基于稀疏特征的小视场高速检测算法应用研究

Aiming at the problem that the features of small field of view detection in the high-speed pipeline are difficult to mine and the millisecond cycle is fast, this paper takes the cigarette filter bead streamline detection as an example, and proposes a small field of view high-speed detection algorithm based on sparse features. Firstly, by adjusting the light source, a 'light spot' feature with strong robustness is designed. Secondly, the sparse representation and dictionary learning are used to obtain the projection histogram features of the light spot. To overcome the interference of unstructured backgrounds, the algorithm is combined with Markov-Bayesian reasoning to reduce the spot detection rate, and finally realize the high-speed accurate recognition of the bead in low contrast. The view high speed detection algorithm of small field based on sparse features was verified on the simulation and experimental platform. The conclusions show that the extracted light spot features can overcome the interference of the color, size and low contrast of the bead, and maintain the stability of the feature. The fused Markov-Bayesian sparse representation algorithm can improve the recognition accuracy of the spot. The method can achieve 3 000 tests per minute, and the detection accuracy can reach 99.5%.