Design And Implementation Of Laser Target Recognition Based On DSP And Binocular Camera

Laser target detection, based on the spot recognition technology, by converting the position of the target into the position of the spot on the imaging device, thus achieving the determination of the target position information extracted by the image signal processing.With a high detection accuracy and fast speed, it is widely used in aerospace, military reconnaissance,medical equipment, security monitoring, industrial measurement and other fields.In this study, based on analysis of domestic and international design of laser spot position system, make research on the key problems of system, such as the optical imaging system, parallel image processing unit and image recognition algorithm. Making system solutions of Laser spot identification system design based on binocular camera and DSP, with high accuracy and high real-time. System apply the principle of scanning imaging, with binocular TDI-CCD with high speed stage, and the image processing unit based on DSP array real-time extract laser light spot target with strong reflection properties.First of all, according to characteristics of spot image, design a kind of light detection arithmetic, using gray information, shape information and frequency information as the characteristics of target detection,with strong real-time performance and high recognition efficiency.To solve complex threshold segmentation problems under the condition of atmospheric turbulence and the larger ratio of background and the target, and problem that spot center location is not accurate caused by the phenomenon of "supersaturated".In order to further improve the recognition efficiency, reduce the false targets, to produce false spot under complicated environment object abstraction and classification, puts forward a method to deal with multiple false target selection.Second, complete the system implementation. Gives design of image acquisition module, image processing module and data output module, etc. Spot recognition algorithm is transplanted to DSP array based on the image processing unit, and make parallel algorithm design according to the characteristics of the system hardware and algorithm.In the end, in view of the images parallax caused by binocular structure, estimates the scope of binocular parallax size within different detecting distances, puts forward the dense disparity map based on the binocular image registration algorithm for pixel level correction methods, and on the base of classic zhangZY binocular matching algorithm, proposed binocular image registration method based on a coarse-pure strategy.After the systems integration test, proved that its reliability and real-time performance are conform to the requirements of the performance and this design has high practical value.