-State Drive (SSD), and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the memory 106, including known, related art, and/or later developed technologies.
[0054] In an embodiment of the present invention, the communication unit 108 may be operatively coupled to the processing unit 104 and may be configured to transmit the generated detection results to the computing device 200 for further analysis and monitoring. The communication unit 108 may utilize communication protocols such as, but not limited to, wireless communication, cellular communication, satellite communication, wired communication, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of communication mechanisms, including known, related art, and/or later developed technologies.
[0055] FIG. 2 illustrates a connectivity diagram of the system 100 with the computing device 200, according to an embodiment of the present invention. The computing device 200 may comprise a user interface 202 configured to present detection results generated by the system 100. The system 100 may be communicatively coupled to the computing device 200 through the communication unit 108 to enable transmission of detection results and facilitate real-time monitoring and analysis.
[0056] In an embodiment of the present invention, the computing device 200 may be configured to receive the transmitted detection results and perform further analysis, visualization, alert generation, decision-making, and so forth. The computing device 200 may be, but not limited to, a control station, a monitoring server, a user interface device, computing system, and so forth, capable of processing the received data.
[0057] In an embodiment of the present invention, the computing device 200 may further configured to receive and process the detection results transmitted by the communication unit 108. The computing device 200 may analyse the received data for identifying suspicious objects, prioritizing alerts based on confidence score and uncertainty level, and supporting decision-making operations.
[0058] In an embodiment of the present invention, the user interface 202 may be configured to present detection results received from the system 100. The user interface 202 may display visual data corresponding to the monitored border region along with overlaid bounding boxes indicating detected objects. Each detected object may be associated with object class information, confidence score, and uncertainty level, wherein the uncertainty level may be visually represented using color-coded indicators or graphical markers to assist a user in assessing reliability of the detection results. The user interface 202 may provide real-time alerts and notifications when suspicious objects are detected or when the uncertainty level exceeds a predefined threshold.
[0059] The user interface 202 may further enable user interaction for filtering, sorting, and prioritizing detection results based on parameters including object type, confidence score, uncertainty level, and so forth. The user interface 202 may allow playback of historical surveillance data, zooming into specific regions of interest, and manual verification of detected objects. In an embodiment of the present invention, the user interface 202 may be further configured to provide interpretability of detection results by presenting uncertainty levels and confidence scores in association with detected objects. The user interface 202 may be configured to visually differentiate detection reliability through graphical indicators, heatmaps, or color-coded overlays, thereby enabling a user to assess trustworthiness of each detection result for decision-making.
[0060] In an embodiment of the present invention, the processing unit 104 may be configured to transmit the generated detection results to the computing device 200 for rendering on the user interface 202. The user interface 202 may present visual data with superimposed bounding boxes, object class information, confidence score, and uncertainty level. The user interface 202 may support interactive visualization including zoom control, object highlighting, and prioritization based on detection parameters.
[0061] The user interface 202 may be, but not limited to, a graphical user interface, a dashboard display, visualization platform, and so forth, capable of presenting surveillance data to a user. Embodiments of the present invention are intended to include or otherwise cover any type of user interface 202 capable of presenting detection results and facilitating user interaction, including known, related art, and/or later developed technologies.
[0062] FIG. 3 illustrates components of the processing unit 104 of the system 100, according to an embodiment of the present invention. The processing unit 104 may comprise a data acquisition module 300, a preprocessing module 302, a feature extraction module 304, a data detection module 306, and an output management module 308.
[0063] In an embodiment of the present invention, the data acquisition module 300 may be configured to receive visual data captured by the surveillance camera 102. The visual data may comprise image frames or video streams representing the monitored border region under varying environmental conditions. The data acquisition module 300 may be further configured to transmit the received visual data to the preprocessing module 302 for further processing.
[0064] In an embodiment of the present invention, the preprocessing module 302 may be configured to receive the visual data from the data acquisition module 300. The preprocessing module 302 may be configured to enhance image quality based on environmental conditions including rain, fog, sandstorm, and snowfall. The preprocessing module 302 may be further configured to apply preprocessing techniques such as, but not limited to, noise reduction, dehazing, contrast enhancement, deblurring, and removal of weather-induced artifacts. The preprocessing module 302 may be configured to transmit the pre-processed visual data to the feature extraction module 304.
[0065] In an embodiment of the present invention, the feature extraction module 304 may be configured to receive the pre-processed visual data from the preprocessing module 302 and extract feature representations from the pre-processed visual data. The feature extraction module 304 may be configured to extract spatial features using the convolutional neural network architecture. The feature extraction module 304 may be configured to extract contextual features using the Swin Transformer architecture configured to capture global relationships within a visual scene. The feature extraction module 304 may be configured to transmit the extracted spatial features and contextual features to the data detection module 306.
[0066] In an embodiment of the present invention, the data detection module 306 may be configured to receive the spatial features and the contextual features from the feature extraction module 304. The data detection module 306 may be configured to combine the spatial features and the contextual features through the dynamic weather-gating mechanism that adjusts feature contribution according to detected environmental condition. Further, the data detection module 306 may be configured to perform localization and classification of objects within the visual data using an object detection framework based on the YOLOv8 backbone. The data detection module 306 may be configured to transmit detection data to the output management module 308.
[0067] In an embodiment of the present invention, the output management module 308 may be configured to receive the detection data from the data detection module 306. The output management module 308 may be configured to estimate prediction uncertainty associated with detected objects through an uncertainty estimation process. The output management module 308 may be configured to generate detection results comprising object class information, bounding box coordinates, confidence score, and uncertainty level for identification of suspicious objects in the monitored border region. The output management module 308 may be configured to transmit the generated detection results to the computing device 200 through the communication unit 108.
