Pakistan Army | News and Discussions
- Thread starter ghazi52
- Start date
ghazi52
THINK TANK: CONSULTANT
- Thread starter
- #408
DG ISPR Lt Gen Ahmed Sharif Chaudhry's Exclusive Interview on Al Jazeera | ISPR
ISPR Official
July 09, 2025.
Be dazzled
THINK TANK: CONSULTANT
B6+ against rpg result is obvious.
Toxic
Trusted Member
Yeah no shit, a MRAP wouldn't have faired much better in the event of a direct RPG impact.
Tariq Habib Afridi
Registered Member
yesterday, i reviewed a research paper on Radio Frequency (RF) based 3-D human action recognition system. I like the idea and i think a similar kind of system can be prepared for Pakistan Army having 2 to 10km range radar which can see through walls and other hard materials and an integrated consumer based GPU system can run a small Machine Learning model identifying human and their actions for counter terrorism option. this device will definitely decrease army casualties as they can before time see any attack someone wants to do on them. I think this kind of RF based can be made easily with low cost and it will have huge impact on counter terrorism operation like IBO. Also, it can be highly accurate even detect hand movements and human positions.
@Panzerkiel @Oscar maybe this kind of system may already exist or under development but if it is not i think its very potent solution and can be made easily and cheaply. I can develop an AI software for the detection of human activity detections based on RF signals and even on WIFI signals.
This kind of system can also be deployed on any armour vehicle and it will give early warning to the soldier in the armour vehicle that at certain distance some dash o clock there is an armed men or threat etc.
Here’s a step-by-step rough plan to develop and deploy this kind of system for our forces.
The first step is defining the system’s requirements: a 360-degree threat detection system capable of identifying humans, reconstructing their 3D meshes, and classifying threats (e.g., armed individuals) based on hand and leg movements. With a range of 2–10 km, it will integrate seamlessly with vehicles like the Al-Khalid or Talha, providing real-time alerts to crews. The system will use high-resolution mmWave radar to generate 3D meshes, downsample them to focus on critical areas (hands/legs), and employ a lightweight ML model for rapid threat classification, ensuring robust performance in Pakistan’s diverse terrains, from deserts to urban areas.
For hardware, I recommend Texas Instruments AWR2944 mmWave radar modules (~$1,500 each, 4–6 units per vehicle) for their high angular resolution (~1°) and long-range capability (2–10 km with high-gain antennas). These will be paired with a NVIDIA Jetson AGX Orin ($2,000) for real-time ML and signal processing, powered by the vehicle’s 12V/24V system. High-gain phased-array antennas will ensure extended range, and a ruggedized display will integrate with the Al-Khalid or VT-4’s command interface for crew alerts, making the system cost-effective (~$14,000 per vehicle) and adaptable to Pakistan’s existing or procurable platforms.
Software development involves processing radar data with TI’s mmWave SDK to generate range-Doppler maps and point clouds, followed by 3D mesh creation using Python’s SciPy (Delaunay triangulation, ~20–50 cm resolution at 2 km). We’ll segment meshes to isolate hands and legs using ML (e.g., PointNet++), downsample them with Open3D’s voxel grid filtering to reduce data by ~50%, and deploy a lightweight MobileNetV3 CNN (~5 MB) trained on RF signatures for >90% accuracy in detecting armed individuals. The software stack (Python, C++, TensorFlow Lite, ROS2) will run on Ubuntu 20.04, ensuring real-time operation (<100 ms latency) on vehicles like the Talha APC.
Integration involves mounting radar units on the vehicle’s turret or hull for 360-degree coverage, connecting to the Jetson via Ethernet, and interfacing with the Al-Khalid or VT-4’s CAN bus for alerts. Calibration will account for environmental noise (e.g., dust, metal), and a Qt-based GUI will display 3D meshes and threat locations (e.g., “Armed individual, 3 km, 2 o’clock”). This setup ensures compatibility with Pakistan’s armored fleet, requiring minimal modifications.
Training and testing will involve collecting or simulating RF datasets with human activities (walking, aiming) at 2–10 km, training the ML model on Jetson (4–6 weeks), and conducting field tests in Pakistan’s urban and desert environments. The system will be validated for <100 ms latency and >85% accuracy at 2–5 km, >75% at 5–10 km, ensuring reliability against hidden threats (e.g., behind foliage). Crew training (1–2 days) will cover interpreting alerts and visuals.
Deployment will involve retrofitting the system on Al-Khalid, Talha, or VT-4 vehicles, with regular calibration and software updates. The 7-month development timeline includes prototyping (2 months), software development (3 months), and testing/integration (2 months). This system will enhance Pakistan’s armored forces with all-weather, through-obstacle threat detection, making our vehicles like the Al-Khalid world-class in situational awareness.
