Research On Detection Method Of Water Surface Infrared Target And Realization Of Multi-core DSP

As the key technology in the fields of infrared guidance,early warning and search,infrared target detection is always an important issue in the field of infrared image processing.The background of infrared image is always complicated due to the reflection of sunlight,sky and clouds,and the impact of other factors,which bring much interference to the detection of targets on the water surface.In the past,scholars only studied the detection method of infrared targets on the water surface when the distance is close or far,but in practical applications,the distance between the detector and the target is always changing and the method of target detecting should be also changed at the same time.In this paper,methods of detecting the infrared ship-target on the water surface,are studied when the detecting distance is different.Firstly,several algorithms are studied to enhance the original infrared image on the water surface which has the low signal-to-noise ratio and contrast,the enhancement of the homomorphic filtering is better which has the higher image contrast.Secondly according to the characteristics of the ship-target is small and difficult to be detected in complex background when the detecting distance is far,several background suppression methods are used to suppress the background of the image,the TDLMS method is chosen as the optimal background suppression method in this paper.A method of infrared small target detection based on local statistics of image information entropy is proposed,which we can use it to detect the weak ship-target in the infrared image effectively,and aim at the phenomenon that false targets may still exist when the detection is based on single-frame image,the improved pipeline filtering method is used to verify the detection result and track the weak target.Then,aim at the phenomenon that the ship profile is distorted by the noise when the detecting distance is relatively close,the contoured segmental Gaussian smoothing is used to smooth the noisy contours,and the affine moment invariants are used as the description function of the contour feature to classify and identify the ship-target.Finally,based on TI multi-core DSP processing board TMDSEVM6678L,a system of infrared weak target detection and trajectory tracking is designed,and the feasibility of it is verified by experiment.