Deep neural networks, such as Faster R-CNN, have been widely used in object detection. However, deep neural networks usually require a large-scale dataset to achieve desirable performance. For the specific application, UAV detection, training data is extremely limited in practice. Since annotating plenty of UAV images manually can be very resource intensive and time consuming, instead, we use PBRT to render a large number of photorealistic UAV images of high variation within a reasonable time. Using PBRT ensures the realism of rendered images, which means they are indistinguishable from real photographs to some extent. Trained with our rendered images, the Faster R-CNN has an AP of 80.69% on manually annotated UAV images test set, much higher than the one only trained with COCO 2014 dataset and PASCAL VOC 2012 dataset (43.36%). Moreover, our rendered image dataset contains not only bounding boxes of all UAVs, but also locations of some important parts of UAVs and locations of all pixels covered by UAVs, which can be used for more complicated application, such as mask detection or keypoint detection.

No abstract

Dual-band (DB) harmonic imaging is performed by transmitting and receiving at both fundamental band (f0) and second-harmonic band (2f0). In our previous work, particular chirp excitation has been developed to increase the signal- to-noise ratio in DB harmonic imaging. However, spectral overlap between the second-order DB harmonic signals results in range side lobes in the pulse compression. In this study, a novel range side lobe inversion (RSI) method is developed to alleviate the level of range side lobes from spectral overlap. The method is implemented by firing an auxiliary chirp to change the polarity of the range side lobes so that the range side lobes can be suppressed in the combination of the original chirp and the auxiliary chirp. Hydrophone measurements show that the RSI method reduces the range side lobe level (RSLL) and thus increases the quality of pulse compression in DB harmonic imaging. With the signal bandwidth of 60%, the RSLL decreases from -23 dB to -36 dB and the corresponding compression quality improves from 78% to 94%. B-mode images also indicate that the magnitude of range side lobe is suppressed by 7 dB when the RSI method is applied.

No abstract

No abstract