[0068] FIG. 4 depicts a flowchart of a method for detection of suspicious objects, according to an embodiment of the present invention. At step 402, the system 100 may receive the visual data captured by the surveillance camera 102.
[0069] At step 404, the system 100 may perform weather-aware preprocessing on the visual data to enhance image quality according to environmental conditions including rain, fog, sandstorm, and snowfall.
[0070] At step 406, the system 100 may extract spatial features from the pre-processed visual data through the convolutional neural network architecture.
[0071] At step 408, the system 100 may extract contextual features from the pre-processed visual data through the Swin Transformer architecture configured to capture global relationships within the visual scene.
[0072] At step 410, the system 100 may combine the spatial features and the contextual features through the dynamic weather-gating mechanism that adjusts feature contribution according to the detected environmental condition.
[0073] At step 412, the system 100 may perform localization and classification of objects within the visual data through an object detection framework based on the YOLOv8 backbone using fused features obtained from the dynamic weather-gating mechanism. At step 414, the system 100 may estimate prediction uncertainty associated with detected objects through the uncertainty estimation process.
[0074] At step 416, the system 100 may generate detection results comprising object class information, bounding box coordinates, confidence score, and uncertainty level for identification of suspicious objects in the monitored border region.
[0075] At step 418, the system 100 transmit the generated detection results to the computing device 200 through the communication unit 108 operatively coupled to the processing unit 104 for further analysis and monitoring.
[0076] While the invention has been described in connection with what is presently considered to be the most practical and various embodiments, it is to be understood that the invention is not to be limited to the disclosed embodiments, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the scope of the appended claims.
[0077] This written description uses examples to disclose the invention, including the best mode, and also to enable any person skilled in the art to practice the invention, including making and using any devices or systems and performing any incorporated methods. The patentable scope of the invention is defined in the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements within substantial differences from the literal languages of the claims. , Claims:CLAIMS
I/We Claim:
1. A system (100) for detection of suspicious objects under uncertain climatic conditions in a border surveillance environment, the system (100) comprising:
a surveillance camera (102) configured to capture visual data from a monitored border region;
a processing unit (104) operatively connected to the surveillance camera (102); and
a memory (106) storing instructions that, when executed by the processing unit (104), cause the following operations, characterized in that the processing unit (104) is configured to:
receive the visual data captured by the surveillance camera (102);
perform weather-aware preprocessing on the visual data to enhance image quality according to environmental conditions including rain, fog, sandstorm, and snowfall;
extract spatial features from the pre-processed visual data through a convolutional neural network architecture;
extract contextual features from the pre-processed visual data through a Swin Transformer architecture configured to capture global relationships within a visual scene;
combine the spatial features and the contextual features through a dynamic weather-gating mechanism that adjusts feature contribution according to the detected environmental condition;
perform localization and classification of objects within the visual data through an object detection framework based on a YOLOv8 backbone using fused features obtained from the dynamic weather-gating mechanism;
estimate prediction uncertainty associated with detected objects through an uncertainty estimation process;
generate detection results comprising object class information, bounding box coordinates, confidence score, and uncertainty level for identification of suspicious objects in the monitored border region; and
transmit the generated detection results to a computing device (200) through a communication unit (108) operatively coupled to the processing unit (104) for further analysis and monitoring.
2. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to display the generated detection results onto a user interface (202) installed on the computing device (200).
3. The system (100) as claimed in claim 1, wherein the detected objects are selected from unmanned aerial vehicles, weapons, unauthorized human activity, or a combination thereof present in the monitored border region.
4. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to perform rain removal, dehazing, contrast enhancement, noise suppression, and snow artifact reduction during the weather-aware preprocessing according to the detected climatic condition.
5. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to analyse long-range spatial relationships within the visual scene through the Swin Transformer architecture to support contextual understanding of complex environments.
6. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to determine feature weighting between spatial features and contextual features through the dynamic weather-gating mechanism based on environmental cues present in the visual data.
7. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to perform anchor-free object detection through the YOLOv8 backbone for identification of objects with different spatial scales.
8. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to quantify prediction confidence associated with detected objects under degraded visual conditions through probabilistic inference.
9. The system (100) as claimed in claim 1, wherein the processing unit (104) is configured to initiate a secondary verification process when the uncertainty level associated with detected objects exceeds a predefined threshold.
10. A method (400) for detection of suspicious objects under uncertain climatic conditions in a border surveillance environment, the method (400) is characterized by steps of:
receiving visual data captured by a surveillance camera (102);
performing weather-aware preprocessing on the visual data to enhance image quality according to environmental conditions including rain, fog, sandstorm, and snowfall;
extracting spatial features from the pre-processed visual data through a convolutional neural network architecture;
extracting contextual features from the pre-processed visual data through a Swin Transformer architecture configured to capture global relationships within a visual scene;
combining the spatial features and the contextual features through a dynamic weather-gating mechanism that adjusts feature contribution according to the detected environmental condition;
performing localization and classification of objects within the visual data through an object detection framework based on a YOLOv8 backbone using fused features obtained from the dynamic weather-gating mechanism;
estimating prediction uncertainty associated with detected objects through an uncertainty estimation process;
generating detection results comprising object class information, bounding box coordinates, confidence score, and uncertainty level for identification of suspicious objects in the monitored border region; and
transmitting the generated detection results to a computing device (200) through a communication unit (108) operatively coupled to the processing unit (104) for further analysis and monitoring.
Date: April 08, 2026
Place: Noida

Nainsi Rastogi
Patent Agent (IN/PA-2372)
Agent for the Applicant