These are some avenues for Drone Video Sensing Analytics (DVSA): 

1. Palladyne AI: 

Palladyne AI is quietly rewriting the rules of robotic intelligence. Born from decades of robotics innovation and headquartered in Salt Lake City, Utah, the company has emerged as a leader in edge-native autonomy—building software that allows robots to perceive, reason, and act in real time, without relying on cloud connectivity or brittle pre-programmed routines. At the heart of its platform is Palladyne IQ, a cognitive engine that transforms industrial and collaborative robots into adaptive agents capable of navigating uncertainty, learning from their environment, and executing complex tasks with minimal human intervention. 

What sets Palladyne apart is its commitment to closed-loop autonomy. Unlike traditional robotic systems that operate on static instructions or require constant cloud-based updates, Palladyne IQ runs directly on the edge—processing sensor data locally, making decisions on the fly, and adjusting behavior in response to real-world feedback. This architecture mimics the human cognitive cycle: observe, interpret, decide, and act. It enables robots to handle nuanced tasks like sanding aircraft fuselages, inspecting weld seams, or navigating cluttered factory floors—jobs that demand both precision and adaptability. 

The company’s deployments speak volumes. In collaboration with the U.S. Air Force’s Warner Robins Air Logistics Complex, Palladyne-powered robots are used for aircraft sustainment operations, including media blasting and surface preparation. These are high-stakes, labor-intensive tasks where consistency and safety are paramount. By automating them with intelligent edge robotics, the Air Force has reduced downtime, improved throughput, and minimized human exposure to hazardous environments. Similar applications are emerging in advanced manufacturing, logistics, and infrastructure maintenance, where Palladyne AI’s software enables robots to operate autonomously in dynamic, unstructured settings. 

This is where an aerial drone video analytics initiative could become a transformative layer. Palladyne’s robots are already equipped with rich sensor arrays—LiDAR, cameras, force sensors—but the real value lies in interpreting that data in context. A cloud-optional analytics pipeline, built for real-time geospatial reasoning and object detection, could extend Palladyne’s capabilities beyond the factory floor. Let us consider a scenario where a drone captures overhead footage of a construction site, and this system flags structural anomalies, maps terrain changes, or identifies safety violations. That data could then be handed off to a Palladyne-enabled ground robot, which autonomously navigates to the flagged area and performs inspection or remediation—closing the loop between aerial sensing and terrestrial action. 

Expertise in multimodal vector search and transformer-based perception models could also enhance Palladyne’s semantic understanding. By embedding the proposed DVSA analytics into their platform, robots could not only detect objects but understand their relevance to the task at hand. For example, in a warehouse setting, a robot might recognize a misaligned pallet not just as an obstacle, but as a deviation from standard operating procedures—triggering a corrective workflow or alerting a human supervisor. This kind of contextual intelligence is the next frontier in robotics, and our initiative is well-positioned to deliver it. 

Moreover, our focus on low-latency, edge-compatible inference aligns perfectly with Palladyne’s design philosophy. Their clients—ranging from defense contractors to industrial OEMs—demand autonomy that works offline, in real time, and under strict security constraints. Our analytics layer, especially if containerized and optimized for deployment on embedded GPUs or ARM-based edge devices, could be seamlessly integrated into Palladyne’s runtime environment. Together, we could offer a unified autonomy stack: one that spans air and ground, perception and action, cloud, and edge. 

Palladyne AI is building a nervous system for the next generation of intelligent machines. Our initiative could serve as its perceptual cortex—infusing those machines with the ability to see, interpret, and adapt with unprecedented clarity. It’s a partnership that doesn’t just add value—it completes the vision. 

2. Draganfly: 

Draganfly, a veteran in the drone industry with over 25 years of innovation, has consistently pushed the boundaries of unmanned aerial systems across defense, public safety, agriculture, and industrial sectors. Headquartered in Saskatoon, Canada, the company has earned a reputation for pairing robust hardware with intelligent software, delivering mission-ready solutions that span from life-saving emergency response to battlefield agility. Its recent pivot toward FPV (first-person view) drone systems marks a strategic evolution—one that aligns perfectly with the growing demand for decentralized, high-performance aerial platforms capable of rapid deployment and real-time decision-making. 

In 2025, Draganfly secured a landmark contract with the U.S. Army to supply Flex FPV drone systems and establish embedded manufacturing facilities at overseas military bases. This shift toward in-theater production reflects a broader transformation in drone warfare and logistics: FPV drones are no longer niche tools but frontline assets, valued for their maneuverability, cost-efficiency, and adaptability. By enabling soldiers to build, train, and deploy drones on-site, Draganfly is helping the military achieve operational agility and reduce supply chain vulnerabilities. The company’s embedded manufacturing model also supports rapid iteration, allowing drone designs to evolve in response to real-time battlefield feedback. 

