Multi-spectral Target Detection Algorithm And System Implementation Research

With the wide application of spectral detection technology,a variety of spectral target detection algorithms have emerged.The traditional detection algorithms,which are much complex and time-consuming,greatly reduce the efficiency.This paper applies the visual selective attention mechanism to multispectral target detection,which achieves the outstanding results,and the algorithm is simple,and easy to implement on the hardware,so it can highly improve the detection efficiency.This paper analyzes the three classical frequency-domain models,and summarize saliency features on frequency domain.In addition,it is demonstrated that the saliency detection based on frequency is not only a pure mathematical calculation,but consistent with the selective attention mechanism.By using multi-scale analysis and morphological processing,the PFT model is optimized,which improves the detection accuracy and integrality.Based on PFT model and quaternion transformation,an efficient spectral target detection algorithm is proposed.The original spectral images in the two special spectral bands are represented by using quaternion,then the two saliency maps are fused by PCA.The paper achieves excellent effect,and is robust to noise.The saliency model based on frequency domain can only highlight the global saliency,combining the local space-domain model,and the hybrid detection algorithm is proposed which adopts the two domains and two complementary features.A large number of experiments demonstrate that the effectiveness and robustness of our proposed algorithm in both colorful images and spectral images.The above algorithm analysis and experimental results manifest that the spectral target detection based on saliency can achieve balance between effect and complexity.This paper builds the multispectral imaging system by using AOTF and near-infrared camera,and designs the spectral target detection system by integrating the FPGA processor.In view of the FPGA's powerful parallel processing and poor signal operation,the detection algorithm based on Hadamard transform,is proposed,designs the corresponding hardware modules,and achieves real-time and efficient effect.