@Panzerkiel @Oscar maybe this kind of system may already exist or under development but if it is not i think its very potent solution and can be made easily and cheaply. I can develop an AI software for the detection of human activity detections based on RF signals and even on WIFI signals.
This kind of system can also be deployed on any armour vehicle and it will give early warning to the soldier in the armour vehicle that at certain distance some dash o clock there is an armed men or threat etc.
Here’s a step-by-step rough plan to develop and deploy this kind of system for our forces.
The first step is defining the system’s requirements: a 360-degree threat detection system capable of identifying humans, reconstructing their 3D meshes, and classifying threats (e.g., armed individuals) based on hand and leg movements. With a range of 2–10 km, it will integrate seamlessly with vehicles like the Al-Khalid or Talha, providing real-time alerts to crews. The system will use high-resolution mmWave radar to generate 3D meshes, downsample them to focus on critical areas (hands/legs), and employ a lightweight ML model for rapid threat classification, ensuring robust performance in Pakistan’s diverse terrains, from deserts to urban areas.
For hardware, I recommend Texas Instruments AWR2944 mmWave radar modules (~$1,500 each, 4–6 units per vehicle) for their high angular resolution (~1°) and long-range capability (2–10 km with high-gain antennas). These will be paired with a NVIDIA Jetson AGX Orin ($2,000) for real-time ML and signal processing, powered by the vehicle’s 12V/24V system. High-gain phased-array antennas will ensure extended range, and a ruggedized display will integrate with the Al-Khalid or VT-4’s command interface for crew alerts, making the system cost-effective (~$14,000 per vehicle) and adaptable to Pakistan’s existing or procurable platforms.
Software development involves processing radar data with TI’s mmWave SDK to generate range-Doppler maps and point clouds, followed by 3D mesh creation using Python’s SciPy (Delaunay triangulation, ~20–50 cm resolution at 2 km). We’ll segment meshes to isolate hands and legs using ML (e.g., PointNet++), downsample them with Open3D’s voxel grid filtering to reduce data by ~50%, and deploy a lightweight MobileNetV3 CNN (~5 MB) trained on RF signatures for >90% accuracy in detecting armed individuals. The software stack (Python, C++, TensorFlow Lite, ROS2) will run on Ubuntu 20.04, ensuring real-time operation (<100 ms latency) on vehicles like the Talha APC.
Integration involves mounting radar units on the vehicle’s turret or hull for 360-degree coverage, connecting to the Jetson via Ethernet, and interfacing with the Al-Khalid or VT-4’s CAN bus for alerts. Calibration will account for environmental noise (e.g., dust, metal), and a Qt-based GUI will display 3D meshes and threat locations (e.g., “Armed individual, 3 km, 2 o’clock”). This setup ensures compatibility with Pakistan’s armored fleet, requiring minimal modifications.
Training and testing will involve collecting or simulating RF datasets with human activities (walking, aiming) at 2–10 km, training the ML model on Jetson (4–6 weeks), and conducting field tests in Pakistan’s urban and desert environments. The system will be validated for <100 ms latency and >85% accuracy at 2–5 km, >75% at 5–10 km, ensuring reliability against hidden threats (e.g., behind foliage). Crew training (1–2 days) will cover interpreting alerts and visuals.
Deployment will involve retrofitting the system on Al-Khalid, Talha, or VT-4 vehicles, with regular calibration and software updates. The 7-month development timeline includes prototyping (2 months), software development (3 months), and testing/integration (2 months). This system will enhance Pakistan’s armored forces with all-weather, through-obstacle threat detection, making our vehicles like the Al-Khalid world-class in situational awareness.
Existing Systems and Research
- DARPA’s RF-based Sensing: Programs like DARPA’s Sense through the Wall (STTW) use RF to detect humans and activities, adaptable for vehicles.
- MIT’s RF-Pose3D: Reconstructs 3D human poses using WiFi, achieving ~20 cm accuracy in controlled settings.
Last edited:
Ali_Baba
VIP Member
This is what Pakistan requires after the recent conflict, where local base defence proved to be a problem for. There are other CIWS based solutions, so hopefully they settle on a solution soon.
Users who are viewing this thread
Pakistan Defence Latest
Country Watch Latest
-
-
-
TF-X / KAAN / Hürjet Turkish Fighter & Trainer Aircrafts News & Discussions (9 Viewers)
- Latest: royalharris
-
-
Latest Posts
-
-
-
Off topic - Query to answer by Chinese friends regarding Three body Problem season 2 (5 Viewers)
- Latest: Salza
-
-
UK Racism against Muslims and Brown people: Why are blacks in the UK given a free pass by white english racists? (18 Viewers)- Latest: Master Chief