Simulation Study Of Edge Detectors To Coherent Ladar Intensity Images And The Performance Evaluation

The development of coherent ladar is rapid and long-range imaging. The coherent ladar can obtain the intensity image and range image at the same time and has the advantages of high-resolution imagery, anti-jamming effectiveness, which can be applied in many fields. Edge detection plays an important role in ladar image processing which provides the data to target segmentation, shape abstraction, target recognition and localization. It is the basis of subsequent steps and its capability has direct influence on precision and performance of ladar imaging system. Therefore it is necessary to study the edge detectors which are suitable to detector ladar images and evaluate the performance of a detector.In this paper, the morphological filter based on ROA detector is introduced, which combines the ROA (ratio of averages) detector and morphological filter based on parametric edge detector. Edges of multi-pixel width are got in ROA detector's results, though this algorithm is simple. Morphological filter based on parametric edge detector has been reported to be successfully applied in actual coherent ladar intensity images corrupted by speckle. One pixel wide edge can be obtained, but the calculation is relatively complicated and time-consuming. The morphological filter based on ROA detector has a similar results with the morphological filter based on parametric edge detector, and significantly reduces the image processing time. Specific confidence levels of the morphological filter based on ROA detector and morphological filter based on parametric edge detector are caculated to evaluate these two edge detectors'performance. And the confidence levels are caculated when morphological filter based on parametric edge detector is utilized to the rectangles of different width, so that whether the edge detector is suitable to detect step edges and roof edges can be judged.In order to describe the performance of several edge detector objectively, coefficient of performance based on Abdou figure of merit is defined, which calculates the sum of all actual detected edges'contribution to true edges localization. The CNR of targets in speckle degraded coherent ladar simulated intensity images is from 14dB to 4.23dB. And there are single objective, multiple objectives with simple background and complicated background, which simulate the ideal conditions and full-scale condition, respectively. So that edge detectors'performances are calculated in different situation. ROA detector, ROA-MSP detector, morphological filter based on ROA detector, parametric edge detector and morphological filter based on parametric edge detector, which are some special detectors based on ladar images, are applied to detect ladar simulated intensity images. At the same time, three conventional edge detectors Sobel detector, Log detector and Canny detector are compared with the detectors mentioned above. The coefficient of performance of different detectors are obtained and compared. Finally, the advantages and disadvantages of specific confidence level and the coefficient are presented.