Vision Detection-Based Path Planning For Plug Tray Seedling Transplanting

Seedling transplanting has many advantages and has replaced the traditional direct seeding method.The development of specialised transplanting equipment to replace traditional manual work is of great significance in improving operational efficiency,reducing labour intensity and alleviating labour shortages.Machine vision and path planning are key technologies for automated transplanting equipment,and their performance will directly determine the efficiency of transplanting operations.Based on this,this study investigates the growth quality inspection of tomato seedlings in greenhouse and the path planning of replanting based on the quality inspection.The details of the research are as follows:(1)An image processing algorithm system based on Python was developed to detect the growth quality of seedlings in different ages of plug trays.Extraction of seedling information from plug trays by the Cive green factor extraction function to obtain seedling information without excess noise.The Hough Lines linear fitting function was used to extract the outline of the seedlings,which greatly improved the accuracy of the detection by avoiding any possible misjudgement due to the cluttered pixel information.The results show that the system is able to meet the practical requirements for the quality inspection of greenhouse tomato plug tray seedling with an average accuracy of 100% at 12 day,with an average accuracy of 98.2% at 22 day.The average time required to complete the inspection and labelling of a 12 day standard 50-hole tomato seedling is 1.18 s.The results show that the developed visual inspection system is able to meet the practical requirements for the quality inspection of seedling trays,and can provide accurate information on the location of holes for replanting.(2)On the basis of quality detection,for the specifinode of the case of replanting seedlings of the same size of hole trays,this thesis tries to use the ant colony algorithm to train and guide the heuristic function of the A* algorithm,so that it has the ability to search in both directions.The simulation model of the path planning algorithm is based on the numerical matrix of the growth quality information of the plug tray seedlings outputted from the visual inspection.The results show that,in terms of path length and operation time,the path planning length of the Imp-A* algorithm model is optimized compared to the common sequence method one(CSM1),the Dijkstra algorithm model(DA),the ant colony algorithm model(ACA)and the A* algorithm model(A*)at 50 holes by up to 16.4%,6.3%,11.7% and 3.1% respectively.In the 50-hole transplanting validation test,the time required for the Imp-A* model was 79.9%,71.5% and 20.5% higher than that of the Dijkstra,ACA and A* models respectively.In the 50-hole transplanting validation test,the total transplanting time of the Imp-A* algorithm model was reduced by 2.42%,12.84%,11.63% and 14.27%compared to the A* algorithm model,the ACA model,the DA model and the CSM1,respectively.Both simulation and validation tests showed that the Imp-A* algorithm model is stable,reliable and efficient,and is expected to be used as a new transplanting path planning scheme instead of the traditional one.