To avoid colliding with trees during its operation,a lawn mower robot must detect the trees.Existing tree detection methods suffer from low detection accuracy(missed detection)and the lack of a lightweight model.In th...To avoid colliding with trees during its operation,a lawn mower robot must detect the trees.Existing tree detection methods suffer from low detection accuracy(missed detection)and the lack of a lightweight model.In this study,a dataset of trees was constructed on the basis of a real lawn environment.According to the theory of channel incremental depthwise convolution and residual suppression,the Embedded-A module is proposed,which expands the depth of the feature map twice to form a residual structure to improve the lightweight degree of the model.According to residual fusion theory,the Embedded-B module is proposed,which improves the accuracy of feature-map downsampling by depthwise convolution and pooling fusion.The Embedded YOLO object detection network is formed by stacking the embedded modules and the fusion of feature maps of different resolutions.Experimental results on the testing set show that the Embedded YOLO tree detection algorithm has 84.17%and 69.91%average precision values respectively for trunk and spherical tree,and 77.04% mean average precision value.The number of convolution parameters is 1.78×10^(6),and the calculation amount is 3.85 billion float operations per second.The size of weight file is 7.11MB,and the detection speed can reach 179 frame/s.This study provides a theoretical basis for the lightweight application of the object detection algorithm based on deep learning for lawn mower robots.展开更多
Cadra(Ephestia)cautella(Walker)is a moth that attacks dates from ripening stages while on tree,throughout storage,and until consumption,causing enormous qualitative and quantitative damages,resulting in economic losse...Cadra(Ephestia)cautella(Walker)is a moth that attacks dates from ripening stages while on tree,throughout storage,and until consumption,causing enormous qualitative and quantitative damages,resulting in economic losses.Image-processing algorithms were developed for detecting and differentiating between three Cadra egg categories based on the success of Trichogramma bourarachae(Pintureau and Babaul)parasitization.These categories were parasitized(black and dark red),unparasitized fertile unhatched(yellow),and unparasitized hatched(white)eggs.Color,light intensity,and shape information was used to develop detection algorithms.Two image processing methods were developed based on three randomly selected images and were tested on a larger validation image set of 40 images:(i)segmentation and extractions of color and morphological features followed by Watershed delineation,and is referred to as Algorithm 1(ALGO1),(ii)finding circular objects by Hough Transformation followed by convolution filtering,and is referred to as Algorithm 2(ALGO2).ALGO1 and ALGO2 achieved correct classification rates(CCRs)for parasitized eggs of 92%and 96%,respectively.Their CCRs for unhatched eggs were 48%and 94%,and for hatched eggs were 42%and 73%,respectively.Regarding parasitized eggs,both methods performed satisfactorily,but,in general,ALGO2 outperformed ALGO1.These results ensure automatic evaluation of the efficiency of biological control of Cadra cautella by the egg parasitoid Trichogramma bourarachae by quantifying the rate of parasitization.The developed detection methods can be used by producers of biocontrol agents for online monitoring of Trichogramma and similar insect natural enemies during mass production and before release against crop pests.Moreover,with few adjustments these methods can be used in similar applications such as detecting plant diseases.展开更多
The quality of the stator winding coil directly affects the performance of the motor.A dual-camera online machine vision detection method to detect whether the coil leads and winding regions were qualified was designe...The quality of the stator winding coil directly affects the performance of the motor.A dual-camera online machine vision detection method to detect whether the coil leads and winding regions were qualified was designed.A vision detection platform was designed to capture individual winding images,and an image processing algorithm was used for image pre-processing,template matching and positioning of the coil lead area to set up a coordinate system.After eliminating image noise by Blob analysis,the improved Canny algorithm was used to detect the location of the coil lead paint stripped region,and the time was reduced by about half compared to the Canny algorithm.The coil winding region was trained with the ShuffleNet V2-YOLOv5s model for the dataset,and the detect file was converted to the Open Neural Network Exchange(ONNX)model for the detection of winding cross features with an average accuracy of 99.0%.The software interface of the detection system was designed to perform qualified discrimination tests on the workpieces,and the detection data were recorded and statistically analyzed.The results showed that the stator winding coil qualified discrimination accuracy reached 96.2%,and the average detection time of a single workpiece was about 300 ms,while YOLOv5s took less than 30 ms.展开更多
