Crowd density estimation is an important topic in the fields of machine learning and video surveillance. Existing methods do not provide satisfactory classification accuracy; moreover, they have difficulty in adapting to complex scenes. Therefore, we propose a method based on convolutional neural networks (CNNs). The proposed method improves performance of crowd density estimation in two key ways. First, we propose a feature pooling method named mixed pooling to regularize the CNNs. It replaces deterministic pooling operations with a parameter that, by studying the algorithm, could combine the conventional max pooling with average pooling methods. Second, we present a classification strategy, in which an image is divided into two cells and respectively categorized. The proposed approach was evaluated on three datasets: two ground truth image sequences and the University of California, San Diego, anomaly detection dataset. The results demonstrate that the proposed approach performs more effectively and easily than other methods.
We have been developing a computer-aided diagnosis (CAD) system for automatically recognizing cervical cancer cells from Papanicolaou smear. Considering that pathological changes of cervix can be indicated by the abnormity of the nucleus of intermediate cell, the key task of this system is to find the intermediate cells and segment the nucleus precisely. This paper presents a novel approach for automatic segmentation of microscopic cervical cell images using multispectral imaging techniques. In order to capture images at different wavelengths, a Liquid Crystal Tunable Filter (LCTF) device is used to provide wavelength selection from 400nm to 720nm with an increment of 10nm. Considering the spectral variances of background, nucleus and cytoplasm, background is extracted firstly from the microscopic images by calculating pixel intensity variance at 470nm, 530nm, 570nm, 580nm and 650nm. Then superficial cells are extracted apart from intermediate cells easily at 530nm 650nm because of the different pixel intensity distribution of the two kinds of cells at these two wavelengths. To segment the nucleus from intermediate cells, we adopt two procedures. Firstly, the nuclei are roughly segmented apart by using an iterative maximum deviation between-cluster algorithm. Secondly, a novel rigorous algorithm based on active contour model is adopted to achieve more exact nuclei segmentation. Using the method proposed in this paper, we did experiments on over 300 cervical smears, and the results show that this method is more robust and precise.
Counting of different classes of white blood cells in bone marrow smears can give pathologists valuable information regarding various cancers. But it is tedious to manually locate, identify, and count these classes of cells, even by skilled hands. This paper presents a novel approach for automatic detection of White Blood Cells in bone marrow microscopic images. Different from traditional color imaging method, we use multispectral imaging techniques for image acquisition. The combination of conventional digital imaging with spectroscopy can provide us with additional useful spectral information in common pathological samples. With our spectral calibration method, device-independent images can be acquired, which is almost impossible in conventional color imaging method. A novel segmentation algorithm using spectral operation is presented in this paper. Experiments show that the segmentation is robust, precise, with low computational cost and insensitive to smear staining and illumination condition. Once the nuclei and cytoplasm have been segmented, more than a hundred of features are extracted under the direction of a pathologist, including shape features, textural features and spectral ratio features. In pattern recognition, a maximum likelihood classifier (MLC) is implemented in a hierarchical tree. The classification results are also discussed. This paper is focused on image acquisition and segmentation.
Cervical cancer is the second most common cancer among women worldwide. Early detection of cervical cancer is very important for successful treatment and increasing survival. Papanicolaou test is the most popular and effective screening test for cervical cancer, but it is highly subjective and skilled-labor intensive. We report a multispectral imaging microscopic system for Papanicolaou smear analysis for early detection of cervical cancer. Different from traditional color imaging method, we use multispectral imaging techniques for image acquisition, which can simultaneously record spectral and spatial information of a sample. A liquid crystal tunable filter (LCTF) is coupled in the light microscope for fast wavelength selection and a two-dimensional cooled charge-coupled device (CCD) for image capture. In this paper, the multispectral image acquisition method is introduced, including exposure control and spectral calibration, which makes the images not so dependent on imaging devices. In the image segmentation process, an effective algorithm using spectral ratio method is applied for cell nuclei detection. This segmentation method can easily detect the nuclei and diminish the influence of the cytoplasm overlap. Results show that our segmentation is more robust and precise than conventional color imaging method which is heavily dependant on sample staining and illumination conditions while with high speed. Once the nuclei have been segmented, cell features including morphological and textural features are measured. A genetic algorithm is used for feature selection and a support vector machine(SVM)is used for training and classification. This paper is focused on image acquisition and segmentation.
