Capsicum annuum (Bell Pell, Shouhe Co., Ltd, Shandong, China) seeds were sown in plug tray containing cultivated soil (organic matter: 478 g/1000 g, N + P + k: 6.75 mg/1000 g, probiotics: 90 µg/1000 g, trace element: 1.7 µg/1000 g, humic acid: 130 µg/1000 g, pH 6.5–6.8). After germinating in greenhouse in Nanjing Agricultural University, Engineering College (32°18′N, 118°46′E) for 18 days, pepper seedlings with two true leaves were transplanted to 3 experimental pots (1.2 m × 1.2 m × 0.4 m). The plant spacing was 25 cm and each plant was irrigated uniformly of 200 ml water every day with sprinkle from the field planting day. The soil (brown clay soil, organic matter: 318 g/1000 g, total N: 1.6 g/1000 g) in the pots was taken from local vegetable growing area, and was filtered 3 times with 1 cm opening to filter soil blocks larger than 1 cm³. The soil compaction was 0.81 kg/cm² at the field planting day.

In the first experiment (root architecture 3D reconstruction), plants were sampled in 7 days, 13 days, 19 days, 25 days, 31 days and 39 days after field planting, 3 replicates. In the second experiment (interception probability in different installation patterns), there were totally 8 installation patterns, 12 replicates. In the third experiment (comparing the microrhizotron with soil sampling), 16 samples were used to compare microrhizotron obtained root indicators with manual measured root indicators.

Since traditional minirhizotron was always large in diameter, it was not suitable to detect roots of shallow plant like peppers, which might cause great effects on root development. In order to observe roots development of shallow root plant in real time and multiple points, a kind of micro acquisition and processing system (mcrorhizotron) for multiple roots observation was proposed and designed in this paper.

The mcrorhizotron consisted of a micro camera (HDMINICAM, Guangdong, China), an optical amplifier (amplified by 20 times), LED, a power supply and wireless module integrated in the camera, and image processing part (Fig. 1). The camera is 12 million pixel CMOS (complementary metal oxide semiconductor), resolution 1080p, angle of view 120°, working temperature − 20 °C − 80 °C, relative humidity 15–85%. Four LED light sources were installed on the top of the microrhizotron to illuminate imaging area. In order to prevent short circuit troubles, the circuit part of the system was sealed with hot melt glue and sprayed with waterproof paint for circuit board (A-2577, China) [35]. The size of the system was 1.1 cm × 1.1 cm × 1.2 cm. The collected root image was transmitted to the terminal receiving device (mobile phone or personal computer) through the WIFI signal formed by the wireless control module (Model: MEDIATEK MT7601).Fig. 1

Structure diagram and photo of micro root observation system. a Structure diagram of microrhizotron. b Picture of microrhizotron

The JPEG files that contained the images of roots were used to determine the root length (including root hair and very fine roots) and root diameter. The root image was preprocessed through image normalization, enhancement, and median filter, and then through image morphology operation, skeleton image in single pixel were extracted [36, 37] as is shown in Fig. 2. In Fig. 2a1, a3 were the same root with different root hair or fine roots observed in different days. Figure 2a2, a4 were extracted root skeleton in single pixel. It was obvious that Fig. 2a2 had more and longer root hair or fine roots than that of Fig. 2a4. Root length was measured through calculating connections between adjacent pixels including diagonal connection or straight-line connection in skeleton image [37]. If two adjacent pixels is straight-line connection, the distance between the two pixels is 1; if it is diagonal connection, the distance is \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt 2$$\end{document}2. Code and script were developed by authors. However, in this way of automatic calculation, both the root axes and very fine root (root hairs) were included. And in root images (Fig. 2), fine root (root hairs) composed the most part. In order to analyze different root traits and to improve measuring accuracy, root axes and root hairs should be separated. Therefore, after the total root length was calculated automatically in skeleton image, root axes length was measured using ruler tools in Adobe Photoshop CS6 Extended. Root surface area was measured by counting the root pixels and calculating total area by multiplying area of each pixel based on the segmented binary image [35]. Root length and diameter was calculated by:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$l = (\sqrt 2 N_{1} + N_{2} )K/{\text{f}}$$\end{document}l=(2N1+N2)K/f2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S_{r} = K \times P/f$$\end{document}Sr=K×P/f3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d = S_{r} /l$$\end{document}d=Sr/lwhere l is root length, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S_{r}$$\end{document}Sr is root surface area, f is the magnification of the optical element, P is the total number of root pixels, K is the corresponding relation between the pixel distance on the image and the actual pixel distance, d is mean root diameter. N1 represents the number of pixels connected diagonally, and N2 represents the number of pixels connected horizontally and vertically (Additional files 1, 2). Fig. 2