Weeds normally grow in patches and spatially distributed in field. Patch spraying to control weeds has advantages of chemical saving, reduced cost and environmental pollution. Advent of electro-optical sensing capabil...Weeds normally grow in patches and spatially distributed in field. Patch spraying to control weeds has advantages of chemical saving, reduced cost and environmental pollution. Advent of electro-optical sensing capabilities has paved the way of using machine vision technologies for patch spraying. Machine vision system has to acquire and process digital images to make control decisions. Proper identification and classification of objects present in image holds the key to make control decisions and use of any spraying operation performed. Recognition of objects in digital image may be affected by background, intensity, image resolution, orientation of the object and geometrical characteristics. A set of 16, including 11 shape and 5 texture-based parameters coupled with predictive discriminating analysis has been used to identify the weed leaves. Geometrical features were indexed successfully to eliminate the effect of object orientation. Linear discriminating analysis was found to be more effective in correct classification of weed leaves. The classification accuracy of 69% to 80% was observed. These features can be utilized for development of image based variable rate sprayer.展开更多
The conventional Close circuit television(CCTV)cameras-based surveillance and control systems require human resource supervision.Almost all the criminal activities take place using weapons mostly a handheld gun,revolv...The conventional Close circuit television(CCTV)cameras-based surveillance and control systems require human resource supervision.Almost all the criminal activities take place using weapons mostly a handheld gun,revolver,pistol,swords etc.Therefore,automatic weapons detection is a vital requirement now a day.The current research is concerned about the real-time detection of weapons for the surveillance cameras with an implementation of weapon detection using Efficient–Net.Real time datasets,from local surveillance department’s test sessions are used for model training and testing.Datasets consist of local environment images and videos from different type and resolution cameras that minimize the idealism.This research also contributes in the making of Efficient-Net that is experimented and results in a positive dimension.The results are also been represented in graphs and in calculations for the representation of results during training and results after training are also shown to represent our research contribution.Efficient-Net algorithm gives better results than existing algorithms.By using Efficient-Net algorithms the accuracy achieved 98.12%when epochs increase as compared to other algorithms.展开更多
Angle detection is a crucial aspect of industrial automation,ensuring precise alignment and orientation ofcomponents in manufacturing processes.Despite the widespread application of computer vision in industrialsettin...Angle detection is a crucial aspect of industrial automation,ensuring precise alignment and orientation ofcomponents in manufacturing processes.Despite the widespread application of computer vision in industrialsettings,angle detection remains an underexplored domain,with limited integration into production lines.Thispaper addresses the need for automated angle detection in industrial environments by presenting a methodologythat eliminates training time and higher computation cost on Graphics Processing Unit(GPU)from machinelearning in computer vision(e.g.,Convolutional Neural Networks(CNN)).Our approach leverages advanced imageprocessing techniques and a strategic combination of algorithms,including contour selection,circle regression,polar warp transformation,and outlier detection,to provide an adaptive solution for angle detection.By configuringthe algorithm with a diverse dataset and evaluating its performance across various objects,we demonstrate itsefficacy in achieving reliable results,with an average error of only 0.5 degrees.Notably,this error margin is 3.274times lower than the acceptable threshold.Our study highlights the importance of accurate angle detection inindustrial settings and showcases the reliability of our algorithm in accurately determining angles,thus contributingto improved manufacturing processes.展开更多
基金the National Natural Science Foundation of China (No.51275223)。
文摘To avoid colliding with trees during its operation,a lawn mower robot must detect the trees.Existing tree detection methods suffer from low detection accuracy(missed detection)and the lack of a lightweight model.In this study,a dataset of trees was constructed on the basis of a real lawn environment.According to the theory of channel incremental depthwise convolution and residual suppression,the Embedded-A module is proposed,which expands the depth of the feature map twice to form a residual structure to improve the lightweight degree of the model.According to residual fusion theory,the Embedded-B module is proposed,which improves the accuracy of feature-map downsampling by depthwise convolution and pooling fusion.The Embedded YOLO object detection network is formed by stacking the embedded modules and the fusion of feature maps of different resolutions.Experimental results on the testing set show that the Embedded YOLO tree detection algorithm has 84.17%and 69.91%average precision values respectively for trunk and spherical tree,and 77.04% mean average precision value.The number of convolution parameters is 1.78×10^(6),and the calculation amount is 3.85 billion float operations per second.The size of weight file is 7.11MB,and the detection speed can reach 179 frame/s.This study provides a theoretical basis for the lightweight application of the object detection algorithm based on deep learning for lawn mower robots.
