Search

The rapid developments of mobile robotics and autonomous navigation over the years are largely empowered by public datasets for testing and upgrading, such as SLAM and localization tasks. Impressive demos and benchmark results have arisen, indicating the establishment of a mature technical framework. However, from the view point of real-world deployments, there are still critical defects of robustness in challenging environments, especially in large-scale, GNSS-denied, textural-monotonous, and unstructured scenarios. To meet the pressing validation demands in such scope, we build a novel challenging robot navigation dataset in a large botanic garden of more than 48000m2. Comprehensive sensors are employed, including high-res/rate stereo Gray&RGB cameras, rotational and forward 3D LiDARs, and low-cost and industrial-grade IMUs, all of which are well calibrated and accurately hardware-synchronized. An all-terrain wheeled robot is configured to mount the sensor suite and provide odometry data. A total of 32 long and short sequences of 2.3 million images are collected, covering scenes of thick woods, riversides, narrow paths, bridges, and grasslands that rarely appeared in previous resources. Excitedly, both highly-accurate ego-motions and 3D map ground truth are provided, along with fine-annotated vision semantics. Our goal is to contribute a high-quality dataset to advance robot navigation and sensor fusion research to a higher level., Submitted to IEEE RA-L for possible publications

Laser Doppler vibrometers (LDVs) are widely used for vibration testing in various fields. Nonlinearity errors are the key factor affecting the measurement accuracy of LDVs. The conventional Heydemann method cannot correct nonlinearity errors produced by noisy environments. Thus, we establish a novel model to describe dynamic nonlinearity errors produced in noisy environments and propose a compensation method to mitigate signal distortion. The performance of the proposed method is assessed by performing both simulations and experiments. The results of experiments carried out in a noisy environment indicate that the proposed method suppresses the nonlinearity to 30 nm compared to 737 nm using the conventional Heydemann correction. The proposed method can improve the accuracy of LDV measurements in industrial environments.

The energy accuracy of laser beams is an essential property of the inertial confinement fusion (ICF) facility. However, the energy gain is difficult to control precisely by traditional Frantz-Nodvik equations due to the dramatically-increasing complexity of the huge optical system. A novel method based on ensemble deep neural networks is proposed to predict the laser output energy of the main amplifier. The artificial neural network counts in 39 more related factors that the physical model neglected, and an ensemble method is exploited to obtain robust and stable predictions. The sensitivity of each factor is analyzed by saliency after training to find out the factors which should be controlled strictly. The identification of factor sensitivities reduces relatively unimportant factors, simplifying the neural network model with little effect on the prediction results. The predictive accuracy is benchmarked against the measured energy and the proposed method obtains a relative deviation of 1.59% in prediction, which has a 2.5 times improvement in accuracy over the conventional method.

We study dispersion mismatches and compensation in high-resolution frequency scanning interference measurements. Compared to semiconductor lasers, external cavity lasers can, in theory, provide wideband frequency modulation and high-resolution measurements. However, experiments indicate that the dispersion mismatch cannot be ignored in measurement systems containing wideband tuning lasers. Fibre auxiliary interferometer external clock sampling results indicate that the beat frequency of measurement signals exhibits linear chirp. In turn, the frequency spectrum of the measurement signal broadens, thus lowering the measurement resolution and limiting the measurement distance. We present a dispersion mismatch model and analyse the associated measurement effects. We also propose a dispersion compensation method based on a focused definition evaluation function. By studying the relationship between the compensation coefficient and evaluation function, the optimal compensation coefficient can be determined quickly and accurately. Our experiments demonstrate that within a tuning bandwidth of 40 nm, the peak full width at half maximum (FWHM) after dispersion compensation is 32.2 µm at a distance of 3.8931671 m, which is close to the theoretical resolution. Our method provides a technical approach to realizing large, high-resolution distance measurements for broadband laser frequency scanning interferometers.

Under the condition of inhomogeneous total internal reflection illumination, a novel approach based on machine learning is proposed to solve the problem of damage online inspection in large-aperture final optics. The damage online inspection mainly includes three problems: automatic classification of true and false laser-induced damage (LID), automatic classification of input and exit surface LID and size measurement of the LID. We first use the local area signal-to-noise ratio (LASNR) algorithm to segment all the candidate sites in the image, then use kernel-based extreme learning machine (K-ELM) to distinguish the true and false damage sites from the candidate sites, propose autoencoder-based extreme learning machine (A-ELM) to distinguish the input and exit surface damage sites from the true damage sites, and finally propose hierarchical kernel extreme learning machine (HK-ELM) to predict the damage size. The experimental results show that the method proposed in this paper has a better performance than traditional methods. The accuracy rate is 97.46% in the classification of true and false damage; the accuracy rate is 97.66% in the classification of input and exit surface damage; the mean relative error of the predicted size is within 10%. So the proposed method meets the technical requirements for the damage online inspection.