Source picture and processed image of roots. a1 Source picture of roots observed in 28 days. a2 Skeleton of a1. a3 Source picture of roots observed in 60 days. a4 Skeleton of a3

The root samples were excavated using a columnar sampler every 7 days (10 days after field planting, taking the center of the plant root system as the center, 10 cm as the radius (Fig. 3a)). The root architecture was measured using a three-dimensional caliper as shown in Fig. 3b. This instrument was mainly composed of height ruler, horizontal direction ruler, rotating disk ruler and laser centering device. First, the sample of the soil column and rotating disk were centered with a line laser, and then the top soil was gently removed with a small brush from the upper layer. The underlying soil of the root was preserved to make the root attached to it, so as not to destroy the original spatial position. The exposed root was measured with the tip of the three-dimensional caliper at each 2–3 mm interval, and it was cut off from the base after the last point was measured (Fig. 3c). The measured data were transformed from cylindrical coordinate system to Cartesian coordinate system, and the root system was reconstructed and analyzed with Matrix laboratory (version: 7.11.0.584). The code or script was developed by author.Fig. 3

Root spatial structure measuring instrument. a Sampling of pepper roots. b Measuring instrument. c Measurement of lateral roots architecture

Roots of capsicum belong to diarch, and the lateral roots initiate from two opposite poles of protoxylem pole pericycle cell. Lateral roots are generally spaced along the longitudinal axis of the main root. Although it has long been known that lateral roots are symmetrically located, the spatial distribution of lateral roots after elongation remains unknown. In order to figure out root system architecture traits, the spatial structure after root elongation was further analyzed in circumferential direction, radial direction, and vertical direction (Fig. 4). α was used to characterize the angular change in the vertical direction, and it was the angle between lateral root (each elongation of 20 mm) and plant center line. L was used to characterize the circumferential spreading length of lateral roots, and it was the distance between lateral root (each elongation of 30 mm) and plant center line in horizontal direction.Fig. 4

Installation patterns of microrhizotrons. a Installation patterns of microrhizotrons. M1, M2, M3 are microrhizotrons in different vertical angles and depth. α is vertical angle, L is lateral roots extension, and R1 and R2 are root length density sampling radius. b, c, and d Roots images from M1, M2 and M3

Root length density was calculated every R = 10 mm radically by the formula as follows:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\text{RLD}} = \frac{{\mathop \sum \nolimits_{i = 1}^{n} l_{i} }}{V}$$\end{document}RLD=∑i=1nliVwhere RLD is root length density, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$l_{i}$$\end{document}li is length of root i, and V is volume of the soil with roots.

In order to reduce the disturbance caused by installation, the microrhizotrons were presetted at the field planting day. 1.5 cm³ (1.1 × 1.1 × 1.2 cm) of soil was removed (replaced by microrhizotron), and the camera of the micro observation system was set carefully towards the roots. The rest of the soil was covered and compacted gently by rubber hammer to ensure that the soil compactness was consistent with the original growing soil environment. We declared a root interception success when the captured root length (root diameter) increased or decreased between t day and t + 1 day.

As is shown in Fig. 4, there were a lot of possibilities to set microrhizotrons. Circumferential position around the plant, installation radiuses of microrhizotron to the root center line, vertical angle and depth may all contribute to the effectiveness of microrhizotrons. For example, M1, M2 and M3 were set at random into soil. M1 was able to intercept with roots, and M2 would intercept with more; but M3 was further from roots, so this set of microrhizotron might be invalid in observing roots. Therefore, to improve the capture of roots, microrhizotrons should be presetted reasonably. And our solution was to install microrhizotrons based on biological characteristics and spatial distribution characteristics of root system, including determining optimal circumferential position, radical position and vertical angle.

