Speakers

Landing on any celestial body with stringent engineering and mission constraints poses an inordinate challenge on any landing mission. Hazard Detection and Avoidance (HDA) becomes the crucial technology to be employed in landers for exploring scientifically interesting and landing near hazardous regions.​

Recently landing missions to Mars and Moon fully relied on computer vision techniques for safe and pin-point landing on surface.​

Chandrayaan-3 Vikram Lander landed on the lunar surface with the help of a computer vision-based HDA approach employed onboard for safe landing near the south pole. In this talk, we will discuss, computer vision techniques adopted in landing missions and with a focus on Chandrayaan Missions.

The integration of Artificial Intelligence (AI) and Machine Learning (ML) has revolutionized Earth Observation Analytics, transforming our capacity to process, evaluate, and derive insights from vast datasets generated by Earth-observing satellites and sensors. This presentation explores the historical development of AI and ML in this field, emphasizing key achievements and innovations by SatSure that address large-scale client needs. AI and ML have facilitated automated feature extraction, anomaly detection, and predictive modeling, with applications ranging from environmental monitoring to disaster management. Additionally, this presentation will discuss the challenges and opportunities presented by this evolution and provide insights into the future potential of AI and ML in advancing Earth observation analytics.

AI and ML models have great success in many fields related to obtaining large amounts of image data to aid in pattern recognition and create algorithms through computer systems. AI can improve in the analysis of large areas of interest, to classify objects, detect and monitor land use, data fusion, cloud removal, and spectral analysis of environmental changes from satellite or aerial imagery. AI can aid in data collection, processing, and understanding using neural networks and deep learning through Computer Vision models to allow data users to better understand and handle data more efficiently in a timely manner, at multiple spatial resolutions. ​

However, despite the numerous opportunities presented by AI and ML in Earth observation, there are also challenges that need to be addressed like issues related to data quality, scalability, interpretability of AI models, and ethical considerations surrounding the use of AI in decision-making processes.​

This talk holistically presents the existing satellites in orbit, future satellite missions, current technology, AI&ML use cases on various geospatial applications, challenges, and opportunities and overall big picture about Earth AI.

The talk will explore the advanced vision systems developed for the ISRO Inertial Systems Unit (IISU) and their crucial roles in various robotic applications. These systems are utilized in humanoid robots, flying robots, robotic manipulators, and horizontal and vertical landing missions, enabling autonomous navigation, precise manipulation, and safe landings. The challenges addressed by these systems, including varying lighting conditions, surface texture recognition, and real-time obstacle avoidance, will be discussed in detail. Innovative solutions and technologies, such as image processing, machine learning, and sensor fusion, will be presented to demonstrate how these challenges were overcome. The transformative impact of vision-based robotics in the space exploration and aerospace sectors will be emphasized, providing a comprehensive understanding of their current state and future prospects. 

Wire crimp visual inspection for quality is a very critical operation in the realization of aerospace systems. Being a process done in very large numbers on daily basis in aerospace manufacturing, the detection of the defects in wire crimp by manual visual inspection can be a daunting task. This talk proposes the use of a computer vision based solution for the quality inspection of each wire crimp joint. 

As part of subsystem level and vehicle level testing of a launch vehicle, each actuator is commanded with positive and negative command and the corresponding movements of actuators are verified against the expected direction of movement. Currently this verification is done manually in the presence of a group of people who witness the test. This method needs a substantial amount of man hours and is prone to errors. This talk proposes the use of a computer vision based solution for automated movement verification of actuators and nozzles in hardware in loop (HIL) simulation of a launch vehicle.​

A raspberry pi based computer with a camera captures the images. Object tracking algorithms are used to identify and find the position of an AruCo marker placed on the object of interest. The velocity of the movement of the marker is calculated and the information is displayed on the video for further processing or broadcast. This setup can be extended to track the exact motion of the body by calibrating the camera as well.