A high precision inspection method based on image mosaic was proposed for the final optics inspection in an Inertial Confinement Fusion(ICF).The image mosaic method contains image registration stage and image fusion stage.In the image registration stage,a registration method based on feature matching of local damaged area was used to overcome the exposure difference between the subimages and the lack of image information,and the moving information in the inspection system was used in image matching to improve the inspection precision and inspection efficiency.In the image fusion stage,the Poisson image fusion was adopted to eliminate the "stitching line" caused by an exposure difference and to implement the seamless mosaic.The images of final optics inspected by Final Optics Online Inspection(FOOI)was compared with the proposed method and Scale Invariant FeatureTransform(SIFT)method,and the results show that the proposed method is superior to SIFT method.The 4×4mosaic image shows that the inspection resolution of the FOOI device in the ICF has been improved significantly by proposed method.

In machine vision measurement, the edge is a key point for fitting geometric parameter. There are two problems in the edge detection process. First, there is redundant information for the object with complex shape in the field of the view. Second, the surface of the object is full of texture which is misidentified as the edge. The texture processes similar feature to the edge and cannot be removed by filter. To solve the above problems, vision sight is proposed to get an interesting region and remove redundant information. A new algorithm based on fuzzy entropy is used to auto-estimate the edge detection direction from the pure region to mixed region in order to avoid the textures which misidentified as the edge. Comparing the algorithm with Canny, the former gets less texture points than the latter. A mask film is used as a standard to weight the validity of the algorithm. The experimental result shows that the algorithm proposed by this paper is robust and accuracy in detecting edge.

In the paper we study the method of reducing environmental influence in broadband laser frequency scanning interferometer. Target displacement caused by vibration will result in Doppler shift in measurement beat frequency. The extent of frequency shift is usually much larger than the actual target displacement. So the direct calculating of the target distance will cause ranging precision to decrease. In this paper, we establish a model for the influence of environmental vibration on the measurement and analyze the influence of the vibration on ranging result. To suppress the vibration effect, the Kalman filter is combined with the overlapping Chirp Z transform to estimate the measured data. The general process is described as follows. Firstly, the tuning nonlinearity will lead to the frequency spectrum broadening, so this paper we use the frequency sampling method to correct the frequency modulation nonlinearity of the laser. The frequency sampling method has the advantages of high speed and high precision. Secondly, the measurement system has the dispersion mismatch effect due to the use of broadband frequency swept laser. To solve this problem, the influence of the dispersion on the measurement is reduced by using the method of dispersion chirp slope calibration. Thirdly, because of the long frequency sweep period of the external cavity swept frequency laser, the vibration process of the target cannot be recorded in real time by single sweep, so in this paper we propose segmenting the measurement signal of single sweep and conducting Chirp Z transform to calculate target distance at different times. Compared with FFT algorithm, Chirp Z transform can achieve arbitrary narrow band spectrum subdivision, with the advantages of high accuracy and fast frequency measurement. Lastly, the Chirp Z ranging result is further combined with the method of Kalman filter to estimate the state of the target distance information. The experimental results indicate that the measurement standard is reduced from 185.4 μm to 9 μm by the proposed method. Without changing the absolute distance measuring device of broadband laser frequency scanning interferometer, this method provides a solution for further improving the ranging accuracy in the vibration environment, and reduces the complexity and cost of the device.

Automotive detection of defects plays an important role in Final Optics Damage Online Inspection (FODOI). Because of the tiny size of defects compared to the image, detection of the defects is a challenge. In a 2k×2k FODOI image, the size of the defects is only several pixels. Moreover, the gray value of different image areas is different because of the uneven distribution of illumination. In this paper, a robust defects detecting method based on Local Area Signal Strength (LASS) and 2-D histogram is theoretically and experimentally proposed. After the image stretching process achieved by grayscale morphology arithmetic, we computed the LASS at every pixel in the image, resulting in a new image with LASS values for each pixel. Thus, an adaptive threshold selection algorithm in 2-D histogram formed by LASS was achieved. The proposed algorithm was compared with some other fast algorithms, such as the traditional 2-D histogram algorithm, Otsu algorithm, and the FCM clustering algorithm; the results show that the proposed algorithm has higher detection precision and is robust in uneven illumination.

The paper proposes a method of using dots in an array as targets for photogrammetry to reconstruct the surface of a large-size flexible antenna efficiently. An analysis demonstrates the impacts of the spot size and the regional feature on the point extraction accuracy, and introduces a fitting-ellipse algorithm in this paper that eliminates the deformation and the shielding. In the feature extraction and matching processes, this article puts forward a feature identification technique based on the area and a matching algorithm depending on the order constraint. The paper closes with an experiment that shows the equipment that can measure a gossamer area of 1m×1.2m in time of less than 1s, which shows its advantages of high accuracy and high efficiency in detecting the shape of a large field.