SBIR-STTR Award

Data links are considered to be of major importance for the Unmanned Air Vehicle (UAV) in its role as a reconnaissance/surveillance platform. Secure, reliable and timely data communication will increase the acceptance of UAVs in the military community. The large amount of communication data needed for video images together with jamming-resistant overhead techniques may not permit timely transmission of images. Thus data compression techniques are needed. This proposal intends to investigate a data compression concept named Walsh Neural Network Unmanned Vehicle Image Transmission (WALNUT) based on vector quantization techniques. Vector quantization (VQ) is emerging as a leading and powerful data compression technique because of its ability to utilize peraceptually meaningful nonlinear correlation techniques. Competitive learning based on neural network (NN) technology applied to VQ brings effective adaptation to motion and scene changes and superior performance needed for near real time compression of image sequences obtained from UAVs operating in a military environment. Such a system has the potential of providing: (1) Economical quantization of pixel information, (2) fast compression and expansion techniques, and (3) protection against channel noise and jamming. In addition VQ with NN learning also inherently maps into parallel computing design makeing possible an even greater data communication speedup.

Video imagery sent from unmanned air vehicles (UAV) over their communication link to an operator is a key to the successful use of UAVs. The limitation of the UAV on-board communication link requires computationaly simple algorithms, and high compression rates. On the other hand, high frame rate and good image quality are needed for successful operation performance for detection, recognition, designation and tracking of targets. The goal of the proposed video compression method, called WALNUT, is to develop a variable near real-time video image compression scheme for UAVs, achieve compression performance surpassing commercial technologies such as JPEG and maintain image quality sufficient to avoid operator fatigue and allow detail analysis of individual frames. WALNUT is based on vector quantization (VQ) image compression. The features of WALNUT which make it ideal for UAV communication links are : 1) high compression ration, 2) fast compression/reconstruction, 3) no dependence on interframe correlation, 4) robust under transmission errors and noise, 5) low distortion, 6) improved quality for specific image types, 7) adaptable to line scanning equipment such as ATARS. Objectives of this initiative are to provide near real-time performance, develop a functional prototype with off-the-shelf equipment, provide compression performance evaluation, explore variable (low to high) compression rates, provide an integration plan with UAV comm. link and design a flyable video image compression prototype.