This paper describes a novel multispectral imaging microscope that can simultaneously record both spectral and spatial information of a sample, which can take advantage of spatial image processing and spectroscopic analysis techniques. A Liquid Crystal Tunable Filter device is used for fast wavelength selection and a cooled two-dimensional monochrome CCD for image detection. In order to acquire images that are not so dependent on imaging devices, a clever CCD exposure time control and a software based spectral and spatial calibration process is performed to diminish the influence of illumination, optic ununiformity, CCD’s spectral response curve and optic throughput property. A set of multispectral image processing and analysis software package is developed, which covers not only general image processing and analysis functions, and also provides powerful analysis tools for multispectral image data, including multispectral image acquisition, illumination and system response calibration, spectral analysis and etc. The combination of spatial and spectral analysis makes it an ideal tool for the applications to biomedicine. In this paper, two applications in biomedicine are also presented. One is medical image segmentation. Using multispectral imaging techniques, a mass of experiments on both marrow bone and cervical cell images showed that our segmentation results are highly satisfactory while with low computational cost. Another is biological imaging spectroscopic analysis in the study of pollen grains in rice. The results showed that the transmittance analysis of multispectral pollen images can accurately identify the pollen abortion stage of male-sterile rice, and can easily distinguish a variety of male sterile cytoplasm.
This paper presents a novel approach for urban texture analysis. The approach applies the artificial immune theory in learning the texture filters for urban texture classifications. In this paper, urban textures are regarded as non-self, and non-urban textures are regarded as self. Texture filters are regarded as antibodies. The clonal selection algorithm is employed to evolve antibodies. Experimental results of urban texture analysis on aerial images are presented to illustrate the feasibility of the proposed method.
This paper introduces a novel methodology for texture object detection using genetic algorithms. The method employs a kind of high performance detection filter defined as 2D masks, which are derived using genetic algorithm operating. The population of filters iteratively evaluated according to a statistical performance index corresponding to object detection ability, and evolves into an optimal filter using the evolution principles of genetic search. Experimental results of texture object detection in high resolution satellite images are presented to illustrate the merit and feasibility of the proposed method.
This paper introduces a novel methodology for object detection using genetic algorithms and morphological processing. The method employs a kind of object oriented structuring element, which are derived using genetic algorithm operating. The population of morphological filters iteratively evaluated according to a statistical performance index corresponding to object extraction ability, and evolves into an optimal structuring elements using the evolution principles of genetic search. Experimental results of object extraction in high resolution satellite images are presented to illustrate the merit and feasibility of the proposed method.
Image textures have been playing an important role in image recognition and interpretation. The selection of image texture features is the key part to classify image objects. An evolution approach for texture image feature selection is prosed in this paper. The approach uses evolution algorithms as the primary search component. Based on selected texture features, a fuzzy cluster method is used for texture image classification. Experimental result on color aerial images show the feasibility of the proposed method.
The effect of He-Ne laser on the intracellular bactericidal and digestive function to C albicans of mice peritoneal M(theta) has been studied with the fluorescence microscope after acridine organge staining. The results indicated that the bactericidal and digestive function of M(theta) in irradiated groups, expressed more active than that in the non-irradiated group, and showed significant difference. The comparison between the different irradiated groups also showed marked difference. Ultrastructure changes of M(theta) were observed under the E/M and the content of a-Acetate Naphthy esterase in lysosome were measured by image analysis, the results demonstrated that M(theta) in the irradiated groups present marked change in ultrastructure and the GN, GA, GA/CA, IOD of esterase increased significantly. The results suggested that the He-Ne laser with appropriate dosage could activate M(theta) , and enhance anti-infection immunity.