文摘Cadra(Ephestia)cautella(Walker)is a moth that attacks dates from ripening stages while on tree,throughout storage,and until consumption,causing enormous qualitative and quantitative damages,resulting in economic losses.Image-processing algorithms were developed for detecting and differentiating between three Cadra egg categories based on the success of Trichogramma bourarachae(Pintureau and Babaul)parasitization.These categories were parasitized(black and dark red),unparasitized fertile unhatched(yellow),and unparasitized hatched(white)eggs.Color,light intensity,and shape information was used to develop detection algorithms.Two image processing methods were developed based on three randomly selected images and were tested on a larger validation image set of 40 images:(i)segmentation and extractions of color and morphological features followed by Watershed delineation,and is referred to as Algorithm 1(ALGO1),(ii)finding circular objects by Hough Transformation followed by convolution filtering,and is referred to as Algorithm 2(ALGO2).ALGO1 and ALGO2 achieved correct classification rates(CCRs)for parasitized eggs of 92%and 96%,respectively.Their CCRs for unhatched eggs were 48%and 94%,and for hatched eggs were 42%and 73%,respectively.Regarding parasitized eggs,both methods performed satisfactorily,but,in general,ALGO2 outperformed ALGO1.These results ensure automatic evaluation of the efficiency of biological control of Cadra cautella by the egg parasitoid Trichogramma bourarachae by quantifying the rate of parasitization.The developed detection methods can be used by producers of biocontrol agents for online monitoring of Trichogramma and similar insect natural enemies during mass production and before release against crop pests.Moreover,with few adjustments these methods can be used in similar applications such as detecting plant diseases.
基金National Natural Science Foundation of China(No.U1831123)。
文摘The quality of the stator winding coil directly affects the performance of the motor.A dual-camera online machine vision detection method to detect whether the coil leads and winding regions were qualified was designed.A vision detection platform was designed to capture individual winding images,and an image processing algorithm was used for image pre-processing,template matching and positioning of the coil lead area to set up a coordinate system.After eliminating image noise by Blob analysis,the improved Canny algorithm was used to detect the location of the coil lead paint stripped region,and the time was reduced by about half compared to the Canny algorithm.The coil winding region was trained with the ShuffleNet V2-YOLOv5s model for the dataset,and the detect file was converted to the Open Neural Network Exchange(ONNX)model for the detection of winding cross features with an average accuracy of 99.0%.The software interface of the detection system was designed to perform qualified discrimination tests on the workpieces,and the detection data were recorded and statistically analyzed.The results showed that the stator winding coil qualified discrimination accuracy reached 96.2%,and the average detection time of a single workpiece was about 300 ms,while YOLOv5s took less than 30 ms.
文摘Weeds normally grow in patches and spatially distributed in field. Patch spraying to control weeds has advantages of chemical saving, reduced cost and environmental pollution. Advent of electro-optical sensing capabilities has paved the way of using machine vision technologies for patch spraying. Machine vision system has to acquire and process digital images to make control decisions. Proper identification and classification of objects present in image holds the key to make control decisions and use of any spraying operation performed. Recognition of objects in digital image may be affected by background, intensity, image resolution, orientation of the object and geometrical characteristics. A set of 16, including 11 shape and 5 texture-based parameters coupled with predictive discriminating analysis has been used to identify the weed leaves. Geometrical features were indexed successfully to eliminate the effect of object orientation. Linear discriminating analysis was found to be more effective in correct classification of weed leaves. The classification accuracy of 69% to 80% was observed. These features can be utilized for development of image based variable rate sprayer.
文摘The conventional Close circuit television(CCTV)cameras-based surveillance and control systems require human resource supervision.Almost all the criminal activities take place using weapons mostly a handheld gun,revolver,pistol,swords etc.Therefore,automatic weapons detection is a vital requirement now a day.The current research is concerned about the real-time detection of weapons for the surveillance cameras with an implementation of weapon detection using Efficient–Net.Real time datasets,from local surveillance department’s test sessions are used for model training and testing.Datasets consist of local environment images and videos from different type and resolution cameras that minimize the idealism.This research also contributes in the making of Efficient-Net that is experimented and results in a positive dimension.The results are also been represented in graphs and in calculations for the representation of results during training and results after training are also shown to represent our research contribution.Efficient-Net algorithm gives better results than existing algorithms.By using Efficient-Net algorithms the accuracy achieved 98.12%when epochs increase as compared to other algorithms.
文摘Angle detection is a crucial aspect of industrial automation,ensuring precise alignment and orientation ofcomponents in manufacturing processes.Despite the widespread application of computer vision in industrialsettings,angle detection remains an underexplored domain,with limited integration into production lines.Thispaper addresses the need for automated angle detection in industrial environments by presenting a methodologythat eliminates training time and higher computation cost on Graphics Processing Unit(GPU)from machinelearning in computer vision(e.g.,Convolutional Neural Networks(CNN)).Our approach leverages advanced imageprocessing techniques and a strategic combination of algorithms,including contour selection,circle regression,polar warp transformation,and outlier detection,to provide an adaptive solution for angle detection.By configuringthe algorithm with a diverse dataset and evaluating its performance across various objects,we demonstrate itsefficacy in achieving reliable results,with an average error of only 0.5 degrees.Notably,this error margin is 3.274times lower than the acceptable threshold.Our study highlights the importance of accurate angle detection inindustrial settings and showcases the reliability of our algorithm in accurately determining angles,thus contributingto improved manufacturing processes.
