Bionic Swarm Intelligence Optimization Algorithm And Its Application In Point Clouds Registration

Bionic swarm intelligence optimization algorithm is a goal optimization strategy to simulate biological behavior in nature,and has certain application in engineering optimization problems.Researching more efficient intelligent optimization strategy of bionic swarm optimization and applying it to solve complex three-dimensional point cloud registration problem has an ideal development prospect.This paper focuses on the improved cuckoo search algorithm and artificial bee colony algorithm.Using pattern search chemotaxis,global reconnaissance strategy and second-order oscillation mechanism,a new improved swarm intelligence optimization algorithm is proposed to improve the performance of the algorithm,and is applied to solve the point cloud registration optimization problem.Bionic swarm intelligence optimization algorithm is a goal optimization strategy to simulate biological behavior in natural world,and has certain application in engineering optimization problems.Research on more efficient intelligent optimization strategy of bionic swarm optimization and on applying it to solve complex three-dimensional point clouds registration problem has an ideal developing prospect.This paper focuses on the improved cuckoo search algorithm and artificial bee colony algorithm,and on using pattern search,global reconnaissance strategy and second-order oscillation mechanism,a new improved swarm intelligence optimization algorithm is proposed to improve the performance of the algorithm,and is applied to solve the point clouds registration optimization problem.The main innovative contributions of this paper include four aspects:1)A cuckoo search algorithm based on pattern search is proposed.Cuckoo search algorithm is a new intelligent optimization algorithm based on Lévy flight search strategy.The single Lévy Flight Random Search Renewal Strategy,however,has some shortcomings,such as limited local mining capacity and low optimization accuracy.To solve this problem,an improved cuckoo global optimization algorithm is proposed.The main features of the improved algorithm are as follows: firstly,the strategy of pattern search alternately carried out by global detection and pattern movement is used to realize the organic combination of global detection of Cuckoo Lévy flight and local optimization of pattern search,so as to avoid blind search and enhance the local mining ability of the algorithm;secondly,the adaptive competition mechanism is adopted to dynamically select the optimal number of solutions,which achieves an effective balance between the search speed and the diversity of the solution in the iterative process.Finally,the dominant set search mechanism is used to realize the effective cooperation and sharing of the optimal solution,and strengthen the study of the dominant experience.The algorithm is applied to the optimization of numerical functions.The results show that the algorithm not only substantially improves the accuracy and efficiency of optimization,but also has strong robustness,and is suitable for multi-modal and complex high-dimensional space global optimization problems.Compared with the typical improved cuckoo optimization algorithm and other swarm intelligence optimization strategies,the local mining performance and optimization accuracy of the improved algorithm has more advantages and better results.Point clouds registration is a key problem in three-dimensional digital processing technology.Traditional point clouds registration methods are sensitive to initial registration position and easy to fall into local optimum.This kind of problem can be effectively solved by using bionic swarm intelligence optimization algorithm.Cuckoo search based on pattern search is used to solve the optimization problem of point clouds registration.In the whole registration process,point clouds simplification and feature extraction are used first.Then,the objective function is optimized by the improved cuckoo search global optimization method,and the global optimal parameters of point clouds transformation matrix are obtained.The final point clouds registration effect can be obtained by precision registration.The performance of the algorithm is tested by different model data.The results show that the point clouds registration based on improved cuckoo global optimization algorithm proposed for the first time can solve the problem that the traditional Iterate Closed Point(ICP)registration algorithm relies heavily on the initial position of the point clouds and restrain premature.Compared with the traditional ICP registration algorithm,the global optimization ability is improved,and the solution accuracy is also greatly improved.It has good robustness and application value in point clouds registration.2)An artificial bee colony algorithm based on global reconnaissance search is proposed.Artificial bee colony algorithm is a swarm intelligence optimization algorithm proposed in recent years to simulate the foraging behavior of bees.Due to the insufficiency of the escaping behavior of the reconnaissance bee in the algorithm,this algorithm has the problems of insufficient global search performance,premature convergence and easy to fall into local optimum.According to the latest research results on the behavior of reconnaissance bees,the reconnaissance bees have the characteristics of rapid flight,global reconnaissance and guiding other bee colonies for foraging.The algorithm utilizes the feature that the reconnaissance bee first detects the honey source globally and rapidly and then cooperates with other bee colonies in the process of foraging.An improved bee colony optimization algorithm is proposed to simulate the reconnaissance bee's global fast reconnaissance search in natural world.Firstly,the algorithm uses the new reconnaissance search strategy to conduct global rapid reconnaissance in the allocated subspace,which can effectively avoid premature convergence of the algorithm and avoid falling into local optimum.Secondly,the reconnaissance bee colony uses the global reconnaissance heuristic information to guide other bee colony foraging search,and then the interaction between the two parties achieves the optimization performance of the algorithm and improves the accuracy of the solution.Finally,the prediction and selection mechanism is introduced by the algorithm to improve the search strategy of the leading and follower bees,which further enhances the performance of local search in the neighborhood of the algorithm.The algorithm is applied to numerical function optimization.The results show that compared with the classical improved artificial bee colony algorithm and other improved swarm intelligence optimization algorithms,the global search performance of the algorithm is enhanced,the premature convergence is avoided effectively,the optimization accuracy is improved significantly,and it can be applied to the optimization problems in high-dimensional space.3)An artificial bee colony algorithm based on second-order oscillation perturbation is proposed.Artificial bee colony algorithm is a search strategy based on the role assignment of bee colony and the mechanism of cooperative work.However,in the latter stage of the search,local mining is gradually exhausted,and the ability of global reconnaissance and escape is insufficient.At the later part of the search,the algorithm has insufficient population diversity and premature convergence,which often shows strong search ability and weak mining ability.Its essence is the imbalance between global exploration and local mining ability.In order to solve this problem,combined with the advantage that artificial bee colony algorithm is easy to mix with other technologies,the second-order oscillation perturbation strategy is introduced in the process of hiring bee colony foraging,and a second-order oscillation mechanism artificial bee colony algorithm based on asynchronous change learning is proposed.Firstly,by introducing the second-order oscillatory search mechanism,prematurity can be effectively suppressed and local search ability can be improved.Secondly,at the beginning of iteration perturbation strategy is used in the search process to enhance global detection and increase the diversity of spatial search.Finally,through asynchronous change learning mechanism,the algorithm enhances the solution accuracy by intensifying local mining performance in the later part of the search process.The algorithm is applied to the optimization of numerical functions.The experimental results of typical test functions show that the algorithm can effectively achieve the balance between global exploration and local mining ability of the artificial bee colony algorithm,overcome the shortage of search performance,increase the diversity of search,and significantly improve the search rate.Compared with other typical strategies,the algorithm has a strong competitive advantage.The second-order oscillating artificial bee colony algorithm based on asynchronous change learning is applied to the registration of three-dimensional point clouds.A point clouds registration method based on improved artificial bee colony algorithm is proposed.Through uniform sampling of input point clouds and extraction of intrinsic shape feature points based on domain radius constraints,point clouds is further simplified.Then better initial registration of point clouds is completed by improved artificial bee colony algorithm,and the parameters of spatial transformation matrix are obtained.Finally,the k-d tree nearest neighbor search method is used to speed up the corresponding point search,so as to improve the efficiency of the ICP registration algorithm.Through registration experiments of point clouds models and scene data at different initial positions,testing results show that this algorithm has better anti-noise,higher registration accuracy and stronger robustness compared with traditional registration methods.