No abstract

Defect detection on solid wood surface has two main problems: (1) the real-time performance of the available methods are poor despite good detection accuracy, and (2) the defect extraction process is complicated. Here, we propose a mixed, fully convolutional neural network (Mix-FCN) to detect the location of wood defects and classify the types of defects from the wood surface images automatically. The images were collected first by a data acquisition device developed in our laboratory. We then employed TensorFlow and Python language to construct a VGG16 model. We used two kinds of datasets (dataset1 and dataset2) to maximize the limited, collected data and enable the Mix-FCN to converge rapidly during training. The weights of the filters in front of the Mix-FCN during training were initialized from the trained VGG16 model. The weights of the VGG16 net were learned by dataset1. Our model was trained, validated, and tested by dataset 2. Overall classification accuracy (OCA), pixel accuracy (PA), mean intersection over union, detection rate, missing alarm, false alarm rate, and precision were used to evaluate the network, and the performance was good based on the seven evaluation indicators. We achieved 99.14% OCA and 91.31% PA, and a batch of 50 images required only 0.368 s of detection time. Our proposed method has better accuracy and less detection time compared to the previous methods of wood detection. INDEX TERMS Deep learning, full convolutional neural network, transfer learning, wood defects detection.

No abstract

No abstract

Optical microscopy allows us to study living fluorescent biological samples. Optical sectioning is a technique to obtain three-dimensional (3D) information about the observed object by acquiring a stack of two-dimensional (2D) images at different depths through the sample. However, the specific shape of the 3D optical transfer function of the optical microscope leads to images presenting defects, such as, for example, an apparent elongation along the vertical axis. It is therefore necessary to preprocess the images before any quantitative measurement is performed. This image restoration can be obtained by deconvolution of the acquired 3D image. We have tested several deconvolution algorithms on synthetic images, obtained by convolution of a solid sphere with a measured point spread function. We have compared the restored image with the original one (shape and volume). The linear least-squares method is fast, but artefacts are still present in the restored images. The Carrington method is well adapted to thin objects. The maximum likelihood-expectation maximization method leads to a good reconstruction of the object, but is very time consuming.Résumé. La microscopie optique permet l'étude de spécimens biologiques vivants et fluorescents. La technique par coupes sériées donne des informations tridimensionnelles (3D) sur l'objetétudié par l'acquisition d'une pile d'images bidimensionnellesà différentes profondeurs de focalisationà travers l'échantillon. Les spécificités de la fonction de transfert optique 3D du microscope conduisentà des images présentant des défauts, comme par exemple uneélongation apparente selon l'axe vertical. Il est donc nécessaire de traiter les images avant toute mesure quantitative. On procèdeà une déconvolution de l'image 3D obtenue. Nous avons testé différents algorithmes de déconvolution sur des images de synthèse obtenues par convolution d'une bille pleine avec une fonction de transfert optique mesurée. Nous avons comparé, en forme et en volume, les images restaurées avec l'image d'origine. La méthode 'linear least square' est rapide, mais l'image restaurée présente des artefacts. La méthode de Carrington est bien adaptéeà la restauration d'objets fins. La méthode 'maximum likelihood-expectation maximization' permet une bonne reconstruction des images, mais demande de grands temps de calcul.