This is precisely where our aerial drone video analytics initiative could become a force multiplier. Draganfly’s FPV platforms, while agile and expendable, generate vast amounts of visual data—footage that, if intelligently processed, could unlock new layers of tactical insight and operational efficiency. Our cloud-based analytics pipeline, designed for real-time geospatial interpretation and object detection, could transform raw FPV footage into actionable intelligence. Whether it’s identifying vehicle-sized targets, mapping terrain anomalies, or detecting patterns in troop movement, our system could elevate Draganfly’s drones from mere reconnaissance tools to autonomous decision-makers. 

Expertise in multimodal vector search and transformer-based object detection could enable semantic indexing of drone footage, allowing operators to query past missions with natural language or visual prompts. This capability would be invaluable in defense scenarios where rapid retrieval of mission-critical data can shape outcomes. For Draganfly’s clients in public safety, insurance, and infrastructure, our analytics could streamline post-disaster assessments, automate damage classification, and support predictive maintenance—all while operating at the edge, without reliance on cloud connectivity. 

Draganfly’s commitment to NDAA-compliant supply chains and secure logistics also aligns well with our architecture’s emphasis on privacy-preserving inference and decentralized control. By integrating our analytics layer into their FPV ecosystem, Draganfly could offer a vertically integrated solution: drones that not only fly and film but also think, interpret, and respond. This would position them not just as hardware providers, but as intelligence partners—delivering end-to-end situational awareness from takeoff to insight. 

In essence, our initiative could help Draganfly close the loop between aerial sensing and autonomous action. It’s a convergence of vision and capability that could redefine what FPV drones are capable of—not just in combat zones, but across industries where speed, precision, and adaptability are paramount. 

#codingexercise: CodingExercise-11-04-2025.docx

 Transient and Transit objects in aerial drone scene sequences 

Time, Location and frequency are the dimensions we would like to ideally capture for each object we detect in aerial drone image scene and their sequences but objects don’t have a generalized signature and often require training and deep learning supervision to detect them, especially for transient and transit objects such as pedestrians and vehicles.  Given several pedestrians and vehicles in scene sequences, we apply Density Based Clustering of Applications with Noise to work robustly on clusters with different shapes, without requiring the number of clusters and especially easy to filter out noise.  

Each image clip sequence is preprocessed to construct a high-quality, neural-network-friendly representation. For each frame, we extract three features: normalized spectral signatures, estimated uncertainty (e.g., motion blur or sensor noise), and timestamp. Spectral signature values are converted from log scale to linear flux using calibration constants derived from the UAV sensor specifications. We then subtract the median and standardize using the interquartile range (IQR), followed by compression into the [-1, 1] range using the arcsinh function. 

Time values are normalized to [0, 1] based on the total observation window, typically spanning 10–30 seconds. Uncertainty values are similarly rescaled and compressed to match the flux scale. The final input tensor for each sequence is a matrix of shape T × 3, where T is the number of frames, and each row contains spectral signature, uncertainty, and timestamp. 

This representation ensures any model can handle sequences of varying length and sampling rate, a critical requirement for aerial deployments where cadence may fluctuate due to flight path, altitude, or environmental conditions. 

The model that we use to leverage frequency domain has three core components: 
Wavelet Decomposition: A one-dimensional discrete wavelet transform (DWT) is applied to the spectral signature vector to suppress noise and highlight localized changes. This is particularly effective in identifying transient objects that appear briefly and then vanish. 

Finite-Embedding Fourier Transform (FEFT): A modified discrete Fourier transform is applied to the time series to extract periodic and harmonic features. FEFT enables detection of transit-like behavior, such as vehicles passing through occluded regions or pedestrians crossing paths. 

Convolutional Neural Network (CNN): The frequency-domain tensor is passed through a series of convolutional and fully connected layers, which learn to discriminate between the four object states. The model is trained using a categorical cross-entropy loss function and optimized with Adam. 

Each entry of the output vector v is called a logit and represents the predicted likelihood that the star is of each class (we use v0 = null, v1 = transient, v2 = pulsator, v3 = transit). We compare the output vectors to one-hot target vectors t with the formula: 

 

And 

 

 

We apply the DFT on an N-long signal x(n) as follows: 

 

To extract features from variable length sequences, we introduce a vector u from writing the result of the DFT as a product of u ⮾ v with u being a parameter to the model and its dimension as a hyperparameter named samples. When 

uk$$u_k$$

 ranges from 0 to N-1, DFT samples those N frequencies in the data. We do not modify vector v but construct is as multiples of a factor

 up to the length of the input vector. 

Then we initialize the finite-embedding Fourier Transform as 

The outer product and element wise exponentiation of the matrix is fast operations. 

The DWT comprises of two operations: a high pass filtering and a low pass filtering. With wavelet decomposition applied to the result, we get two wavelets: 1 representing downscaled and smoothed version of a function and the other as 2. the variations and they are taken as bi-orthogonal. 

Samples from the DOTA dataset are taken to ensure generalization across diverse environments. The model is trained on a four-class scheme: null (no object), transient (brief appearance), stable (persistent presence), and transit (periodic occlusion or movement). 
#codingexercise: CodingExercise-11-04-2025.docx