Design And Experiment Of Intra-row Seedling Avoidable Weeding Control Device Based On An Improved U-Net Model For Chinese Cabbage

Chinese Cabbage is the second largest vegetable crop in China,occupying an important position in China’s "vegetable basket" project.Autumn cabbage has recently been a significant source of income for farmers in China throughout the autumn due to the "a ripe surplus,two ripe shortage" of agroecological circumstances in the northern cold areas.After the harvest of fresh corn,autumn cabbage is replanted to gain access to superior economic returns.Currently,chemical weeding is the major method used to control autumn cabbage weeds in northeast China,with manual and mechanical weeding as supplemental measures.However,chemical weed control not only harms the environment and reduces farmland’s biodiversity,but it also leaves pesticide residues on crops from the application of herbicides,which poses a threat to human health.The issue of pesticide residues is particularly serious for vegetable crops.With the improvement of people’s environmental awareness and consumption concept,people have gradually realized the environmental pollution and residue problems caused by chemical pesticides.Growing interest has been given to mechanical weeding as an environmentally friendly,clean,and non-polluting weeding technique that satisfies the needs of sustainable agricultural development.However,due to the difficulty of control and the high level of intelligence required,mechanical weeding in autumn cabbage weed management is limited to the eradication of some inter-row weeds,and intra-row weeding is still in the exploration stage.In order to address the absence of intelligent cabbage intra-row weeding control devices in Northeast China,this research designs a cabbage intra-row weeding control device based on an improved U-Net model.The main research of this paper is as follows:(1)An intra-row seedling avoidable weeding control device based on an improved U-Net model for Chinese cabbage is designed.The device mainly consists of a robot mobile platform,a seedling recognition and positioning system,a seedling avoidable weeding control system,and a weeding actuator.The robot mobile platform is responsible for mounting the seedling avoidable weeding control device.The seedling recognition and positioning system must identify and locate Chinese cabbage and weed in the farmland and transmit the processed image information to the seedling avoidable weeding control system.The seedling avoidable weeding control system will control the tiller in the weeding actuator after receiving the image information,and the tiller will control the opening and closing of the weeding shovel to achieve seedling avoidable weeding.(2)A semantic segmentation model for Chinese cabbage and weed based on multi-scale input and attention mechanism is constructed.A semantic segmentation model based on improved U-Net is proposed in this paper to address the issue of efficient and accurate identification of vegetable crops and weeds.First,the simplified Visual Group Geometry 16(VGG16)network is used as the coding network of the improved model,and then the input images are continuously and naturally downsampled using the average pooling layer to create feature maps of various sizes,and these feature maps are laterally integrated from the network into the coding network of the improved model.Then,cut the number of convolutional layers of the decoding network of the model and introduce the Efficient Channel Attention(ECA)before the feature fusion of the decoding network,so that the feature maps from the jump connection in the encoding network and the up-sampled feature maps in the decoding network pass through the ECA module together before feature fusion.Finally,the study uses the obtained Chinese cabbage and weed images as the dataset to compare the improved model with the original U-Net model and the current commonly used semantic segmentation models PSPNet and Deep Lab V3+.The improved model in this paper can provide strong technical support for the implementation of seedling avoidable weeding action of the weeding device.(3)The seedling avoidable weeding control system and weeding actuator are designed.After recognizing the seedlings,the seedling recognition and positioning system will extract the information from the recognized seedling image,obtain the location information of the seedlings and the distance information between the adjacent seedlings,and transmit it to the seedling avoidable weeding control system.The seedling avoidable weeding control system will judge whether there are weeds between the adjacent crops and whether the distance between the two crops meets the weeding conditions according to the obtained information.The seedling avoidable weeding control system won’t send weeding instructions to the weeding actuator until both requirements have been satisfied.A swing-type weeding actuator is designed by analyzing the seedling avoidable weeding action principles,and the movement trajectory of the weeding shovel and key weeding actuator components are studied in detail,which can serve as a guide for the design of the swing-type weeding actuator.(4)The performance tests of the cabbage intra-row seedling avoidable weeding device are conducted.Soil tank tests are used to confirm the precision of the seedling recognition and positioning system,as well as the viability and stability of its interconnection with other systems.Field testing is used to confirm the effectiveness of the seedling avoidance and weeding device’s weeding capabilities.The machine advance speed,cabbage plant distance,and weeding shovel entry depth are used as test factors in the field test,while the weeding rate and seedling injury rate are used as test indexes.The results of the single-factor test show that the optimal intervals of machine speed,cabbage spacing and weeding shovel depth are 0.3 m/s~0.5 m/s,45 cm~55 cm,and 11 mm~17 mm,respectively.The best intervals for each test component are used as the level values in a three-factor,three-level Box-Behnken test.The test result indicates that the best combination of parameters for the seedling avoidable weeding device is: machine forward speed 0.36 m/s,cabbage plant spacing 51 cm,and weeding shovel entry depth 16 mm.The field verification test shows that the weeding rate is 86.36% and the seedling injury rate is3.72% at this time.An intra-row seedling avoidable weeding control device based on an improved U-Net model for Chinese cabbage in this paper can meet the requirements of cabbage field seedling avoidable weeding operation,and the research results can provide new ideas for the design and development of intelligent weeding machinery for field vegetables.An improved U-Net semantic segmentation model based on multi-scale input and attention mechanism is proposed in this paper to address the issue of efficient and accurate identification of vegetable crops and weeds,and an oscillating mechanical weeding device is designed based on the improved model,which provides a new method and approach to solve the problems of chemical pesticide pollution and residue in the field.