Comparison experiment was conducted with soil sampling using the microrhizotron developed. In the experiment, each plant was installed with 2 microrhizotrons (opposite) in installation pattern horizontal angle 45° + radius 30 mm + vertical angle 45° (depth 50 mm) with 16 replicates. At four dates, root image was captured and processed by microrhizotron system. At the same time, soil samples were extracted with a cut syringe (diameter 2 cm) as is shown in Fig. 5. The syringe was used to extract soil cylinder opposite the camera. After the soil cylinder together with the microrhizotron was extracted, they were carefully put out and cut off layer by layer until 3 mm to the camera [24, 25], thus the roots in the view of camera was acquired. The roots were then washed carefully with distilled water and wiped with absorbent paper, and then measured root length and diameter manually with microcallipe.Fig. 5

Soil sampling of roots opposite to the camera in 3 mm

Root 3D reconstruction for 4, 6, 10, 16, 24, and 35 leaves stage is shown in Fig. 6. Root system architecture depicted spatial configure of lateral roots, and root system in different stage showed very different architecture. There were only a small number (8 roots) of lateral roots at the 4 leaves stage, and these roots were very small and just existed within 5 cm soil layer. Roots in 6 leaves were much denser than that in 4 leaves. As the plant grew to 10 leaves, root number increased to 37; most roots spread in the top soil, and a few extended to deep soil. At the 35 leaves stage, there was only a slight increase in root number, but a large number of roots extended to soil layer deeper than 5 cm.Fig. 6

Root system architecture 3D reconstruction. 4, 6, 10, 16, 24 and 35, are numbers of true leaves, and they were sampled in 7 days, 13 days, 19 days, 25 days, 31 days and 39 days after field planting

From vertical view, lateral roots distributed radiative along primary root. Two perpendicular lines were introduced to divide the roots into 4 parts. Each line was about 45° to initials of lateral roots. Interestingly, roots tended to gather in red arrow area, but in blue arrow area, roots were much less (Fig. 7). For understanding of spatial distribution of lateral roots after elongation, the circumferential gathering traits was quantified by calculating root length in red arrow area and in blue arrow area in Fig. 8. The result showed that lateral roots were highly focused in the red arrow area. The highest root length percentage of red arrow area in the 4 leaves stage reached 96%. Although in 16 leaves stage, the percentage was a little lower (84%) than the beginning, the average percentage was 90%, indicating highly gathering of lateral roots.Fig. 7

Lateral root gathering characteristics in vertical view

Pepper root system is diarch, and the lateral roots initiate from a patch of protoxylem pole pericycle cell, and are symmetrically located in two rows [38, 39]. To some extent, our findings implied that spatial distribution of lateral roots after elongation was also restricted to some range, and restricted horizontal angle was a highly-conserved mechanism shared by most lateral roots.

Vertical angle represented geotropism and deepen ability of lateral roots. As is depicted in Figs. 9 and 10, root length increased as plant leaves increased. The shortest root length was 60 mm in 4 leaves and extended to 120 mm in 35 leaves. For most lateral roots, vertical angles were between 80° and 90° at the initial position, almost perpendicular to the root center line. This indicated that lateral roots extended circumferentially at first to support the plant. Vertical angle since 20 mm was much lower than that of the beginning, and plant roots in different leaves showed different variation trend. Vertical angle at 4 leaves stage and 6 leaves stage seemed to be larger than that of 10 leaves stage, 16 leaves stage, and 24 leaves stage, but not that in 35 leaves stage. These differences might be caused by plant root development, since a smaller vertical implied lateral roots plunged into deep soil, which helped to search for water and nutrition, and to improve lodging resistance. Also, root phenotype was not only controlled by gene, but also by environment. Therefore, roots architecture may adapt appropriately to the environment. Although there was some difference as root exploring in soil, vertical angle of lateral roots was between 30° and 60° after elongation until 120 mm.Fig. 9

Vertical angle of lateral roots

Root extension was the distance of root to plant center line. Individually, root extension had some difference, for example, root extension in 4 leaves was the lowest, and that in 6 leaves was the highest. This is because plant had individual difference, and also different reflection to environments. Generally, lateral roots extension linearly increased when root length was less than 80 mm. Then the increase slowed down, and reached peak value in root end. The longer the roots extent, the more water and nutrition it could access, and also more stable the plant would be. Therefore, the fast root extension in early days enabled a strong root basis for later development.

