Research On Straw Mulch Image Segmentation Based On Gray Wolf Optimizer Algorithm

Straw is one of the important biomass in the process of agricultural production.Straw returned not only can reduce the environmental pollution and the waste of natural resources brought by the burning of straw,but also they can leave enough water for the local soil,improve the structure of the soil,increase the organic matter in the soil,promote the healthy growth of crops,and play an important role in environmental protection and the sustainable development of modern agriculture.Nutrient release is slow after straw returning to the field.In order not to seriously affect the absorption of nutrients in the next crop,the time of straw returning to the field is very limited.However,the majority of farmers lack scientific knowledge,straw returning to the field is still difficult to popularize.Based on this situation,the detection of straw returning to the field and supervision work will follow.Because the point is wide,the lack of regulatory personnel,regulatory incompetence.Straw coverage is an important index to measure conservation tillage techniques.The research on rapid detection method of straw coverage rate can improve the efficiency and accuracy of straw coverage rate detection,which is of great significance to straw mulch test.At this stage,the popularity of unmanned aerial vehicle(UAV)applications,make photographs of vegetation facilitate testing,compared with the artificial ground test,aerial drones have a shooting range,high flexibility,shooting time is short,high resolution image data,and the advantages of easy to store data,but the images of straw,weeds and other land close to color,and image in color too influenced by the external environment factors,such as,directly affect the accuracy of the segmentation of straw,and aerial image size is larger,the processing speed is severely affected.Efficient segmentation of land,straw and weeds in field aerial straw images is the key link of straw coverage detection.Therefore,it is imperative to find a fast and accurate segmentation method of multi-threshold images that synchronizes straw and land.To solve the problem of poor stability and easy to get into local optimal solutions of GWO algorithm,an improved grey wolf optimizer(GWO)algorithm based on differential evolution(DE)algorithm and OTSU algorithm is proposed.Multi-threshold OTSU,Tsallis entropy and DE algorithms are combined with GWO algorithm.Multi-threshold OTSU algorithm is used to calculate the fitness of the initial population,which made the initial stage basically stable.The GWO algorithm and DE algorithm are combined to carry out iterative update.DE algorithm can solve the local optimal solution of GWO algorithm.Tsallis entropy algorithm is used to update the population,which can improve the flexibility and adaptability of the algorithm.CEC2005 benchmark functions(23 test functions)were used to test the performance of DE-OTSU-GWO algorithm.Compared with the existing particle swarm optimization(PSO)algorithm and GWO algorithm,the experimental result shows that DE-OTSU-GWO algorithm is more stable and accurate in solving functions.In addition,the high quality of DE-OTSU-GWO algorithm is proved by convergence behavior analysis.In the result of the classical agricultural image recognition problem,compared with GWO,PSO,DE-GWO and 2D-OTSU-FA(2D OTSU Firefly Algorithm),DE-OTSU-GWO algorithm has accuracy in straw image recognition and good effect in segmentation evaluation.The OTSU algorithm improves the precision of the whole algorithm while increasing the running time.After adding DE algorithm,the time complexity will increase,but the solving time will shorten.In order to meet the requirement of large-scale processing of straw coverage image acquired by aerial photography and improve the image segmentation quality and speed of the current DE-GWO algorithm.Differential evolution artificial bee colony survey multi-objective grey wolf optimizer(DE-AS-MOGWO)algorithm is upgraded to the multi-objective GWO algorithm from the original GWO algorithm,and integrates a fixed size external archive,which is used to save or retrieve the currently obtained non-dominated Pareto optimal solution.However,the result are still unstable,and it is easy to fall into the local optimal solution in the process of operation,resulting in inaccurate image segmentation.Therefore,on the basis of MOGWO algorithm,the observation strategy of bee observation in bee colony algorithm is added to obtain an efficient global optimization multi-objective random search method,which makes up for the lack of exploration ability of MOGWO algorithm.The DE algorithm is added to solve the problem that the effect is not stable because it is easy to fall into the local optimum.Tsallis entropy is used to evaluate the efficiency of the algorithm,which can improve the flexibility and adaptability of the algorithm.This algorithm not only inherits the automatic segmentation feature of DE-GWO algorithm,but also has the efficient convergence of AS-MOGWO algorithm,which improves the accuracy and processing speed of image segmentation.The analysis result shows that the matching error between DE-AS-MOGWO optimization algorithm proposed in this study and the manual actual measurement method can be controlled within 8% under the condition of no external influence.In terms of algorithm performance,compared with PSO,GWO,DE-GWO and DE-MOGWO,the average matching rate of DE-AS-MOGWO has been improved by4.967%,3.617%,2.188% and 3.404%,respectively.The average error rate has been reduced by0.168%,0.131%,0.089% and 0.116%,respectively.The algorithm time has been reduced by82%,84%,17% and 32%,respectively.The experimental result shows that the multi-threshold multi-target image segmentation method can achieve better segmentation effect in large scale images,and has universal applicability for different straw coverage images,providing efficient algorithm support for large area straw coverage detection and other related image detection.