Root length density is heterogeneous in soil in different depths and radius. As is shown in Table 1, the root length density near the root center was the largest. With the increase of sampling radius, root length density decreased in general. The root system was concentrated in the shallow soil within 5 cm in 4 leaves, 6 leaves, and 10 leaves stages. When plants had more than 16 leaves, root system gradually appeared in the soil layer of 10 cm, but compared with 5 cm soil layer, root length density was much smaller in 10 cm soil layer. Few roots were observed in depth more than 10 cm.Table 1

Root length density distribution

Average root length density was only 0.09 in 4 leaves, and it increased to 1.081 in 24 leaves stage. Compared with average value, root length density in R = 10 mm was much greater, and root length density in R > 40 mm was much smaller. Root length density between R = 20 mm and R = 40 mm was approximate to the average value. Interestingly, from time series root length density increased to the maximal value at 24 leaves, and then decreased. On one hand, this may be caused by circumferential expansion. As is shown in 35 leaves, although root length density near to the center decreased, it increased in large radius. On another hand, this may be related to root senescence and death, and the growth of the shoot, but still needs to be proved rigorously.

According to the analysis of root initiate principle and root elongation discussed before, 8 installation patterns (12 replicates) for microrhizotron were tested and verified to determine high probability of root interception. They were the permutation and combination of 2 horizontal angles (45° and 90°), 2 radiuses of microrhizotron to the root center line (30 mm and 60 mm), and 2 vertical angles [18] (30° and 45°). Each plant was equipped with 3 microrhizotrons, and the depth of all the presets was 50 mm. It was convenient to observe root interception with microrhizotron every day, and in Fig. 9 we illustrated interception probability every 6 days.

As is shown in Fig. 11, probabilities of root interception under different installation patterns were greatly different. The highest interception probability appears in combination 001 with averaged value 96.7%, followed by 000 with probability 95%, and the lowest probability was only 46.7% in combination 100. Combination 0×× showed much higher interception probabilities than combination 1××, indicating that a smaller horizontal angle will make the camera intercept with more roots. Interception probabilities in combination ××1 was slightly higher than that of ××0, implying installation of 45° in vertical direction may be more proper for root interception in depth 50 mm.Fig. 11

Probabilities of root interception under different installation patterns. Notes The first number in the combined number represents horizontal angle of microrhizotron to referred plane, where 0 is 45° and 1 is 90°. The second number in the combined number represents radius of microrhizotron to the root center line, where 0 is 30 mm and 1 is 60 mm. The last number in the combined number represents vertical angle of microrhizotron to the root center line, where 0 is 30° and 1 is 45°

From time sequence, averaged interception probabilities increased slightly from 68.7 to 81.2% as days increased. The reason for this was that the development and elongation of roots improved interception with microrhizotron.

It was also obvious in the figure that if installation of microrhizotron were at random, interception probability might be only 73.5%, while presetting according to root architecture features may increase this value to 96.7%. The results proved that interception probability could be improved through proper and right preset of microrhizotrons according to roots architecture features.

As is shown in Tables 2 and 3, when root hair and very fine root were excluded, the relative error was 12.1% between microrhizotron and soil sampling in root length, and 15.4% in diameter. If root hair and very fine root was included, the relative increased high to 93.5% in root length and 48.9% in diameter. This was because most root hair and very fine root were broken and lost during soil sampling. So, in most cases, very fine roots could not be evaluated by soil sampling. However, excluding root hair and very fine root, average error of root length was 2.42 mm. Average root diameter error was 0.05 mm, and it was able to observe fine roots about 0.1 mm in diameter.Table 2

Comparison of root length acquired from microrhizotron and soil sampling (mm)

Additional file 1. Script for root system architecture construction and analysis.